Analysing Nontraditional Students' ChatGPT Interaction, Engagement, Self-Efficacy and Performance: A Mixed-Methods Approach
Saved in:
| Title: | Analysing Nontraditional Students' ChatGPT Interaction, Engagement, Self-Efficacy and Performance: A Mixed-Methods Approach |
|---|---|
| Language: | English |
| Authors: | Mohan Yang (ORCID |
| Source: | British Journal of Educational Technology. 2025 56(5):1973-2000. |
| Availability: | Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us |
| Peer Reviewed: | Y |
| Page Count: | 28 |
| Publication Date: | 2025 |
| Document Type: | Journal Articles Reports - Research |
| Education Level: | Higher Education Postsecondary Education |
| Descriptors: | Nontraditional Students, Artificial Intelligence, Technology Uses in Education, Man Machine Systems, Interaction, Learner Engagement, Self Efficacy, Academic Achievement, Performance, College Students, Predictor Variables, Prompting, Novices, Resistance (Psychology), Computer Attitudes, Adoption (Ideas), Assignments, Digital Literacy |
| DOI: | 10.1111/bjet.13588 |
| ISSN: | 0007-1013 1467-8535 |
| Abstract: | Generative artificial intelligence brings opportunities and unique challenges to nontraditional higher education students, stemming, in part, from the experience of the digital divide. Providing access and practice is critical to bridge this divide and equip students with needed digital competencies. This mixed-methods study investigated how nontraditional higher education students interact with ChatGPT in multiple courses and examined relationships between ChatGPT interactions, engagement, self-efficacy and performance. Data were collected from 73 undergraduate and graduate students through chat logs, course reflections and artefacts, surveys and interviews. ChatGPT interactions were analysed using four metrics: prompt number, depth of knowledge (DoK), prompt relevance and originality. Results showed that ChatGPT prompt numbers ([beta] = 0.256, p < 0.03) and engagement ([beta] = 0.267, p < 0.05) significantly predicted performance, while self-efficacy did not. Students' DoK (r = 0.40, p < 0.01) and prompt relevance (r = 0.42, p < 0.01) were positively correlated with performance. Text mining analysis identified distinct interaction patterns, with 'strategic inquirers' demonstrating significantly higher performance than 'exploratory inquirers' through more sophisticated follow-up questioning. Qualitative findings revealed that while most students were first-time ChatGPT users who initially showed resistance, they developed growing acceptance. Still, students tended to use ChatGPT sparingly and, even then, as only a starting point for assignments. The study highlights the need for targeted guidance in prompt engineering and AI literacy training to help nontraditional higher education students leverage ChatGPT more effectively for higher-order thinking tasks. |
| Abstractor: | As Provided |
| Entry Date: | 2025 |
| Accession Number: | EJ1480022 |
| Database: | ERIC |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwFT7hgHXCdtIzYL8WxGp8QQAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDM0Oeek8oAkZEqPymAIBEICBmmgPpK30iBJ8jKxHeuYfaMDKAkjPnal_VwuB0pfgqwA9h5Bo_KmGKdM0YGzIdq5BdsLunCi-1gMvHwP1MKDN5L794o-D2YMEMnaUl_DLkxG8xPJghZP62_S2yMkOdu7PrTqHnipkMFFSee77Jq1QpklOmJTWtBJhOpZtWU4oLEK9Qt1oWYQcFuh72LZcFCTf_G6C7p3l5Mi7mwY= Text: Availability: 1 Value: <anid>AN0187257453;58i01sep.25;2025Aug14.00:21;v2.2.500</anid> <title id="AN0187257453-1">Analysing nontraditional students' ChatGPT interaction, engagement, self‐efficacy and performance: A mixed‐methods approach </title> <p>Generative artificial intelligence brings opportunities and unique challenges to nontraditional higher education students, stemming, in part, from the experience of the digital divide. Providing access and practice is critical to bridge this divide and equip students with needed digital competencies. This mixed‐methods study investigated how nontraditional higher education students interact with ChatGPT in multiple courses and examined relationships between ChatGPT interactions, engagement, self‐efficacy and performance. Data were collected from 73 undergraduate and graduate students through chat logs, course reflections and artefacts, surveys and interviews. ChatGPT interactions were analysed using four metrics: prompt number, depth of knowledge (DoK), prompt relevance and originality. Results showed that ChatGPT prompt numbers (β = 0.256, p &lt; 0.03) and engagement (β = 0.267, p &lt; 0.05) significantly predicted performance, while self‐efficacy did not. Students' DoK (r = 0.40, p &lt; 0.01) and prompt relevance (r = 0.42, p &lt; 0.01) were positively correlated with performance. Text mining analysis identified distinct interaction patterns, with 'strategic inquirers' demonstrating significantly higher performance than 'exploratory inquirers' through more sophisticated follow‐up questioning. Qualitative findings revealed that while most students were first‐time ChatGPT users who initially showed resistance, they developed growing acceptance. Still, students tended to use ChatGPT sparingly and, even then, as only a starting point for assignments. The study highlights the need for targeted guidance in prompt engineering and AI literacy training to help nontraditional higher education students leverage ChatGPT more effectively for higher‐order thinking tasks. Practitioner notesWhat is already known about this topic Nontraditional students face unique challenges in higher education, such as limited technological literacy and digital access.The emergence of generative AI tools presents both opportunities and challenges for addressing educational disparities.Existing studies on AI implementation predominantly focus on traditional students.What this paper adds Empirical evidence of how nontraditional students interact with ChatGPT through multiple metrics (prompt number, DoK, relevance and originality).Distinct interaction patterns and their relationship to performance outcomes.The relationship among ChatGPT interactions, engagement, self‐efficacy and performance.Implications for practice and/or policy Need for explicit instruction in prompt engineering as a critical skill for higher‐order thinking.Importance of providing targeted technology training and self‐paced learning resources for nontraditional students.Value of developing comprehensive AI literacy training that addresses both tool capabilities and limitations.</p> <p>Keywords: ChatGPT; depth of knowledge; engagement; generative AI; nontraditional students; performance; self‐efficacy</p> <hd id="AN0187257453-2">INTRODUCTION</hd> <p>The landscape of higher education is undergoing a dramatic transformation driven by two concurrent trends: the growing presence of nontraditional students (Dolch &amp; Zawacki‐Richter, [<reflink idref="bib18" id="ref1">18</reflink>]) and the rapid emergence of artificial intelligence (AI) tools, particularly generative AI (GenAI) technologies (Bolick &amp; da Silva, [<reflink idref="bib9" id="ref2">9</reflink>]; Yan et al., [<reflink idref="bib70" id="ref3">70</reflink>]). When taken together, they present a unique opportunity for the field to centre both opportunities and challenges around this rapid advancement, promoting the equitable design, implementation and use of AI within higher education. Equitable human–AI partnerships require that humans of varied backgrounds and identities are included in the design and deployment of the technologies, have agency during the use of the technologies and see positive outcomes that meet their individual and community goals as they use the technologies for learning and life.</p> <p>Providing equal access to and increasing the diversity of the student body to reflect population diversity have been initiatives in higher education (Holmegaard et al., [<reflink idref="bib34" id="ref4">34</reflink>]). Nontraditional students constitute a significant portion of the higher education population, with enrolment continuing to rise even as other subpopulations have remained stable or declined (Dolch &amp; Zawacki‐Richter, [<reflink idref="bib18" id="ref5">18</reflink>]). Among prime‐age adults (25–54), approximately 61 million of them without a bachelor's degree were in the labour force in the United States in 2021, and roughly 4.3 million adults were enrolled in postsecondary education in the same year (Bauer et al., [<reflink idref="bib6" id="ref6">6</reflink>]), indicating a critical gap in expanding higher education access. Nontraditional students are typically defined in the United States as those who are older than 24, employed full‐time, financially independent or have dependents (Bean &amp; Metzner, [<reflink idref="bib7" id="ref7">7</reflink>]; Choy, [<reflink idref="bib13" id="ref8">13</reflink>]; Cleveland‐Innes, [<reflink idref="bib15" id="ref9">15</reflink>]; Jones &amp; Watson, [<reflink idref="bib38" id="ref10">38</reflink>]). In some European countries, however, nontraditional students can refer to those from underrepresented groups, such as minority or low socio‐economic backgrounds (Holmegaard et al., [<reflink idref="bib34" id="ref11">34</reflink>]; Sapir, [<reflink idref="bib56" id="ref12">56</reflink>]). Importantly, unlike traditional students, these demographics often intersect in ways that require complex balancing acts for students simultaneously managing academic, professional and personal lives (Markle, [<reflink idref="bib49" id="ref13">49</reflink>]). These students often face unique challenges, including limited technological literacy and access to digital resources, creating a persistent digital divide that can impede their academic success (Banerjee, [<reflink idref="bib5" id="ref14">5</reflink>]). Mere digital access is not sufficient (Fairlie &amp; London, [<reflink idref="bib22" id="ref15">22</reflink>]). Targeted support and training to empower nontraditional students in bridging the digital divide are necessary (Jesnek, [<reflink idref="bib36" id="ref16">36</reflink>]).</p> <p>The advent of AI tools, particularly ChatGPT, presents both opportunities and challenges for addressing these educational disparities. While these technologies offer potential benefits, such as personalized learning support and automated feedback, they also risk widening the existing digital divide if not thoughtfully implemented. Exposing nontraditional students to AI tools could help bridge this gap by providing accessible technological support for their learning needs (Banerjee, [<reflink idref="bib5" id="ref17">5</reflink>]). However, understanding how these students interact with such tools and the subsequent impact on their learning outcomes remains crucial for curriculum design, effective implementation to develop their digital literacy, a critical skill in today's workforce. Yet, extant research on digital technology usage, especially GenAI tools, primarily focuses on traditional students, leaving a gap in understanding how nontraditional students interact with these technologies. While existing studies have shed light on the digital exclusion faced by underrepresented students (Kuo, [<reflink idref="bib41" id="ref18">41</reflink>]), exploring how AI can be leveraged to contribute to student empowerment is a more strengths‐based approach to examining the interactions of marginalized learners. Learner agency is intentionally centred (Sulecio de Alvarez &amp; Dickson‐Deane, [<reflink idref="bib62" id="ref19">62</reflink>]), reconceptualizing human–AI interaction as transformative instead of transactional (Lombard, [<reflink idref="bib46" id="ref20">46</reflink>]).</p> <p>Despite the growing adoption of AI in education, several critical gaps persist in our understanding. First, there is limited empirical evidence regarding the impact of these tools on learning and instructional design processes, particularly for nontraditional students. We lack comprehensive metrics for evaluating learner‐AI interactions, making it difficult to assess the effectiveness of these tools in supporting student learning. Third, the intricacies of human–AI interaction patterns among students remain poorly understood. These gaps hinder our ability to support nontraditional learners in navigating 'jagged technological frontier' (Dell'Acqua et al., [<reflink idref="bib17" id="ref21">17</reflink>]). In responding to these gaps, we adopted GenAI tools in the target courses within which the majority of students identified themselves as nontraditional students. Through 2023 and 2024, we integrated GenAI tools, such as ChatGPT in multiple instructional design and human resource development courses to help students explore the technological integration in learning and design process as an effort for developing their digital awareness and competencies. The purpose of this study is to address the mentioned gaps by examining how nontraditional undergraduate and graduate students interact with ChatGPT and the relationships among their ChatGPT interaction, engagement, self‐efficacy and performance for a holistic picture of students' ChatGPT interaction. Specifically, we evaluate their interactions through multiple metrics, including the number of prompts, depth of knowledge in questioning, prompt relevance and evolution, and originality. By analysing these interactions and their impact on students' instructional design and writing performance, we aim to develop a more nuanced understanding of how GenAI tools can be effectively leveraged to support nontraditional students' academic success.</p> <hd id="AN0187257453-3">LITERATURE REVIEW</hd> <p></p> <hd id="AN0187257453-4">Nontraditional learners in the AI age</hd> <p>Nontraditional students differ from the conventional undergraduate archetype—18–22 year‐old, full‐time enrolled, parent‐dependent students pursuing immediate postsecondary education. Bean and Metzner's ([<reflink idref="bib7" id="ref22">7</reflink>]) foundational framework emphasizes age (&gt;24), non‐residential status and part‐time enrolment as primary identifiers, which was expanded with new identifiers added including first‐generation status and non‐standard high school credentials (The National Postsecondary Student Aid Study, [<reflink idref="bib51" id="ref23">51</reflink>]) and delayed enrolment, employment, financial independence and caretaking responsibilities (Horn et al., [<reflink idref="bib35" id="ref24">35</reflink>]). Increasingly, nontraditional populations are seeking career advancement or personal enrichment (Cavazos et al., [<reflink idref="bib10" id="ref25">10</reflink>]; Dolch &amp; Zawacki‐Richter, [<reflink idref="bib18" id="ref26">18</reflink>]; Shepherd &amp; Sheu, [<reflink idref="bib58" id="ref27">58</reflink>]). Yet, the complex time management in balancing academic, professional and personal responsibilities poses challenges for these students (Markle, [<reflink idref="bib49" id="ref28">49</reflink>]).</p> <p>In supporting nontraditional students' needs, AI applications in higher education span four main areas: profiling and prediction, intelligent tutoring systems, assessment and evaluation, and adaptive systems and personalization (Zawacki‐Richter et al., [<reflink idref="bib72" id="ref29">72</reflink>]). In instructional design specifically, Bolick and da Silva ([<reflink idref="bib9" id="ref30">9</reflink>]) highlighted how AI tools can enhance workflows through content creation, outlining, storyboarding, scenario generation and personalization. For nontraditional students, these AI‐assisted capabilities could significantly streamline their workflow, allowing focus on higher‐order aspects of design. However, effectively leveraging these tools requires navigating the 'jagged technological frontier', where AI capabilities vary significantly across tasks. Understanding these human–AI interaction styles is crucial for supporting nontraditional students, particularly given the digital divide that encompasses not just technology access but also the skills required to use it effectively (Banerjee, [<reflink idref="bib5" id="ref31">5</reflink>]).</p> <hd id="AN0187257453-5">Large language models (LLM) in education</hd> <p>The complexity of AI integration is particularly evident in LLMs interactions. While Kumar et al. ([<reflink idref="bib40" id="ref32">40</reflink>]) found that GenAI assistance can build system trust and improve performance, Zamfirescu‐Pereira et al. ([<reflink idref="bib71" id="ref33">71</reflink>]) revealed significant challenges non‐experts face in prompt design, including over‐generalization and misapplied interaction expectations, potentially exacerbating rather than alleviating nontraditional students' educational barriers. Dell'Acqua et al. ([<reflink idref="bib17" id="ref34">17</reflink>]) identified two successful AI use patterns: 'Centaurs', who strategically delegate tasks, and 'Cyborgs', who deeply integrate AI into their workflow. Similarly, Nguyen et al. ([<reflink idref="bib52" id="ref35">52</reflink>]) found that 'Structured Adaptivity' approaches yield better outcomes than 'Unstructured Streamline' methods. However, achieving these optimal interaction patterns requires specific guidance and support that many nontraditional students may lack. While example‐based instruction (Wittwer &amp; Renkl, [<reflink idref="bib68" id="ref36">68</reflink>]) and metacognitive strategies (Alt &amp; Raichel, [<reflink idref="bib2" id="ref37">2</reflink>]) show potential for supporting AI tool use, understanding how to effectively adapt these approaches for nontraditional students requires dedicated research.</p> <p>Studies have also examined prompt frequency and complexity (White et al., [<reflink idref="bib67" id="ref38">67</reflink>]). The Human‐AI Language‐based Interaction Evaluation (HALIE) framework evaluates human‐LLM interaction across three dimensions: targets (process vs. output), perspectives (first‐person vs. third‐party) and criteria (quality vs. preference; Lee et al., [<reflink idref="bib42" id="ref39">42</reflink>]). HALIE was validated across diverse tasks, demonstrating that interaction‐based evaluation can reveal insights not captured by traditional non‐interactive benchmarks.</p> <p>Student engagement with educational technology encompasses behavioural, emotional and cognitive dimensions (Fredricks &amp; McColskey, [<reflink idref="bib24" id="ref40">24</reflink>]; Henrie et al., [<reflink idref="bib33" id="ref41">33</reflink>]). In the context of LLM interaction, emotional engagement reflects students' affective reactions, including interest, enjoyment and anxiety when using AI tools (Uddin et al., [<reflink idref="bib65" id="ref42">65</reflink>]). Behavioural engagement manifests through observable participation patterns, such as the frequency and consistency of LLM use (Lo et al., [<reflink idref="bib45" id="ref43">45</reflink>]), while cognitive engagement involves the psychological investment in learning and self‐regulation strategies (Escalante et al., [<reflink idref="bib20" id="ref44">20</reflink>]; Zhang &amp; Tur, [<reflink idref="bib74" id="ref45">74</reflink>]). HALIE, similarly, delineates engagement into emotional dimensions (student confidence and comfort in AI tool usage), cognitive aspects (depth of understanding and learning strategy development) and behavioural patterns (interaction frequency and consistency). The relationship between these engagement patterns and learning outcomes remains particularly unexplored for nontraditional students, despite evidence suggesting that interaction quality may be more important than quantity for learning outcomes (Lee et al., [<reflink idref="bib42" id="ref46">42</reflink>]).</p> <hd id="AN0187257453-6">Depth of knowledge and prompt relevance</hd> <p>Webb's Depth of Knowledge (DoK; Webb, [<reflink idref="bib66" id="ref47">66</reflink>]) analyses learning activities complexity and guides educators in designing assessments (Harris &amp; Patten, [<reflink idref="bib26" id="ref48">26</reflink>]; Hess et al., [<reflink idref="bib27" id="ref49">27</reflink>]), developing learning activities (Barber, [<reflink idref="bib4" id="ref50">4</reflink>]; Francis, [<reflink idref="bib21" id="ref51">21</reflink>]) and aligning curriculum with learning objectives (Traynor et al., [<reflink idref="bib64" id="ref52">64</reflink>]; Wyse &amp; Viger, [<reflink idref="bib69" id="ref53">69</reflink>]) including curriculum designed specifically for neurodivergent learners (Morgan et al., [<reflink idref="bib50" id="ref54">50</reflink>]). Its applications now extend to analysing learner engagement with AI systems, where prompts reflect the cognitive complexity of student interactions with AI systems. The sophistication of prompts can range from basic information retrieval to complex analytical queries, indicating different levels of cognitive processing and understanding (see Figure 1 and Appendix A).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0001.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0001.jpg" title="1 Webb's DOK levels." /> </p> <p></p> <p>Prompt relevance measures how effectively students' queries relate to and build upon each other. Li et al. ([<reflink idref="bib43" id="ref55">43</reflink>]) found that successful learners craft sequences of interconnected prompts that progressively develop their understanding. Prompt evolution examines how students' interaction patterns with AI systems change over time. Students who progressively refine their prompts show deeper engagement with the material and better learning outcomes (Chang et al., [<reflink idref="bib12" id="ref56">12</reflink>]). This evolutionary process often involves students moving from simple, direct questions to more nuanced and contextual queries that demonstrate increasing mastery of both the subject matter and the AI tool itself.</p> <hd id="AN0187257453-8">RESEARCH QUESTIONS</hd> <p>To bridge the gaps as mentioned above, we seek to explore students' ChatGPT interactions, engagement, self‐efficacy and performance with the following guiding research questions (RQ):</p> <p></p> <ulist> <item> How did students perform in interacting with ChatGPT to facilitate their learning and design process?</item> <p></p> <item> How were students' ChatGPT interactions, engagement, self‐efficacy and learning performance related to each other?</item> <p></p> <item> What were students' experiences in using ChatGPT to facilitate their learning and design?</item> </ulist> <hd id="AN0187257453-9">METHODOLOGY</hd> <p></p> <hd id="AN0187257453-10">Research design</hd> <p>To answer the research questions, we conducted a sequential explanatory mixed‐methods study (Creswell &amp; Clark, [<reflink idref="bib16" id="ref57">16</reflink>]) with text mining techniques applied to extract the patterns of students' ChatGPT interaction (see Figure 2).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0002.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0002.jpg" title="2 Study process." /> </p> <p></p> <hd id="AN0187257453-12">Context and participants</hd> <p>This study was conducted at a mid‐Atlantic, research‐intensive, minority‐serving public university in the United States. Approximately 38% of students are from underrepresented ethnic groups, over half of the undergraduate students receive financial aid, one‐third of students are enrolled part‐time, and approximately one‐third of freshmen are first‐generation students. From 2023 to 2024, GenAI was integrated into instructional design projects in two undergraduate and one graduate course with a total of 75 students in the areas of technology, instructional design and human resource development. According to students' demographic information, between 78% and 80% of students (missing data from five students) in those classes were nontraditional students that met at least one of the characteristics of the traditional definition, such as age, job or parenting. When underrepresented students were considered as nontraditional students as well, the rate could reach approximately 90% or higher based on students' demographics, but we were unable to verify the specific number for this study due to missing data. Prior to course activities, students were supported with a ChatGPT guide and video demos. Data collection included end‐of‐course surveys yielding 73 complete matched projects for quantitative analysis and 82 projects/chat logs for text mining analysis.</p> <hd id="AN0187257453-13">Data collection</hd> <p>Data were collected at multiple stages and from multiple sources (see Table 1).</p> <p>1 TABLE Data sources.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Data sources&lt;/th&gt;&lt;th align="left"&gt;Description&lt;/th&gt;&lt;th align="left"&gt;RQs&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;ChatGPT logs&lt;/td&gt;&lt;td align="left"&gt;As part of the assignments, all students were required to use ChatGPT towards their designs following specific instructions (eg, when and how to use ChatGPT). During classes, students needed to record their' interaction with ChatGPT are recorded and submit it including their prompts and ChatGPT outputs and submit it either as a link or document along with their projects. For the purposes of the course assignments, the grade of ChatGPT interactions was weighted equally with the designed project to ensure students' effort in using the tool&lt;/td&gt;&lt;td align="left"&gt;1 &amp; 2&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Class artefacts&lt;/td&gt;&lt;td align="left"&gt;During classes, students submitted their design or writings that were completed with the facilitation of ChatGPT. Artefacts were examined and evaluated along with their ChatGPT interactions&lt;/td&gt;&lt;td align="left"&gt;2&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Reflection &amp; discussion&lt;/td&gt;&lt;td align="left"&gt;Students reflected on their usage of ChatGPT and discussed the role of technology at different points of the courses. Reflection or discussion prompts were provided&lt;/td&gt;&lt;td align="left"&gt;2 &amp; 3&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Survey&lt;/td&gt;&lt;td align="left"&gt;A one&amp;#8208;time postsurvey on students' perceived self&amp;#8208;efficacy and engagement was conducted following a 5&amp;#8208;Likert scale at the end of the semester. Open&amp;#8208;ended questions were also added to collect students' experience and insights of using the tool in the courses&lt;/td&gt;&lt;td align="left"&gt;2&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Interview/focus group&lt;/td&gt;&lt;td align="left"&gt;A group of four students from each class was formed for a focus group focusing on their experience using ChatGPT. Two individual interviews were conducted in one of the courses due to scheduling challenges. Each interview/focus group lasted approximately 60&amp;#8201;minutes&lt;/td&gt;&lt;td align="left"&gt;1, 2, &amp; 3&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd id="AN0187257453-14">Research variables and measurement</hd> <p>We adapted HALIE (Lee et al., [<reflink idref="bib42" id="ref58">42</reflink>]) as our theoretical framework to guide the development of our metrics in evaluating students' ChatGPT interactions (see Figure 3). While originally designed for evaluating language models, HALIE's structure and focus on interaction processes, targets (process vs. output), perspectives (first‐person vs. third‐party) and criteria (quality vs. preference) makes it equally valuable for studying user behaviour, adaptation patterns and experience. The framework provides researchers with a systematic methodology for collecting and analysing both quantitative and qualitative data about how users interact with AI systems over time. It aligns with our investigation of interaction characteristics, engagement patterns and performance indicators.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0003.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0003.jpg" title="3 Research theoretical framework. Source: Adapted from Lee et al. ([42])." /> </p> <p></p> <p>Table 2 illustrates how HALIE's three‐dimensional structure guided our assessment development.</p> <p>2 TABLE Mapping HALIE framework dimensions to assessment indicators.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;HALIE dimension&lt;/th&gt;&lt;th align="left"&gt;Component&lt;/th&gt;&lt;th align="left"&gt;Assessment indicators&lt;/th&gt;&lt;th align="left"&gt;Measurement approach&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;Targets&lt;/td&gt;&lt;td align="left"&gt;Process&lt;/td&gt;&lt;td align="left"&gt;Prompt number&lt;/td&gt;&lt;td align="left"&gt;Count of student&amp;#8208;initiated prompts per project&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Depth of Knowledge (DoK)&lt;/td&gt;&lt;td align="left"&gt;Four&amp;#8208;level rubric assessing cognitive complexity&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Prompt relevance&lt;/td&gt;&lt;td align="left"&gt;Four&amp;#8208;level rubric evaluating query progression&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Output&lt;/td&gt;&lt;td align="left"&gt;Originality&lt;/td&gt;&lt;td align="left"&gt;Four&amp;#8208;level rubric measuring personal contribution&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Performance score&lt;/td&gt;&lt;td align="left"&gt;Rubric&amp;#8208;based assessment by external evaluator&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Perspectives&lt;/td&gt;&lt;td align="left"&gt;First&amp;#8208;person&lt;/td&gt;&lt;td align="left"&gt;Self&amp;#8208;efficacy&lt;/td&gt;&lt;td align="left"&gt;Five&amp;#8208;item scale (Avey et&amp;#160;al.,&amp;#160;&lt;xref ref-type="bibr" rid="bibr3"&gt;2009&lt;/xref&gt;)&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Emotional engagement&lt;/td&gt;&lt;td align="left"&gt;Two&amp;#8208;item scale (Skinner et&amp;#160;al.,&amp;#160;&lt;xref ref-type="bibr" rid="bibr59"&gt;2009&lt;/xref&gt;)&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Student reflections&lt;/td&gt;&lt;td align="left"&gt;Qualitative analysis of reflections/discussions&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Third&amp;#8208;party&lt;/td&gt;&lt;td align="left"&gt;Expert evaluation&lt;/td&gt;&lt;td align="left"&gt;Independent assessment by two researchers&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Interaction analysis&lt;/td&gt;&lt;td align="left"&gt;Text mining of query patterns&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Criteria&lt;/td&gt;&lt;td align="left"&gt;Quality&lt;/td&gt;&lt;td align="left"&gt;DoK levels&lt;/td&gt;&lt;td align="left"&gt;Evaluation of cognitive engagement&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Prompt relevance&lt;/td&gt;&lt;td align="left"&gt;Assessment of strategic development&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Output originality&lt;/td&gt;&lt;td align="left"&gt;Evaluation of creative transformation&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Preference&lt;/td&gt;&lt;td align="left"&gt;Engagement measures&lt;/td&gt;&lt;td align="left"&gt;Combined emotional/cognitive/behavioural scores&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;User experience&lt;/td&gt;&lt;td align="left"&gt;Analysis of qualitative feedback&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd id="AN0187257453-16">Learning performance</hd> <p> <emph>Students' performance was measured by evaluating their course project against corresponding pre‐established rubrics</emph>. To avoid the potential bias from the course instructor, one co‐author conducted the evaluation specifically for the research. All performance scores were converted into percentages first and then transformed into z scores to ensure suitability in the regression model.</p> <hd id="AN0187257453-17">ChatGPT interaction</hd> <p>In measuring ChatGPT interaction, four metrics as four separate variables were developed, including (<reflink idref="bib1" id="ref59">1</reflink>) prompt number, (<reflink idref="bib2" id="ref60">2</reflink>) the depth of knowledge, (<reflink idref="bib3" id="ref61">3</reflink>) the prompt relevance and evolution, and (<reflink idref="bib4" id="ref62">4</reflink>) the originality of work. Prompt number refers to the number of prompts for a corresponding project. Two authors independently examined all chat logs following the DoK and prompt relevance metrics. A final score was assigned after consensus was reached for each project. DoK level indicates a student's cognitive engagement and reasoning in questioning ChatGPT, examined as part of the quality criteria. Analysing quality in human–AI interactions provides insight into students' cognitive engagement and learning development (Lee et al., [<reflink idref="bib42" id="ref63">42</reflink>]). The prompt relevance refers to how prompts are related and evolve as students interact with ChatGPT. As ChatGPT interaction is a required part of the project development, the work originality becomes an essential aspect of understanding the learning and design process, which was evaluated by comparing students' work and the ChatGPT logs. All metrics can be seen in Appendix A.</p> <hd id="AN0187257453-18">Self‐efficacy</hd> <p>Self‐efficacy of using ChatGPT was measured with items adapted from Avey et al. ([<reflink idref="bib3" id="ref64">3</reflink>]). Overall, five items were used.</p> <hd id="AN0187257453-19">Engagement</hd> <p>Engagement has been measured from different perspectives in different contexts. We measured three dimensions, including emotional, cognitive and behavioural. We measured emotional engagement, a dimension that has not been addressed comprehensively in technology‐enhanced learning (Halverson &amp; Graham, [<reflink idref="bib31" id="ref65">31</reflink>]) with two items adapted from Skinner et al. ([<reflink idref="bib59" id="ref66">59</reflink>]). Cognitive engagement focused on the perceived value, persistence and critical thinking effort (Fredricks et al., [<reflink idref="bib23" id="ref67">23</reflink>]; Halverson &amp; Graham, [<reflink idref="bib31" id="ref68">31</reflink>]; Reschly &amp; Christenson, [<reflink idref="bib54" id="ref69">54</reflink>]). Behavioural engagement measures students' participation and concentration with items adapted from Skinner et al. ([<reflink idref="bib59" id="ref70">59</reflink>]). The engagement variable composite score calculated by summing students' self‐reported scores on emotional, cognitive and behavioural engagement items.</p> <hd id="AN0187257453-20">Data analysis</hd> <p></p> <hd id="AN0187257453-21">Quantitative strand</hd> <p>A multiple regression analysis was conducted with performance as the dependent variable (DV) and prompt number, engagement and self‐efficacy as the independent variables (IVs) to investigate their influence on students' ChatGPT‐facilitated performance score. The DoK, prompt relevance and originality variables violated one or more assumptions of the multiple regression technique (eg, linearity and collinearity), so we opted instead to examine the relationships between these variables and performance using Pearson product moment correlations. Additionally, a correlation analysis was conducted among the ChatGPT interactions, engagement and self‐efficacy.</p> <hd id="AN0187257453-22">Qualitative strand</hd> <p>The qualitative analysis involves both deductive and inductive coding and thematic analysis on semi‐structured interview/focus group, students' in‐class reflection/discussion and their responses to the open‐ended survey questions. Two coders followed a deductive approach and coded the data independently, followed by meetings to finalize codes until a consensus was reached. A third coder followed an inductive approach independently. The codes were merged for reconfiguration and categorization.</p> <hd id="AN0187257453-23">Text mining strand</hd> <p>We employed a mixed‐methods approach to analyse their query patterns and their relationship to performance. In this study, we use query and prompt interchangeably. Using the facebook/bart‐large‐mnli model, a state‐of‐the‐art zero‐shot classification model that is ideal for handling the diverse and naturally occurring queries generated by students (Champa et al., [<reflink idref="bib11" id="ref71">11</reflink>]), we classified the queries into five categories: Information Request (seeking specific factual answers), Clarification (resolving ambiguities), Explanation (understanding or reasoning), Follow‐Up (expanding upon previous queries) and Query (broad or unspecific questions). The high semantic understanding of the model guaranteed accuracy in capturing the intent behind each student query. We then generated query sequences, representing the chronological progression of their interactions with ChatGPT (eg, Information Request and Clarification). These sequences captured the dynamic nature of student interactions and allowed for an analysis of their query progression patterns. Hierarchical clustering was applied to the query sequences to identify distinct interaction patterns (Jiang et al., [<reflink idref="bib37" id="ref72">37</reflink>]). A dendrogram analysis helped to determine the optimal cluster number, with each one representing a unique interaction strategy. We visualized the query clusters using D3.js for intuitive interpretation.</p> <p>To examine the link between clusters and performance outcomes, we conducted a one‐way ANOVA. We selected a representative student from each cluster based on prompt frequency, performance and students' reflection, to illustrate the query patterns for a more nuanced understanding of the differences. The thematic analysis on reflections (Clarke &amp; Braun, [<reflink idref="bib14" id="ref73">14</reflink>]) provided additional insights and contextualized the interaction patterns and their impact on performance. This combined quantitative and qualitative approach offered a comprehensive understanding of how query behaviours influenced learning performance.</p> <hd id="AN0187257453-24">RESULTS</hd> <p></p> <hd id="AN0187257453-25">Quantitative strand</hd> <p>The descriptive statistics revealed that the mean performance score, DoK and prompt relevance mean scores were relatively low, while the originality had a higher mean score (see Table 3). The engagement and self‐efficacy mean scores indicated that students were engaged in the projects and had a relatively high self‐efficacy level.</p> <p>3 TABLE Descriptive statistics for performance score and ChatGPT interaction variables.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left" /&gt;&lt;th align="left"&gt;&lt;italic&gt;N&lt;/italic&gt;&lt;/th&gt;&lt;th align="left"&gt;Minimum&lt;/th&gt;&lt;th align="left"&gt;Maximum&lt;/th&gt;&lt;th align="left"&gt;Mean&lt;/th&gt;&lt;th align="left"&gt;SD&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;Performance (original percentage score)&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;34.38&lt;/td&gt;&lt;td align="char" char="."&gt;100.00&lt;/td&gt;&lt;td align="char" char="."&gt;67.76&lt;/td&gt;&lt;td align="char" char="."&gt;18.34&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Standardized performance&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;1.82&lt;/td&gt;&lt;td align="char" char="."&gt;1.76&lt;/td&gt;&lt;td align="char" char="."&gt;0.00&lt;/td&gt;&lt;td align="char" char="."&gt;1.00&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;ChatGPT prompt number&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="char" char="."&gt;34&lt;/td&gt;&lt;td align="char" char="."&gt;7.03&lt;/td&gt;&lt;td align="char" char="."&gt;7.04&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Depth of knowledge&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="char" char="."&gt;4&lt;/td&gt;&lt;td align="char" char="."&gt;1.90&lt;/td&gt;&lt;td align="char" char="."&gt;1.00&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Prompt relevance&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="char" char="."&gt;4&lt;/td&gt;&lt;td align="char" char="."&gt;1.99&lt;/td&gt;&lt;td align="char" char="."&gt;1.09&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Originality&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="char" char="."&gt;4&lt;/td&gt;&lt;td align="char" char="."&gt;2.79&lt;/td&gt;&lt;td align="char" char="."&gt;1.12&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Engagement&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;9&lt;/td&gt;&lt;td align="char" char="."&gt;35&lt;/td&gt;&lt;td align="char" char="."&gt;27.77&lt;/td&gt;&lt;td align="char" char="."&gt;6.12&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Self&amp;#8208;efficacy&lt;/td&gt;&lt;td align="char" char="."&gt;73&lt;/td&gt;&lt;td align="char" char="."&gt;10&lt;/td&gt;&lt;td align="char" char="."&gt;25&lt;/td&gt;&lt;td align="char" char="."&gt;20.30&lt;/td&gt;&lt;td align="char" char="."&gt;3.33&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <p>A total of 73 cases had scores for all three variables and were included in the multiple regression analysis. The findings revealed that the three predictors provided a good model fit, explaining a statistically significant proportion of variance (<emph>R</emph><sups>2</sups> = 0.17, <emph>F</emph> = 4.70 (<reflink idref="bib3" id="ref74">3</reflink>, 69), <emph>p</emph> &lt; 0.01). Results indicated that prompt number and engagement were both statistically significant positive predictors of learning performance, but students' self‐efficacy scores did not significantly influence performance (see Table 4).</p> <p>4 TABLE Influence of interaction, engagement and self‐efficacy of ChatGPT on performance.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Independent variable&lt;/th&gt;&lt;th align="left"&gt;Unstandardized coefficients&lt;/th&gt;&lt;th align="left"&gt;&lt;italic&gt;&amp;#946;&lt;/italic&gt;&lt;/th&gt;&lt;th align="left"&gt;&lt;italic&gt;p&lt;/italic&gt;&lt;/th&gt;&lt;/tr&gt;&lt;tr&gt;&lt;th align="left"&gt;&lt;italic&gt;B&lt;/italic&gt;&lt;/th&gt;&lt;th align="left"&gt;SE&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;ChatGPT prompt number&lt;/td&gt;&lt;td align="char" char="."&gt;0.036&lt;/td&gt;&lt;td align="char" char="."&gt;0.016&lt;/td&gt;&lt;td align="char" char="."&gt;0.256&lt;/td&gt;&lt;td align="char" char="."&gt;0.03&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Engagement&lt;/td&gt;&lt;td align="char" char="."&gt;0.044&lt;/td&gt;&lt;td align="char" char="."&gt;0.021&lt;/td&gt;&lt;td align="char" char="."&gt;0.267&lt;/td&gt;&lt;td align="char" char="."&gt;0.05&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Self&amp;#8208;efficacy&lt;/td&gt;&lt;td align="char" char="."&gt;0.026&lt;/td&gt;&lt;td align="char" char="."&gt;0.039&lt;/td&gt;&lt;td align="char" char="."&gt;0.086&lt;/td&gt;&lt;td align="char" char="."&gt;0.51&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <p>Prior to examining correlations among participants' learning performance, number of prompts, DoK, prompt relevance, engagement, self‐efficacy and originality scores, we examined distributions for these variables to determine normality. Skewness for each of the variables was within the acceptable range of ±1.00, so we proceeded with the Pearson correlation analyses. To account for multiple comparisons between learning performance, DoK, prompt relevance and originality, we used an adjusted p value of 0.02 (3 comparisons divided by 0.05), and to account for multiple comparisons between ChatGPT interaction variables, we used an adjusted p value of 0.003, calculated using the conservative Bonferroni approach (15 comparisons divided by 0.05). We found statistically significant positive correlations between learning performance and students' DoK (<emph>r</emph> = 0.40, <emph>p</emph> &lt; 0.02) and prompt relevance (<emph>r</emph> = 0.42, <emph>p</emph> &lt; 0.02), with a non‐significant negative correlation observed between performance and originality (<emph>r</emph> = −0.13, <emph>p</emph> = 0.26). The findings of correlations between students' ChatGPT interactions are shown in Table 5. Prompt number is significantly correlated with DoK and prompt relevance. DoK is significantly correlated with prompt relevance. Originality is significantly negatively correlated with DoK, prompt relevance and self‐efficacy. Self‐efficacy is significantly correlated with engagement. Figure 4 illustrated the relationships between the variables.</p> <p>5 TABLE Correlations between ChatGPT interaction variables.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left" /&gt;&lt;th align="left"&gt;Prompt number&lt;/th&gt;&lt;th align="left"&gt;DoK&lt;/th&gt;&lt;th align="left"&gt;Prompt relevance&lt;/th&gt;&lt;th align="left"&gt;Originality&lt;/th&gt;&lt;th align="left"&gt;Engagement&lt;/th&gt;&lt;th align="left"&gt;Self&amp;#8208;efficacy&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;Prompt number&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;DoK&lt;/td&gt;&lt;td align="char" char="."&gt;0.747&lt;xref ref-type="fn" rid="tfn1" /&gt;&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Prompt relevance&lt;/td&gt;&lt;td align="char" char="."&gt;0.738&lt;xref ref-type="fn" rid="tfn1" /&gt;&lt;/td&gt;&lt;td align="char" char="."&gt;0.917&lt;xref ref-type="fn" rid="tfn1" /&gt;&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Originality&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.204&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.291&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.277&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="left" /&gt;&lt;td align="left" /&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Engagement&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.036&lt;/td&gt;&lt;td align="char" char="."&gt;0.175&lt;/td&gt;&lt;td align="char" char="."&gt;0.140&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.192&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;td align="left" /&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Self&amp;#8208;efficacy&lt;/td&gt;&lt;td align="char" char="."&gt;0.129&lt;/td&gt;&lt;td align="char" char="."&gt;0.221&lt;/td&gt;&lt;td align="char" char="."&gt;0.162&lt;/td&gt;&lt;td align="char" char="."&gt;&amp;#8722;0.237&lt;/td&gt;&lt;td align="char" char="."&gt;0.531&lt;xref ref-type="fn" rid="tfn1" /&gt;&lt;/td&gt;&lt;td align="char" char="."&gt;1&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <p>1 ** Correlation is significant at the 0.003 level (2‐tailed).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0004.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0004.jpg" title="4 Relationships between the variables. ** Correlation is significant at the 0.003 level (2‐tailed)." /> </p> <p></p> <hd id="AN0187257453-27">Text mining strand</hd> <p>Our analysis identified three distinct clusters of query patterns, highlighting diverse approaches to interacting with ChatGPT, with 11 students in Cluster 1, 16 students in Cluster 2 and 48 students in Cluster 3 (Figure 5). The additional seven students fell above the threshold defined by the horizontal line in the hierarchical clustering dendrogram and formed an outlier group. A one‐way ANOVA revealed a significant difference in learning performance between at least two groups (<emph>F</emph>(<reflink idref="bib3" id="ref75">3</reflink>, 78) = 3.451, <emph>p</emph> = 0.02), with Tukey's HSD test showing a significantly different result between Clusters 1 and 3 (<emph>p</emph> = 0.025. 95% CI = 1.513, 31.93). Despite the different sample sizes between groups, the homogeneity of variance was not violated. Below, we described Clusters 1 and 3 in greater detail and used representative students to further demonstrate the differences.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0005.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0005.jpg" title="5 Hierarchical clustering dendrogram of student query patterns. The dendrogram visualizes the clustering of students based on the similarity of their query sequences when interacting with ChatGPT. Each number in the dendrogram represents a unique ID. The horizontal line indicates the threshold used to identify three distinct clusters." /> </p> <p></p> <hd id="AN0187257453-29">Cluster 1: Strategic inquirers</hd> <p>Cluster 1 was characterized by structured and purposeful query sequences, often transitioning from such as Explanation to repeated Follow‐Up queries (Figure 6, left). These 'strategic inquirers' adopted a deliberate approach to iteratively refine and expand their understanding by systematically building on previous responses. It also demonstrated their metacognitive awareness of their learning needs and a strategic use of ChatGPT to meet those needs. After obtaining initial information, they often posed follow‐up questions to clarify nuances, requested additional context or probed deeper into the topic. This iterative engagement effectively created a dialogue with the AI, supporting their ability to construct deeper and more cohesive understandings. Furthermore, these students demonstrated a goal‐oriented manner, shaped by clear objectives. This behaviour aligns with theories of self‐regulated learning, where learners actively monitored and adapted their strategies to achieve specific outcomes. The structured nature of their query sequences implied students' purposeful scaffolding learning processes, which could have contributed to better performance.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0006.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0006.jpg" title="6 Visualization of query patterns for Cluster 1 (left) and Tray (pseudonym; right). Weighted arcs indicate the transition frequency. Top arcs represent left‐to‐right transitions, while bottom arcs are opposite." /> </p> <p></p> <p>Tray (performance score: 100; 11 queries; DoK: 4; prompt relevance: 4; originality: 2; engagement: 32 out of 35; self‐efficacy: 24 out of 25.) displayed a structured and purposeful approach to engaging with ChatGPT (Figure 6, right). He began with a broad Explanation query that established the overarching topic of interest for a more targeted exploration. Following this, Tray made an Information Request to gather foundational knowledge about the subject, anchoring the broader topic within a historical context. Building on this, Tray transitioned into a series of focused Follow‐Up queries, signalling a clear goal to develop a comprehensive lesson plan. As the interaction progressed, Tray maintained focus and dived into detailed aspects (eg, reasonably priced materials and specific features of the target objects). Tray's query sequence reflected a strategic use of ChatGPT to scaffold learning objectives, systematically building a robust understanding of both the process of the topic and the educational framework for teaching it. The repeated Follow‐Up queries revealed a metacognitive approach, with Tray refining goals and clarifying details to achieve a deeper understanding. This culminated in an Explanation query that synthesized prior insights into a format suitable for a specific instructional context. In the survey and reflection, Tray mentioned his experience with ChatGPT at his job and enjoyment using ChatGPT in this course. To him, the best way is to 'act as if I am talking to another person and tell it exactly what I am looking for'. Even though he acknowledged that ChatGPT 'takes away originality', he used 'the information it gives me to form my own ideas' since 'it provided ideas to the prompt I did not think of before'.</p> <hd id="AN0187257453-31">Cluster 3: Exploratory inquirers</hd> <p>The largest group, Cluster 3 or exploratory inquirers, exhibited less structured and more exploratory query behaviours, characterized by shorter queries and a predominance of Explanation queries (Figure 7, left). Unlike the strategic inquirers, the exploratory inquirers relied heavily on instructor‐provided guidelines to frame their interactions with ChatGPT. These assignment guidelines were often directly incorporated into the students' queries. While this approach helped students stay focused on task requirements, it also constrained the breadth of their inquiry, reducing opportunities for deeper, self‐directed exploration. The predominance of Explanation queries further underscores this reliance on external scaffolding, as students typically sought detailed responses to specific assignment components rather than engaging in iterative questioning or refining their understanding. This reactive approach to interacting with ChatGPT may reflect a lack of confidence or familiarity with using AI tools independently. By adhering closely to the provided structure, these students prioritized completing the task over fostering a broader understanding or exploring creative solutions.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01sep25/bjet13588-fig-0007.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13588-fig-0007.jpg" title="7 Visualization of query patterns for Cluster 3 (left) and Bella (pseudonym; right)." /> </p> <p></p> <p>Bella (performance score, 70.83; 8 queries/prompts; DoK: 1; prompt relevance: 2; originality: 1; engagement: 34 out of 35; self‐efficacy: 25 out of 25) relied heavily on instructor‐provided guidelines to shape queries and maintain alignment with task requirements (Figure 7, right). Bella began with an Information Request. Throughout the process, she continued to use Information Request query or Explanation query to directly address key elements of the instructor guidelines, such as asking for 'the goal of the learning module' and 'plan for evaluation'. Overall, Bella's queries were tightly aligned with the instructor's framework, with little evidence of refinement or expansion beyond the prescribed elements of the project. This case highlights a key characteristic of exploratory inquirers: the tendency to use AI tools primarily as a means of reproducing externally defined structures rather than as a resource for creative problem solving or independent knowledge construction. Bella revealed in her survey and reflection that she had been using ChatGPT and 'feel more comfortable with using it', which might explain her self‐efficacy level. To her, ChatGPT 'allowed me to be introduced to new ideas and use the ideas to implement into my work'. Yet, she included portions of ChatGPT outputs without sufficient modifications.</p> <hd id="AN0187257453-33">Diverse explorers and intensive explorers</hd> <p>Cluster 2, diverse explorers, features a more varied mix of query types. While students in Cluster 2 still engage in iterative learning, their inquiries tend to be broader or less targeted, with fewer sustained follow‐up sequences like those seen in Cluster 1. Students who did not belong to the three clusters emerged as outliers (or intensive explorers), distinguished by their much higher frequency of interactions with ChatGPT, averaging 25 queries per student. Notably, performance within this group varied widely, including both very high‐performing and very low‐performing students. This indicates that frequent interaction with ChatGPT alone does not guarantee success; instead, the quality of engagement and how effectively students leverage the tool play critical roles.</p> <hd id="AN0187257453-34">Qualitative strand</hd> <p>The open‐coding process generated 105 initial codes, which were reconfigured and categorized. By examining the categories and their interconnections, we further categorized them and generated six themes following the HALIE framework. The themes, categories and sample codes can be seen in Table 6. Sample quotes can be seen in Appendix B. The majority of students identified them as nontraditional students, who rated their digital skills as average while the skills improved after the class. Some students expressed their discomfort using ChatGPT but stated the possibility of feeling differently 'if I was younger'. Despite that, as most students used ChatGPT for the first time in the classes, they became increasingly comfortable with it and showcased a growing digital capability after the experience. Theme 1 indicates these novice ChatGPT users were engaged in the ChatGPT interactions for their projects at different levels. For instance, some students used it very sparingly for initial ideas while some evaluated the outputs and followed up for more specific information. Most students still questioned their prompt engineering skills, despite a user guide provided to them upfront. Theme 2 reveals the mixed evaluation of ChatGPT outputs. After experiencing ChatGPT, some students realized and emphasized the constraints of GenAI, such as repetitive or vague outputs while others recognized the affordances, such as the visualization and breakdown of complex information. Themes 3 and 4 focus on first‐person perspectives. Many students showed initial concerns about ChatGPT and were hesitant to try the tool for their work. After the experience, there was a growing acceptance of and confidence in using this technology. Most students still regarded it as a tool to jumpstart or complement their project, leading to a perceived ownership of work. While most students used the free version of ChatGPT (3.5) and strongly acknowledged the efficiency while using ChatGPT, quite a few students voiced the limitations, such as outdated and repetitive outputs which further led to a feeling of getting lost, as indicated in Theme 5. Regarding preference in Theme 6, several students spoke very negatively but most people expressed how they were 'fascinated' by the performance of ChatGPT and have a strong willingness to continue using it in future classes and jobs.</p> <p>6 TABLE Themes, categories and sample codes.</p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Themes&lt;/th&gt;&lt;th align="left"&gt;Categories&lt;/th&gt;&lt;th align="left"&gt;Sample codes&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;1. Process: novice ChatGPT users were intentionally and selectively engaged&lt;/td&gt;&lt;td align="left"&gt;Digital skills journey&lt;/td&gt;&lt;td align="left"&gt;Progress in digital capability&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Interactive engagement&lt;/td&gt;&lt;td align="left"&gt;Iterative prompting for better results&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Selective usage&lt;/td&gt;&lt;td align="left"&gt;Using ChatGPT for initial ideas/outline&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;2. Output: mixed evaluation of ChatGPT outputs&lt;/td&gt;&lt;td align="left"&gt;ChatGPT limitations&lt;/td&gt;&lt;td align="left"&gt;Unclear or vague outputs&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;ChatGPT affordances&lt;/td&gt;&lt;td align="left"&gt;Break down complex info&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;3. First&amp;#8208;person perspectives: Growing acceptance of ChatGPT&lt;/td&gt;&lt;td align="left"&gt;Initial concerns&lt;/td&gt;&lt;td align="left"&gt;Ethical concerns&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Resistance&lt;/td&gt;&lt;td align="left"&gt;Initial hesitation/doubt&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Growing acceptance&lt;/td&gt;&lt;td align="left"&gt;Understanding tool's role&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;4. First&amp;#8208;person perspectives: maintaining ownership of work&lt;/td&gt;&lt;td align="left"&gt;Ownership of work&lt;/td&gt;&lt;td align="left"&gt;Claiming ownership of output&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Personal accountability&lt;/td&gt;&lt;td align="left"&gt;Taking responsibility&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Tool versus partner&lt;/td&gt;&lt;td align="left"&gt;Complementary tool&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;5. Quality: Efficiency and limitations coexist&lt;/td&gt;&lt;td align="left"&gt;Time efficiency&lt;/td&gt;&lt;td align="left"&gt;Time&amp;#8208;saving benefits&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Output accuracy&lt;/td&gt;&lt;td align="left"&gt;Outdated information&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Technical limitations&lt;/td&gt;&lt;td align="left"&gt;Get lost in long output&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;6. Preference: Enjoyment and continuous usage intention&lt;/td&gt;&lt;td align="left"&gt;Enjoyment&lt;/td&gt;&lt;td align="left"&gt;Positive experience&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Negative experience&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Willingness to reuse&lt;/td&gt;&lt;td align="left"&gt;Plans for future use&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd id="AN0187257453-35">DISCUSSION</hd> <p>In this study, we tried to provide a comprehensive picture of nontraditional students' interactions with ChatGPT to facilitate their learning through quantitative and qualitative approaches. We analysed four metrics in the quantitative strand, including prompt number, DoK, prompt relevance and originality. The average prompt number was seven, reflecting students' limited time on task as they engaged with instructional design tasks, such as developing training proposals or analysing trends and cases. The qualitative findings showed their growing acceptance of ChatGPT but also revealed their initial concerns and resistance, which might explain the relatively low prompt numbers. The DoK level hovered around Level 2, indicating a focus on lower‐order thinking skills like recall and reproduction, with minimal application of concepts. Nontraditional students in this class were still progressing in their digital capacity being the first‐time user of ChatGPT as they stated in the interviews. Even though they interactively engaged in the activity, the engagement was selective, with most of them using ChatGPT for initial ideation, resulting in low DoK levels. Prompt relevance revealed a pattern of isolated and narrowly scoped queries, showing limited progression in complexity and specificity. Originality, on the other hand, demonstrated moderate levels, with some students significantly modifying ChatGPT's output by incorporating their ideas and creativity. It was also reflected in one of the themes in the qualitative analysis, that learners maintained ownership of work as they used it as a supplementary tool and took accountability of their work. This may, in part, reflect the high creative skills often observed in and associated with nontraditional students, such as neurodivergent learners (Hayashibara et al., [<reflink idref="bib32" id="ref76">32</reflink>]; Pasarín‐Lavín et al., [<reflink idref="bib53" id="ref77">53</reflink>]). While students exhibited originality, the other metrics suggest they are still developing skills in effectively leveraging ChatGPT as a learning tool. Positive correlations between prompt numbers and both DoK and relevance suggest that training students to craft more prompts may deepen their engagement with content. However, originality's negative correlation with both DoK and relevance may suggest potential over‐reliance on ChatGPT (Fuchs, [<reflink idref="bib25" id="ref78">25</reflink>]; Sok &amp; Heng, [<reflink idref="bib60" id="ref79">60</reflink>]; Stojanov et al., [<reflink idref="bib61" id="ref80">61</reflink>]), where dense interactions with the tool may not equate to genuine, independent thought. Our findings on the negative correlation between originality and both DoK and prompt relevance align with recent empirical evidence on AI over‐reliance patterns (Zhai et al., [<reflink idref="bib73" id="ref81">73</reflink>]). While students in our study demonstrated high self‐efficacy and engagement with AI tools, which might partially be attributed to the video tutorials and guidelines, this confidence may actually enable problematic dependency. Malik et al. ([<reflink idref="bib48" id="ref82">48</reflink>]) found in their survey that 75% of studied learners reported reduced critical thinking despite feeling comfortable using AI tools. The interaction patterns we observed, particularly among exploratory inquirers who relied heavily on isolated, narrow queries, mirror the concerning behaviours documented by Duhaylungsod and Chavez ([<reflink idref="bib19" id="ref83">19</reflink>]), where over‐reliance led to complacency and undue dependence on AI dialogue systems.</p> <p>The contrast between our strategic inquirers, who maintained higher originality through purposeful questioning, and other groups who showed lower originality scores despite frequent AI use, is consistent with Kim et al.'s ([<reflink idref="bib39" id="ref84">39</reflink>]) finding that simplified AI interactions can undermine learners' independent analytical capabilities. Our results showing high engagement but reduced originality particularly resonate with Santiago Jr. et al.'s ([<reflink idref="bib55" id="ref85">55</reflink>]) faculty observations about AI tools potentially impeding critical thinking development despite increasing surface‐level task completion. As Grassini ([<reflink idref="bib28" id="ref86">28</reflink>]) demonstrated through empirical analysis, the convenience of AI assistance can mask a decline in students' analytical capabilities. This is a pattern that helps explain why we found higher self‐efficacy correlating with lower originality scores.</p> <p>The findings from the qualitative data may shed light on students' interaction behaviours. Most students had never heard of or used ChatGPT prior to the class. Besides low comfort with and confidence in using such a new tool, some of them had strong privacy or ethical concerns or resistance that prevented them from fully engaging with it, which echoes prior literature (Adel et al., [<reflink idref="bib1" id="ref87">1</reflink>]; Guleria et al., [<reflink idref="bib30" id="ref88">30</reflink>]; Luo et al., [<reflink idref="bib47" id="ref89">47</reflink>]). Despite being impressed by the performance of ChatGPT, students admitted to using ChatGPT sparingly as a starting point. Notably, they were less concerned after the class and showed more willingness to be more engaged. The text mining approach revealed three clusters and one outlier cluster based on their query patterns. Cluster 1 students (13%) were identified as strategic inquirers with distinctly more follow‐ups from information requests and explanations, compared with Cluster 2 diverse explorers who asked a mix of query types, and Cluster 3 exploratory inquirers who mainly used shorter explanation queries.</p> <p>The quantitative analyses revealed the significance of prompt number and engagement level in predicting performance scores and several pairs of significant correlations among the ChatGPT interactions. Prompt numbers, as a behavioural indicator of time on task similarly to the number of discussion posts, naturally aligned with higher performance, echoing prior research linking engagement to learning outcomes (Zheng &amp; Warschauer, [<reflink idref="bib75" id="ref90">75</reflink>]). Self‐efficacy was positively and significantly correlated with engagement, yet it was not a significant predictor for performance, possibly due to calibration issues where students' perceived confidence did not align with their actual capabilities (Bol &amp; Hacker, [<reflink idref="bib8" id="ref91">8</reflink>]). Qualitative findings suggest that as students became more comfortable using ChatGPT, they may have tended to over‐rely on the tool, which inadvertently hindered originality in their assignments. This reliance could explain the negative relationship between self‐efficacy and originality, as students may increasingly adopt ChatGPT's output without substantial personal contribution. Students' DoK and prompt relevance levels are also significantly correlated with their performance, indicating that a higher level of thinking and reasoning in interacting with ChatGPT and more relevant prompts tend to be associated with a higher level of performance. By examining the relationships among the ChatGPT interactions, we found that prompt number, DoK and prompt relevance are significantly related to each other, while originality is significantly negatively related to DoK and prompt relevance. These findings underscore the nuanced and intricate relationships between engagement, self‐efficacy, interactions and performance in a ChatGPT‐facilitated learning environment, warranting further exploration (Liang et al., [<reflink idref="bib44" id="ref92">44</reflink>]; Shahzad et al., [<reflink idref="bib57" id="ref93">57</reflink>]).</p> <p>The findings from Cluster 1, for example, highlighted the potential for AI tools to support purposeful, inquiry‐driven learning when students adopted structured interaction strategies. Educational interventions designed to scaffold such strategies—for example, teaching students how to formulate effective follow‐up questions or encouraging iterative inquiry—could help other learners replicate the successful behaviours observed in this cluster. To support students like Bella from Cluster 3, educational interventions could encourage more active and reflective engagement with AI tools. For example, students could be prompted to go beyond reproducing the guidelines by generating additional follow‐up questions, exploring the reasoning behind their choices or considering alternative approaches to the task. Practically speaking, these results suggest that educators need to provide targeted guidance on integrating ChatGPT into learning and design environments. Despite a user guide and video demo on using ChatGPT, which were perceived as very helpful resources, students' interactions with the tool largely remained at lower to medium cognitive levels as they lacked the skills to engage in ChatGPT at high cognitive levels. Therefore, we suggest that prompt engineering should be explicitly taught as a critical skill to empower learners to engage in higher‐order thinking and problem‐solving (Gregory, [<reflink idref="bib29" id="ref94">29</reflink>]). Educators should also emphasize the importance of balancing ChatGPT's affordances with critical, independent thinking to avoid over‐reliance on GenAI‐generated outputs (Luo et al., [<reflink idref="bib47" id="ref95">47</reflink>]). Academic institutions and organizations should consider investing in AI literacy training that raises awareness about the limitations and ethical considerations of GenAI tools (Luo et al., [<reflink idref="bib47" id="ref96">47</reflink>]). To respect the diversity in students' learning journeys, instructors should consider offering personalized feedback and guidance that aligns with each student's approach and needs.</p> <hd id="AN0187257453-36">IMPLICATIONS FOR RESEARCH AND PRACTICE</hd> <p>This study is among the first to examine and evaluate learners' GenAI interactions and the interplay between interaction, engagement, self‐efficacy and performance in a ChatGPT‐facilitated learning environment using a mixed‐methods approach. We adapted the HALIE framework to guide our study and adopted the DoK levels to understand students' cognitive engagement in the complex problem‐solving process facilitated by ChatGPT. By investigating the interactions through multiple metrics, instructors and researchers can gain insights into the depth of cognitive engagement and the sophistication of AI tool use. Future research should adapt and improve these metrics in other tasks or contexts to identify areas where learners may need additional support in leveraging AI tools more effectively for higher‐order thinking tasks. In this study, we did not find the significance of students' self‐efficacy in predicting their performance, which might be counterintuitive. The survey was deployed at the end of the semester. We recommend future studies to conduct a pre‐survey as well to have a more accurate measurement of learners' self‐efficacy at the time of using the AI tools. Additionally, future research can consider incorporating originality as a criterion in measuring performance, instead of treating it as a separate variable, to better capture a more nuanced understanding of learning outcomes as it can assess learners' genuine learning and creativity while still acknowledging the role of AI as a supportive tool.</p> <p>Examining nontraditional students' technology usage and efficacy remains underexplored (Sutherland et al., [<reflink idref="bib63" id="ref97">63</reflink>]), especially the integration of AI. Nontraditional students, as a growing population in higher education over the past decade, require even more support when it comes to technology integration due to the traditional digital divide (Banerjee, [<reflink idref="bib5" id="ref98">5</reflink>]; Sutherland et al., [<reflink idref="bib63" id="ref99">63</reflink>]). In the era of AI, to prepare the future workforce with the needed digital competencies, more attention should be placed to support nontraditional students. Considering their characteristics, we recommend instructors include more targeted technology training and practice when applicable, along with resources supporting their self‐paced learning. Students in this class recognized the usefulness of a ChatGPT user guide and a video demo created by the research team. Instructors can create microlearning modules (eg, short videos and text‐based job aids) prior to the integration of new tools in coursework to facilitate a more positive learning experience.</p> <hd id="AN0187257453-37">LIMITATIONS</hd> <p>We acknowledge that the quantitative sample size was not large enough, restricting the use of more robust analytical techniques. Considering the unique context of the present study, the generalizability of the findings might be limited when extended to other types of student populations. Our participants primarily consisted of nontraditional students in the United States who often might face unique challenges compared with those in other countries or areas. Those characteristics might not match with traditional student populations. However, we followed a rigorous research design and provided transparency in our methodology for future adaptation and application in other contexts. For example, the metrics to evaluate ChatGPT interactions can be replicated or adapted in studies involving traditional students or other learner demographics. Additionally, synthesizing data from diverse sources posed challenges, and unique nuances across the undergraduate and graduate courses were not fully captured. The study also did not comprehensively account for the varied characteristics of nontraditional learners (ie, age and employment status). Future research should address these limitations by using larger sample sizes and more disaggregated data to explore fine‐grained differences based on nontraditional status. Additionally, future research could benefit from in‐depth qualitative approaches, such as phenomenology or narrative inquiry, to provide valuable contextual insights into the unique lived experiences and perspectives of individual participants using ChatGPT for their learning and how such use may impact learning outcomes around engagement, performance and self‐efficacy.</p> <hd id="AN0187257453-38">CONCLUSION</hd> <p>In this mixed‐methods research, we explored nontraditional students' ChatGPT interactions to facilitate their instructional design, model creation, case development and trends analysis projects. By adapting the HALIE framework, we conducted quantitative tests, qualitative thematic analysis and text mining techniques to examine students' ChatGPT interaction and relationships between the variables. The results revealed different query patterns, a relatively low average depth of knowledge and prompt relevance, yet significant associations with performance. Prompt number and engagement were key significant predictors of performance. All ChatGPT interaction characteristics examined in this study were significantly correlated with each other. The findings underscore the importance of analysing learners' GenAI interactions to inform course designs that foster learners' engagement and critical thinking.</p> <hd id="AN0187257453-39">CONFLICT OF INTEREST STATEMENT</hd> <p>The authors declare no conflicts of interest regarding the research.</p> <hd id="AN0187257453-40">FUNDING INFORMATION</hd> <p>This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors.</p> <hd id="AN0187257453-41">DATA AVAILABILITY STATEMENT</hd> <p>The datasets of the current study are available from the corresponding author upon reasonable request.</p> <hd id="AN0187257453-42">ETHICS STATEMENT</hd> <p>The Old Dominion University Education Human Subjects Review Committee (2103003‐2) has approved this study. Students were assigned pseudonyms in the manuscript.</p> <hd id="AN0187257453-43">A APPENDIX ChatGPT INTERACTION METRICS</hd> <hd1 id="AN0187257453-44">Depth of Knowledge (DoK)</hd1> <p></p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Recall and recognition of factual info (1)&lt;/th&gt;&lt;th align="left"&gt;Skills and concepts processing (2)&lt;/th&gt;&lt;th align="left"&gt;Strategic thinking and reasoning (3)&lt;/th&gt;&lt;th align="left"&gt;Extended thinking (4)&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts that require basic recall of facts, definitions or simple procedures&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Low cognitive demand, straightforward and typically single&amp;#8208;step questions&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Little to no follow&amp;#8208;up or deeper engagement with ChatGPT responses&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts that require comprehension and processing of information or ChatGPT outcomes&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Involve more than one step but still relatively straightforward&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Some follow&amp;#8208;up questions show an attempt to build on previous responses&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts that require reasoning, planning and using evidence&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;More complex and abstract, often involving multiple steps and deeper analysis&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Follow&amp;#8208;up questions build on previous interactions to explore topics more deeply&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts that require complex reasoning, planning, developing and evaluating&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Involve multiple steps and sustained effort over a longer period&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Follow&amp;#8208;up questions show a sophisticated development and refinement of ideas&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd1 id="AN0187257453-45">Prompt relevance and evolution</hd1> <p></p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Low relevance and evolution (1)&lt;/th&gt;&lt;th align="left"&gt;Limited relevance and limited evolution (2)&lt;/th&gt;&lt;th align="left"&gt;Moderate relevance and clear evolution (3)&lt;/th&gt;&lt;th align="left"&gt;High relevance and advanced evolution (4)&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts are largely unrelated to each other&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Little to no narrowing of scope; prompts are broad and disconnected&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;No clear progression or development in the line of questioning&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts show some relevance to each other but lack depth and focus&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Limited narrowing of scope; some prompts build on previous ones but not consistently&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Minimal progression in the complexity and specificity of questions&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts are mostly relevant and build on each other with a clear line of questioning&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Noticeable narrowing of scope; prompts become more focused and detailed&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Clear progression in the complexity and specificity of questions&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Prompts are highly relevant and show a strong development of ideas&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Clear narrowing of scope with highly focused and interrelated prompts&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Advanced progression in the complexity, specificity and depth of questions&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd1 id="AN0187257453-46">Originality</hd1> <p></p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Low originality (1)&lt;/th&gt;&lt;th align="left"&gt;Limited originality (2)&lt;/th&gt;&lt;th align="left"&gt;Moderate originality (3)&lt;/th&gt;&lt;th align="left"&gt;High originality (4)&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;The topic idea was purely from ChatGPT&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Direct copying and pasting from ChatGPT without any modification&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Minimal to no personal input or original content&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Reliance entirely on ChatGPT's responses&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;The selection of the topic idea was generated based on the interaction with ChatGPT&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Some modification of ChatGPT's output but still heavily reliant on AI&amp;#8208;generated content&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Limited personal input and minimal original content&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Basic attempts to paraphrase or slightly modify ChatGPT's responses&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Idea was from student but refined based on the interaction with ChatGPT&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Significant modification of ChatGPT's output with substantial personal input&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Integration of personal ideas, analysis and creativity&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Effective use of ChatGPT as a tool to support and enhance original work&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;td align="left"&gt;&lt;list list-type="Bullet"&gt;&lt;list-item&gt;&lt;p&gt;Students had a clear training idea from the beginning&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Extensive and sophisticated use of ChatGPT to refine and enhance original content&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;High degree of personal input, creativity and critical thinking&lt;/p&gt;&lt;/list-item&gt;&lt;list-item&gt;&lt;p&gt;Consistent and purposeful prompting and training of ChatGPT to obtain desired outputs&lt;/p&gt;&lt;/list-item&gt;&lt;/list&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <hd id="AN0187257453-47">B APPENDIX SAMPLE QUOTES FOR THE THEMES AND CATEGORIES</hd> <p></p> <p> <ephtml> &lt;table&gt;&lt;thead valign="bottom"&gt;&lt;tr&gt;&lt;th align="left"&gt;Themes&lt;/th&gt;&lt;th align="left"&gt;Categories&lt;/th&gt;&lt;th align="left"&gt;Sample codes&lt;/th&gt;&lt;th align="left"&gt;Sample quotes&lt;/th&gt;&lt;/tr&gt;&lt;/thead&gt;&lt;tbody valign="top"&gt;&lt;tr&gt;&lt;td align="left"&gt;1. Process: novice ChatGPT users were intentionally and selectively engaged&lt;/td&gt;&lt;td align="left"&gt;Digital skills journey&lt;/td&gt;&lt;td align="left"&gt;Progress in digital capability&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;Since I was initially unfamiliar with ChatGPT I wasn't confident that I would know how to complete the proposal assignment. By reviewing the resources, videos and instructions I was able to overcome any technological barriers&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Interactive engagement&lt;/td&gt;&lt;td align="left"&gt;Iterative prompting for better results&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;For example, I said give me a proposal for overlanding. It gave me a generalization and a format, but I was like, Okay, well, this is not really college level. So I said give me a college type proposal for overlanding. And then it re&amp;#8208;scribed it on how I wanted it to be with college literacy&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Selective usage&lt;/td&gt;&lt;td align="left"&gt;Using ChatGPT for initial ideas/outline&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I'm using it more because we're required to interact with it...if we were given the opportunity to use it or not use it, I would probably be more of an 80 I probably wouldn't use it and 20 I probably would just to get that extra feedback. Like to get that extra little... like something that I might have missed&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;2. Output: Mixed evaluation of ChatGPT outputs&lt;/td&gt;&lt;td align="left"&gt;ChatGPT limitations&lt;/td&gt;&lt;td align="left"&gt;Unclear or vague outputs&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I say the biggest challenge I would have is the, I will say, ambiguity, but the lack of long established precedents on like, citation and accreditation or, you know, using it without encountering problems with plagiarism, you know what I mean?&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;ChatGPT affordances&lt;/td&gt;&lt;td align="left"&gt;Jump start the project&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;...used my thoughts to create certain things and it allowed me to be introduced to new ideas and use the ideas to implement into my work. ChatGPT acted as a ladder for me to climb to elevate my projects&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;3. First&amp;#8208;person perspectives: Growing acceptance of ChatGPT&lt;/td&gt;&lt;td align="left"&gt;Resistance&lt;/td&gt;&lt;td align="left"&gt;Ethical concerns&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I've had others, like younger students say, well, hey, why don't you just, you know, use the ChatGPT for this, you know, it's so much quicker, so much easier. And I'm like, that's fine, but as a person going back to, you know, saying, morally, our ethics, stuff like that. What am I teaching myself? I'm teaching myself the quick, easy way to get an answer. But it does me no good, especially if I'm in the workforce, and say that I'm in a meeting with the board of directors for something and it's like, hey, hold on a minute. I need to get my ChatGPT up so I can give you this answer. So what have I taught myself morally? Nothing really&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Cognitively challenged&lt;/td&gt;&lt;td align="left"&gt;Think outside the box&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;Its ability to offer diverse perspectives and suggestions that I might not have considered otherwise. It helped me think outside the box and explore different angles for my eLearning proposal&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Growing acceptance&lt;/td&gt;&lt;td align="left"&gt;Understanding tool's role&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I guess we all, it sounds like everyone here figured out that it works and can help. Maybe throughout the course of the class, some of us already knew. But others may have grown to that realization that it's not a boogeyman. It's not going to do anything crazy. It'll do what you ask it to do&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;4. First&amp;#8208;person perspectives: Maintaining ownership of work&lt;/td&gt;&lt;td align="left"&gt;Ownership of work&lt;/td&gt;&lt;td align="left"&gt;Ownership of process&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I felt as if ownership was not 100% mine, but I felt as if it was an easier way to research a topic as you would with anything and get your results all in one place. I did feel I owned more of the learning process as I provided the prompts and tailored them to my specific needs&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Personal accountability&lt;/td&gt;&lt;td align="left"&gt;Taking responsibility&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I feel accountable with anything that I use with chat GPT because I feel like even though it's always right, I always like to fact checked it fact check it. So if it were to like, give me an idea or a topic, I'm like, Okay, let me go find some references that I can use to support this idea and not just use it on my own. Being accountable means that I can stand by what I say no matter if I get the idea or topic from chat GPT or on my own, and being able to like defend it&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Tool versus partner&lt;/td&gt;&lt;td align="left"&gt;Complementary tool&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I see it as a tool. But I know that we have to go further than that. I mean, just because chat and AI are opening all kinds of doors for all kinds of other things. I just wanted to add that what I, what I feel is that it complements my work, that it helps me get it done faster. So it's more efficient for me. I don't get stuck and I don't have to sit there and think for hours on something when I can ask the question if that makes sense&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;5. Quality: Efficiency and limitations coexist&lt;/td&gt;&lt;td align="left"&gt;Time efficiency&lt;/td&gt;&lt;td align="left"&gt;Time&amp;#8208;saving benefits&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I feel as if I'm learning more at a faster rate when using it because it's giving me exactly what I'm asking for, which reduces research time dramatically&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Output accuracy&lt;/td&gt;&lt;td align="left"&gt;Outdated information&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;I think it tells you in the User Agreement that it only goes up to 2022 publication, or either two, I'm pretty sure it's 2022. So, it's not very up to date&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Technical limitations&lt;/td&gt;&lt;td align="left"&gt;Get lost in long output&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;You've already kind of got in mind what you're going to do, and as I started chatting with, you know, the AI stuff...I find myself starting to dig in deeper, okay, I want more detail. I don't want the same thing...So I find myself two hours into this chatting with this thing and it's still not gotten me anywhere, where what I'm looking for&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;6. Preference: Enjoyment and continuous usage intention&lt;/td&gt;&lt;td align="left"&gt;Enjoyment&lt;/td&gt;&lt;td align="left"&gt;Positive experience&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;My experience with ChatGPT was incredibly positive. AI was able to provide me with valuable insights and suggestions to help me refine my proposal development process. I found the responses to be relevant, well&amp;#8208;researched and overall, very helpful in guiding me towards a more polished product&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Negative experience&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;It seems like a true waste of the opportunity of knowledge. I do not like using ChatGPT, it takes all creativity out of the picture and also does not allow us to learn things organically&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;tr&gt;&lt;td align="left"&gt;Willingness to reuse&lt;/td&gt;&lt;td align="left"&gt;Plans for future use&lt;/td&gt;&lt;td align="left"&gt;&lt;italic&gt;Now I see that it can be used for much more than completely doing your assignments. I will be using it more often&lt;/italic&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/tbody&gt;&lt;/table&gt; </ephtml> </p> <ref id="AN0187257453-48"> <title> REFERENCES </title> <blist> <bibl id="bib1" idref="ref59" type="bt">1</bibl> <bibtext> Adel, A., Ahsan, A., &amp; Davison, C. (2024). ChatGPT promises and challenges in education: Computational and ethical perspectives. Education Sciences, 14 (8), 814. https://doi.org/10.3390/educsci14080814</bibtext> </blist> <blist> <bibl id="bib2" idref="ref37" type="bt">2</bibl> <bibtext> Alt, D., &amp; Raichel, N. (2020). Enhancing perceived digital literacy skills and creative self‐concept through gamified learning environments: Insights from a longitudinal study. International Journal of Educational Research, 101, 101561. https://doi.org/10.1016/j.ijer.2020.101561</bibtext> </blist> <blist> <bibl id="bib3" idref="ref61" type="bt">3</bibl> <bibtext> Avey, J. B., Avolio, B. J., Crossley, C. D., &amp; Luthans, F. (2009). Psychological ownership: Theoretical extensions, measurement and relation to work outcomes. Journal of Organizational Behavior, 30 (2), 173 – 191. https://doi.org/10.1002/job.583</bibtext> </blist> <blist> <bibl id="bib4" idref="ref50" type="bt">4</bibl> <bibtext> Barber, J. (2018). Depth of knowledge and conceptual understanding. Science Scope, 41 (9), 76 – 81.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref14" type="bt">5</bibl> <bibtext> Banerjee, M. (2020). An exploratory study of online equity: Differential levels of technological access and technological efficacy among underserved and underrepresented student populations in higher education. Interdisciplinary Journal of e‐Skills and Lifelong Learning, 16, 93 – 121. https://doi.org/10.28945/4664</bibtext> </blist> <blist> <bibl id="bib6" idref="ref6" type="bt">6</bibl> <bibtext> Bauer, L., Burtis, E., Howard, O., &amp; Wang, S. Y. (2023). Which nontraditional students are also in the labor force? The Hamilton Project. https://<ulink href="http://www.hamiltonproject.org/publication/post/which&amp;#8208;nontraditional&amp;#8208;students&amp;#8208;are&amp;#8208;also&amp;#8208;in&amp;#8208;the&amp;#8208;labor&amp;#8208;force/">www.hamiltonproject.org/publication/post/which&amp;#8208;nontraditional&amp;#8208;students&amp;#8208;are&amp;#8208;also&amp;#8208;in&amp;#8208;the&amp;#8208;labor&amp;#8208;force/</ulink></bibtext> </blist> <blist> <bibl id="bib7" idref="ref7" type="bt">7</bibl> <bibtext> Bean, J. P., &amp; Metzner, B. S. (1985). A conceptual model of nontraditional undergraduate student attrition. Review of Educational Research, 55 (4), 485 – 540. https://doi.org/10.3102/00346543055004485</bibtext> </blist> <blist> <bibl id="bib8" idref="ref91" type="bt">8</bibl> <bibtext> Bol, L., &amp; Hacker, D. J. (2012). Calibration research: Where do we go from here? Frontiers in Psychology, 3, 229. https://doi.org/10.3389/fpsyg.2012.00229</bibtext> </blist> <blist> <bibl id="bib9" idref="ref2" type="bt">9</bibl> <bibtext> Bolick, A. D., &amp; da Silva, R. L. (2024). Exploring artificial intelligence tools and their potential impact to instructional design workflows and organizational systems. TechTrends, 68, 91 – 100. https://doi.org/10.1007/s11528‐023‐00894‐2</bibtext> </blist> <blist> <bibtext> Cavazos, J. T., Hauck, K. A., Baskin, H. M., &amp; Bain, C. M. (2024). ChatGPT goes to college: Exploring student perspectives on artificial intelligence in the classroom. Teaching of Psychology. https://doi.org/10.1177/00986283241268829</bibtext> </blist> <blist> <bibtext> Champa, A. I., Rabbi, M. F., Nachuma, C., &amp; Zibran, M. F. (2024). ChatGPT in action: Analyzing its use in software development. In Proceedings of the 21st International Conference on Mining Software Repositories (pp. 182 – 186).</bibtext> </blist> <blist> <bibtext> Chang, D. H., Lin, M. P. C., Hajian, S., &amp; Wang, Q. Q. (2023). Educational design principles of using AI chatbot that supports self‐regulated learning in education: Goal setting, feedback, and personalization. Sustainability, 15 (17), 12921. https://doi.org/10.3390/su151712921</bibtext> </blist> <blist> <bibtext> Choy, S. (2002). Nontraditional undergraduates: Findings from "the condition of education, 2002". National Center for Education Statistics.</bibtext> </blist> <blist> <bibtext> Clarke, V., &amp; Braun, V. (2017). Thematic analysis. The Journal of Positive Psychology, 12 (3), 297 – 298.</bibtext> </blist> <blist> <bibtext> Cleveland‐Innes, M. (1994). Adult student drop‐out at post‐secondary institutions. The Review of Higher Education, 17 (4), 423 – 445. https://doi.org/10.1353/rhe.1994.0011</bibtext> </blist> <blist> <bibtext> Creswell, J. W., &amp; Clark, V. L. P. (2017). Designing and conducting mixed methods research. Sage Publications.</bibtext> </blist> <blist> <bibtext> Dell'Acqua, F., McFowland III, E., Mollick, E. R., Lifshitz‐Assaf, H., Kellogg, K., Rajendran, S., &amp; Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. Harvard Business School Technology &amp; Operations Management. Unit Working Paper (24‐013).</bibtext> </blist> <blist> <bibtext> Dolch, C., &amp; Zawacki‐Richter, O. (2018). Are students getting used to learning technology? Changing media usage patterns of traditional and non‐traditional students in higher education. Research in Learning Technology, 26. https://doi.org/10.25304/rlt.v26.2038</bibtext> </blist> <blist> <bibtext> Duhaylungsod, A. V., &amp; Chavez, J. V. (2023). ChatGPT and other AI users: Innovative and creative utilitarian value and mindset shift. Journal of Namibian studies: History Politics Culture, 33, 4367 – 4378. https://doi.org/10.59670/jns.v33i.2791</bibtext> </blist> <blist> <bibtext> Escalante, J., Pack, A., &amp; Barrett, A. (2023). AI‐generated feedback on writing: Insights into efficacy and ENL student preference. International Journal of Educational Technology in Higher Education, 20, 57. https://doi.org/10.1186/s41239‐023‐00425‐2</bibtext> </blist> <blist> <bibtext> Francis, E. (2021). Deconstructing depth of knowledge. Solution Tree.</bibtext> </blist> <blist> <bibtext> Fairlie, R. W., &amp; London, R. A. (2012). The effects of home computers on educational outcomes: Evidence from a field experiment with community college students. The Economic Journal, 122 (561), 727 – 753. https://doi.org/10.1111/j.1468‐0297.2011.02484.x</bibtext> </blist> <blist> <bibtext> Fredricks, J. A., Blumenfeld, P. C., &amp; Paris, A. H. (2004). School engagement: Potential of the concept, state of the evidence. Review of Educational Research, 74 (1), 59 – 109. https://doi.org/10.3102/00346543074001059</bibtext> </blist> <blist> <bibtext> Fredricks, J. A., &amp; McColskey, W. (2012). The measurement of student engagement: A comparative analysis of various methods and student self‐report instruments. In Handbook of research on student engagement (pp. 763 – 782). Springer US.</bibtext> </blist> <blist> <bibtext> Fuchs, K. (2023). Exploring the opportunities and challenges of NLP models in higher education: Is chat GPT a blessing or a curse? Frontiers in Education, 8, 1166682. https://doi.org/10.3389/feduc.2023.1166682</bibtext> </blist> <blist> <bibtext> Harris, M. A., &amp; Patten, K. P. (2015). Using Bloom's and Webb's taxonomies to integrate emerging cybersecurity topics into a computing curriculum. Journal of Information Systems Education, 26 (3), 219 – 234.</bibtext> </blist> <blist> <bibtext> Hess, K. K., Jones, B. S., Carlock, D., &amp; Walkup, J. R. (2009). Cognitive rigor: Blending the strengths of Bloom's Taxonomy and Webb's Depth of Knowledge to enhance classroom‐level processes. Educational Psychology Review, 21 (4), 383 – 401.</bibtext> </blist> <blist> <bibtext> Grassini, S. (2023). Shaping the future of education: Exploring the potential and consequences of AI and chatGPT in educational settings. Education Sciences, 13 (7), 692. https://doi.org/10.3390/educsci13070692</bibtext> </blist> <blist> <bibtext> Gregory, S. F. (2024). Empowering teaching with prompt engineering: How to integrate curriculum, standards, and assessment for a new age. In Transforming education with generative AI: Prompt engineering and synthetic content creation (pp. 239 – 260). IGI Global.</bibtext> </blist> <blist> <bibtext> Guleria, A., Krishan, K., Sharma, V., &amp; Kanchan, T. (2023). ChatGPT: Ethical concerns and challenges in academics and research. The Journal of Infection in Developing Countries, 17 (9), 1292 – 1299. https://doi.org/10.3855/jidc.18738</bibtext> </blist> <blist> <bibtext> Halverson, L. R., &amp; Graham, C. R. (2019). Learner engagement in blended learning environments: A conceptual framework. Online Learning, 23 (2), 145 – 178. https://doi.org/10.24059/olj.v23i2.1481</bibtext> </blist> <blist> <bibtext> Hayashibara, E., Savickaite, S., &amp; Simmons, D. (2023). Creativity and neurodiversity: Towards an inclusive creativity measure for autism and ADHD. OSF preprint. https://doi.org/10.31219/osf.io/4vqh5</bibtext> </blist> <blist> <bibtext> Henrie, C. R., Halverson, L. R., &amp; Graham, C. R. (2015). Measuring student engagement in technology‐mediated learning: A review. Computers &amp; Education, 90, 36 – 53. https://doi.org/10.1016/j.compedu.2015.09.005</bibtext> </blist> <blist> <bibtext> Holmegaard, H. T., Madsen, L. M., &amp; Ulriksen, L. (2017). Why should European higher education care about the retention of non‐traditional students? European Educational Research Journal, 16 (1), 3 – 11. https://doi.org/10.1177/1474904116683688</bibtext> </blist> <blist> <bibtext> Horn, L., Nevill, S., &amp; Griffith, J. (2006). Profile of undergraduates in US postsecondary education institutions, 2003‐04: With a special analysis of community college students. Statistical Analysis Report, NCES 2006‐184. National Center for Education Statistics.</bibtext> </blist> <blist> <bibtext> Jesnek, L. M. (2012). Empowering the non‐traditional college student and bridging the "digital divide". Contemporary Issues in Education Research, 5 (1), 1 – 8.</bibtext> </blist> <blist> <bibtext> Jiang, S., Tatar, C., &amp; Xie, C. (2021). Augmented reality in science laboratories: Investigating high school students' navigation patterns and their effects on learning performance. Educational Technology Research and Development, 60 (3). https://doi.org/10.1177/07356331211038764</bibtext> </blist> <blist> <bibtext> Jones, D. J., &amp; Watson, B. C. (1990). High‐risk students and higher education: Future trends. ASHE‐ERIC Higher Education Report No. 3. The George Washington University.</bibtext> </blist> <blist> <bibtext> Kim, Y., Lee, M., Kim, D., &amp; Lee, S.‐J. (2023). Towards explainable AI writing assistants for non‐native English speakers. arXiv preprint arXiv:2304.02625.</bibtext> </blist> <blist> <bibtext> Kumar, H., Musabirov, I., Reza, M., Shi, J., Wang, X., Williams, J. J., Kuzminykh, A., &amp; Liut, M. (2024). Impact of guidance and interaction strategies for LLM use on learner performance and perception. arXiv preprint arXiv:2310.13712.</bibtext> </blist> <blist> <bibtext> Kuo, Y. (2018). An exploratory study of minority students' technology usage and perceptions of technology: Nontraditional adult students in technology‐based environments. Journal of Research on Technology in Education, 50 (4), 350 – 364. https://doi.org/10.1080/15391523.2018.1522984</bibtext> </blist> <blist> <bibtext> Lee, M., Srivastava, M., Hardy, A., Thickstun, J., Durmus, E., Paranjape, A., Gerard‐Ursin, I., Li, X. L., Ladhak, F., Rong, F., Bose, S., Kennedy, B., Xiong, C., Strobelt, H., Luu, L., Whittaker, M., Wu, J., Srinivasan, A., Chilton, L., &amp; Liang, P. (2023). Evaluating human‐language model interaction. arXiv preprint arXiv:2212.09746.</bibtext> </blist> <blist> <bibtext> Li, B., Bonk, C. J., Wang, C., &amp; Kou, X. (2024). Reconceptualizing self‐directed learning in the era of generative AI: An exploratory analysis of language learning. IEEE Transactions on Learning Technologies, 17 (4), 1515 – 1529. https://doi.org/10.1109/TLT.2024.3386098</bibtext> </blist> <blist> <bibtext> Liang, J., Wang, L., Luo, J., Yan, Y., &amp; Fan, C. (2023). The relationship between student interaction with generative artificial intelligence and learning achievement: Serial mediating roles of self‐efficacy and cognitive engagement. Frontiers in Psychology, 14, 1285392. https://doi.org/10.3389/fpsyg.2023.1285392</bibtext> </blist> <blist> <bibtext> Lo, C. K., Hew, K. F., &amp; Jong, M. S. Y. (2024). The influence of ChatGPT on student engagement: A systematic review and future research agenda. Computers &amp; Education, 219, 105100. https://doi.org/10.1016/j.compedu.2024.105100</bibtext> </blist> <blist> <bibtext> Lombard, E. (2022). Do you want this double or single spaced? Transactional versus transformational questions. Journal of Education and Learning, 11 (5), 124. https://doi.org/10.5539/jel.v11n5p124</bibtext> </blist> <blist> <bibtext> Luo, T., Muljana, P. S., Ren, X., &amp; Young, D. (2024). Exploring instructional designers' utilization and perspectives on generative AI tools: A mixed methods study. Educational Technology Research and Development. https://doi.org/10.1007/s11423‐024‐10437‐y</bibtext> </blist> <blist> <bibtext> Malik, A. R., Pratiwi, Y., Andajani, K., Numertayasa, I. W., Suharti, S., &amp; Darwis, A. (2023). Exploring artificial intelligence in academic essay: Higher education student's perspective. International Journal of Educational Research Open, 5, 100296. https://doi.org/10.1016/j.ijedro.2023.100296</bibtext> </blist> <blist> <bibtext> Markle, G. (2015). Factors influencing persistence among nontraditional university students. Adult Education Quarterly, 65 (3), 267 – 285. https://doi.org/10.1177/0741713615583085</bibtext> </blist> <blist> <bibtext> Morgan, J. J., Brown, N. B., Hsiao, Y.‐J., Howerter, C., Juniel, P., Sedano, L., &amp; Castillo, W. L. (2014). Unwrapping academic standards to increase the achievement of students with disabilities. Intervention in School and Clinic, 49 (3), 131 – 141. https://doi.org/10.1177/1053451213496156</bibtext> </blist> <blist> <bibtext> National Center for Education Statistics. (1999). National Postsecondary Student Aid Study (NPSAs). U.S. Department of Education. https://nces.ed.gov/statprog/handbook/npsas.asp</bibtext> </blist> <blist> <bibtext> Nguyen, A., Hong, Y., Dang, B., &amp; Huang, X. (2024). Human‐AI collaboration patterns in AI‐assisted academic writing. Studies in Higher Education, 49 (5), 847 – 864. https://doi.org/10.1080/03075079.2024.2323593</bibtext> </blist> <blist> <bibtext> Pasarín‐Lavín, T., García, T., Abín, A., &amp; Rodríguez, C. (2024). Neurodivergent students. A continuum of skills with an emphasis on creativity and executive functions. Applied Neuropsychology: Child, 1 – 13. https://doi.org/10.1080/21622965.2024.2406914</bibtext> </blist> <blist> <bibtext> Reschly, A. L., &amp; Christenson, S. L. (2012). Jingle, jangle, and conceptual haziness: Evolution and future directions of the engagement construct. In Handbook of research on student engagement (pp. 3 – 19). Springer US.</bibtext> </blist> <blist> <bibtext> Santiago Jr, C. S., Embang, S. I., Conlu, M. T. N., Acanto, R. B., Lausa, S. M., Ambojia, K. W. P., Laput, E. Y., Aperocho, M. D. B., Malabag, B. A., &amp; Balilo, B. B., Jr. (2023). Utilization of writing assistance tools in research in selected higher learning institutions in the Philippines: A text mining analysis. International Journal of Learning, Teaching and Educational Research, 22 (11), 259 – 284. https://doi.org/10.26803/ijlter.22.11.14</bibtext> </blist> <blist> <bibtext> Sapir, A. (2022). What student support practitioners know about non‐traditional students: A practice‐based approach. Higher Education Research and Development, 41 (6), 2064 – 2078.</bibtext> </blist> <blist> <bibtext> Shahzad, M. F., Xu, S., &amp; Zahid, H. (2024). Exploring the impact of generative AI‐based technologies on learning performance through self‐efficacy, fairness &amp; ethics, creativity, and trust in higher education. Education and Information Technologies, 30 (3), 1 – 26. https://doi.org/10.1007/s10639‐024‐12949‐9</bibtext> </blist> <blist> <bibtext> Shepherd, M. M., &amp; Sheu, T. S. (2014). The effects of informal faculty‐student interaction and use of information technology on non‐traditional students' persistence intentions and educational outcomes. Journal of Higher Education Theory &amp; Practice, 14 (2), 46 – 60.</bibtext> </blist> <blist> <bibtext> Skinner, E. A., Kindermann, T. A., &amp; Furrer, C. J. (2009). A motivational perspective on engagement and disaffection: Conceptualization and assessment of children's behavioral and emotional participation in academic activities in the classroom. Educational and Psychological Measurement, 69 (3), 493 – 525. https://doi.org/10.1177/0013164408323233</bibtext> </blist> <blist> <bibtext> Sok, S., &amp; Heng, K. (2023). ChatGPT for education and research: A review of benefits and risks. Cambodian Journal of Educational Research, 3 (1), 110 – 121. https://doi.org/10.62037/cjer.2023.03.01.06</bibtext> </blist> <blist> <bibtext> Stojanov, A., Liu, Q., &amp; Koh, J. H. L. (2024). University students' self‐reported reliance on ChatGPT for learning: A latent profile analysis. Computers and Education: Artificial Intelligence, 6, 100243. https://doi.org/10.1016/j.caeai.2024.100243</bibtext> </blist> <blist> <bibtext> Sulecio de Alvarez, M. S., &amp; Dickson‐Deane, C. (2018). Avoiding educational technology pitfalls for inclusion and equity. TechTrends, 62 (4), 345 – 353. https://doi.org/10.1007/s11528‐018‐0270‐0</bibtext> </blist> <blist> <bibtext> Sutherland, K., Brock, G., de Villiers Scheepers, M. J., Millear, P. M., Norman, S., Strohfeldt, T., Downer, T., Masters, N., &amp; Black, A. L. (2024). Non‐traditional students' preferences for learning technologies and impacts on academic self‐efficacy. Journal of Computing in Higher Education, 36 (2), 298 – 319. https://doi.org/10.1007/s12528‐023‐09354‐5</bibtext> </blist> <blist> <bibtext> Traynor, A., Li, T., &amp; Zhou, S. (2020). Gauging uncertainty in test‐to‐curriculum alignment indices. Applied Measurement in Education, 33 (2), 141 – 158. https://doi.org/10.1080/08957347.2020.1732387</bibtext> </blist> <blist> <bibtext> Uddin, S. J., Albert, A., Ovid, A., &amp; Alsharef, A. (2023). Leveraging ChatGPT to aid construction hazard recognition and support safety education and training. Sustainability, 15 (9), 7121. https://doi.org/10.3390/su15097121</bibtext> </blist> <blist> <bibtext> Webb, N. L. (2002). Depth‐of‐knowledge levels for four content areas. Language Arts, 28, 1 – 9.</bibtext> </blist> <blist> <bibtext> White, J., Fu, Q., Hays, S., Sandborn, M., Olea, C., Gilbert, H., Elnashar, A., Spencer‐Smith, J., &amp; Schmidt, D. C. (2023). A prompt pattern catalog to enhance prompt engineering with ChatGPT. arXiv preprint arXiv:2302.11382.</bibtext> </blist> <blist> <bibtext> Wittwer, J., &amp; Renkl, A. (2010). How effective are instructional explanations in example‐based learning? A meta‐analytic review. Educational Psychology Review, 22 (4), 393 – 409. https://doi.org/10.1007/s10648‐010‐9136‐5</bibtext> </blist> <blist> <bibtext> Wyse, A. E., &amp; Viger, S. G. (2011). How item writers understand depth of knowledge. Educational Assessment, 16 (4), 185 – 206. https://doi.org/10.1080/10627197.2011.634286</bibtext> </blist> <blist> <bibtext> Yan, L., Martinez‐Maldonado, R., &amp; Gašević, D. (2024). Generative artificial intelligence in learning analytics: Contextualising opportunities and challenges through the learning analytics cycle. In Proceedings of the 14th International Conference on Learning Analytics &amp; Knowledge (pp. 101 – 111). https://doi.org/10.1145/3636555.3636856</bibtext> </blist> <blist> <bibtext> Zamfirescu‐Pereira, J. D., Wong, R. Y., Hartmann, B., &amp; Yang, Q. (2023). Why Johnny can't prompt: How non‐AI experts try (and fail) to design LLM prompts. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1 – 21). https://doi.org/10.1145/3544548.3581388</bibtext> </blist> <blist> <bibtext> Zawacki‐Richter, O., Marín, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education—Where are the educators? International Journal of Educational Technology in Higher Education, 16 (1), 1 – 27. https://doi.org/10.1186/s41239‐019‐0171‐0</bibtext> </blist> <blist> <bibtext> Zhai, C., Wibowo, S., &amp; Li, L. D. (2024). The effects of over‐reliance on AI dialogue systems on students' cognitive abilities: A systematic review. Smart Learning Environments, 11 (1), 28. https://doi.org/10.1186/s40561‐024‐00316‐7</bibtext> </blist> <blist> <bibtext> Zhang, P., &amp; Tur, G. (2023). A systematic review of ChatGPT use in K‐12 education. European Journal of Education, 59 (2), e12599. https://doi.org/10.1111/ejed.12599</bibtext> </blist> <blist> <bibtext> Zheng, B., &amp; Warschauer, M. (2015). Participation, interaction, and academic achievement in an online discussion environment. Computers &amp; Education, 84, 78 – 89. https://doi.org/10.1016/j.compedu.2015.01.008</bibtext> </blist> </ref> <aug> <p>By Mohan Yang; Shiyan Jiang; Belle Li; Kristin Herman; Tian Luo; Shanan Chappell Moots and Nolan Lovett</p> <p>Reported by Author; Author; Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib18" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib70" firstref="ref3"></nolink> <nolink nlid="nl3" bibid="bib34" firstref="ref4"></nolink> <nolink nlid="nl4" bibid="bib13" firstref="ref8"></nolink> <nolink nlid="nl5" bibid="bib15" firstref="ref9"></nolink> <nolink nlid="nl6" bibid="bib38" firstref="ref10"></nolink> <nolink nlid="nl7" bibid="bib56" firstref="ref12"></nolink> <nolink nlid="nl8" bibid="bib49" firstref="ref13"></nolink> <nolink nlid="nl9" bibid="bib22" firstref="ref15"></nolink> <nolink nlid="nl10" bibid="bib36" firstref="ref16"></nolink> <nolink nlid="nl11" bibid="bib41" firstref="ref18"></nolink> <nolink nlid="nl12" bibid="bib62" firstref="ref19"></nolink> <nolink nlid="nl13" bibid="bib46" firstref="ref20"></nolink> <nolink nlid="nl14" bibid="bib17" firstref="ref21"></nolink> <nolink nlid="nl15" bibid="bib51" firstref="ref23"></nolink> <nolink nlid="nl16" bibid="bib35" firstref="ref24"></nolink> <nolink nlid="nl17" bibid="bib10" firstref="ref25"></nolink> <nolink nlid="nl18" bibid="bib58" firstref="ref27"></nolink> <nolink nlid="nl19" bibid="bib72" firstref="ref29"></nolink> <nolink nlid="nl20" bibid="bib40" firstref="ref32"></nolink> <nolink nlid="nl21" bibid="bib71" firstref="ref33"></nolink> <nolink nlid="nl22" bibid="bib52" firstref="ref35"></nolink> <nolink nlid="nl23" bibid="bib68" firstref="ref36"></nolink> <nolink nlid="nl24" bibid="bib67" firstref="ref38"></nolink> <nolink nlid="nl25" bibid="bib42" firstref="ref39"></nolink> <nolink nlid="nl26" bibid="bib24" firstref="ref40"></nolink> <nolink nlid="nl27" bibid="bib33" firstref="ref41"></nolink> <nolink nlid="nl28" bibid="bib65" firstref="ref42"></nolink> <nolink nlid="nl29" bibid="bib45" firstref="ref43"></nolink> <nolink nlid="nl30" bibid="bib20" firstref="ref44"></nolink> <nolink nlid="nl31" bibid="bib74" firstref="ref45"></nolink> <nolink nlid="nl32" bibid="bib66" firstref="ref47"></nolink> <nolink nlid="nl33" bibid="bib26" firstref="ref48"></nolink> <nolink nlid="nl34" bibid="bib27" firstref="ref49"></nolink> <nolink nlid="nl35" bibid="bib21" firstref="ref51"></nolink> <nolink nlid="nl36" bibid="bib64" firstref="ref52"></nolink> <nolink nlid="nl37" bibid="bib69" firstref="ref53"></nolink> <nolink nlid="nl38" bibid="bib50" firstref="ref54"></nolink> <nolink nlid="nl39" bibid="bib43" firstref="ref55"></nolink> <nolink nlid="nl40" bibid="bib12" firstref="ref56"></nolink> <nolink nlid="nl41" bibid="bib16" firstref="ref57"></nolink> <nolink nlid="nl42" bibid="bib31" firstref="ref65"></nolink> <nolink nlid="nl43" bibid="bib59" firstref="ref66"></nolink> <nolink nlid="nl44" bibid="bib23" firstref="ref67"></nolink> <nolink nlid="nl45" bibid="bib54" firstref="ref69"></nolink> <nolink nlid="nl46" bibid="bib11" firstref="ref71"></nolink> <nolink nlid="nl47" bibid="bib37" firstref="ref72"></nolink> <nolink nlid="nl48" bibid="bib14" firstref="ref73"></nolink> <nolink nlid="nl49" bibid="bib32" firstref="ref76"></nolink> <nolink nlid="nl50" bibid="bib53" firstref="ref77"></nolink> <nolink nlid="nl51" bibid="bib25" firstref="ref78"></nolink> <nolink nlid="nl52" bibid="bib60" firstref="ref79"></nolink> <nolink nlid="nl53" bibid="bib61" firstref="ref80"></nolink> <nolink nlid="nl54" bibid="bib73" firstref="ref81"></nolink> <nolink nlid="nl55" bibid="bib48" firstref="ref82"></nolink> <nolink nlid="nl56" bibid="bib19" firstref="ref83"></nolink> <nolink nlid="nl57" bibid="bib39" firstref="ref84"></nolink> <nolink nlid="nl58" bibid="bib55" firstref="ref85"></nolink> <nolink nlid="nl59" bibid="bib28" firstref="ref86"></nolink> <nolink nlid="nl60" bibid="bib30" firstref="ref88"></nolink> <nolink nlid="nl61" bibid="bib47" firstref="ref89"></nolink> <nolink nlid="nl62" bibid="bib75" firstref="ref90"></nolink> <nolink nlid="nl63" bibid="bib44" firstref="ref92"></nolink> <nolink nlid="nl64" bibid="bib57" firstref="ref93"></nolink> <nolink nlid="nl65" bibid="bib29" firstref="ref94"></nolink> <nolink nlid="nl66" bibid="bib63" firstref="ref97"></nolink> |
|---|---|
| Header | DbId: eric DbLabel: ERIC An: EJ1480022 AccessLevel: 3 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Analysing Nontraditional Students' ChatGPT Interaction, Engagement, Self-Efficacy and Performance: A Mixed-Methods Approach – Name: Language Label: Language Group: Lang Data: English – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Mohan+Yang%22">Mohan Yang</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0003-0856-0814">0000-0003-0856-0814</externalLink>)<br /><searchLink fieldCode="AR" term="%22Shiyan+Jiang%22">Shiyan Jiang</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0003-4781-846X">0000-0003-4781-846X</externalLink>)<br /><searchLink fieldCode="AR" term="%22Belle+Li%22">Belle Li</searchLink><br /><searchLink fieldCode="AR" term="%22Kristin+Herman%22">Kristin Herman</searchLink><br /><searchLink fieldCode="AR" term="%22Tian+Luo%22">Tian Luo</searchLink><br /><searchLink fieldCode="AR" term="%22Shanan+Chappell+Moots%22">Shanan Chappell Moots</searchLink><br /><searchLink fieldCode="AR" term="%22Nolan+Lovett%22">Nolan Lovett</searchLink> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="SO" term="%22British+Journal+of+Educational+Technology%22"><i>British Journal of Educational Technology</i></searchLink>. 2025 56(5):1973-2000. – Name: Avail Label: Availability Group: Avail Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us – Name: PeerReviewed Label: Peer Reviewed Group: SrcInfo Data: Y – Name: Pages Label: Page Count Group: Src Data: 28 – Name: DatePubCY Label: Publication Date Group: Date Data: 2025 – Name: TypeDocument Label: Document Type Group: TypDoc Data: Journal Articles<br />Reports - Research – Name: Audience Label: Education Level Group: Audnce Data: <searchLink fieldCode="EL" term="%22Higher+Education%22">Higher Education</searchLink><br /><searchLink fieldCode="EL" term="%22Postsecondary+Education%22">Postsecondary Education</searchLink> – Name: Subject Label: Descriptors Group: Su Data: <searchLink fieldCode="DE" term="%22Nontraditional+Students%22">Nontraditional Students</searchLink><br /><searchLink fieldCode="DE" term="%22Artificial+Intelligence%22">Artificial Intelligence</searchLink><br /><searchLink fieldCode="DE" term="%22Technology+Uses+in+Education%22">Technology Uses in Education</searchLink><br /><searchLink fieldCode="DE" term="%22Man+Machine+Systems%22">Man Machine Systems</searchLink><br /><searchLink fieldCode="DE" term="%22Interaction%22">Interaction</searchLink><br /><searchLink fieldCode="DE" term="%22Learner+Engagement%22">Learner Engagement</searchLink><br /><searchLink fieldCode="DE" term="%22Self+Efficacy%22">Self Efficacy</searchLink><br /><searchLink fieldCode="DE" term="%22Academic+Achievement%22">Academic Achievement</searchLink><br /><searchLink fieldCode="DE" term="%22Performance%22">Performance</searchLink><br /><searchLink fieldCode="DE" term="%22College+Students%22">College Students</searchLink><br /><searchLink fieldCode="DE" term="%22Predictor+Variables%22">Predictor Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Prompting%22">Prompting</searchLink><br /><searchLink fieldCode="DE" term="%22Novices%22">Novices</searchLink><br /><searchLink fieldCode="DE" term="%22Resistance+%28Psychology%29%22">Resistance (Psychology)</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Attitudes%22">Computer Attitudes</searchLink><br /><searchLink fieldCode="DE" term="%22Adoption+%28Ideas%29%22">Adoption (Ideas)</searchLink><br /><searchLink fieldCode="DE" term="%22Assignments%22">Assignments</searchLink><br /><searchLink fieldCode="DE" term="%22Digital+Literacy%22">Digital Literacy</searchLink> – Name: DOI Label: DOI Group: ID Data: 10.1111/bjet.13588 – Name: ISSN Label: ISSN Group: ISSN Data: 0007-1013<br />1467-8535 – Name: Abstract Label: Abstract Group: Ab Data: Generative artificial intelligence brings opportunities and unique challenges to nontraditional higher education students, stemming, in part, from the experience of the digital divide. Providing access and practice is critical to bridge this divide and equip students with needed digital competencies. This mixed-methods study investigated how nontraditional higher education students interact with ChatGPT in multiple courses and examined relationships between ChatGPT interactions, engagement, self-efficacy and performance. Data were collected from 73 undergraduate and graduate students through chat logs, course reflections and artefacts, surveys and interviews. ChatGPT interactions were analysed using four metrics: prompt number, depth of knowledge (DoK), prompt relevance and originality. Results showed that ChatGPT prompt numbers ([beta] = 0.256, p < 0.03) and engagement ([beta] = 0.267, p < 0.05) significantly predicted performance, while self-efficacy did not. Students' DoK (r = 0.40, p < 0.01) and prompt relevance (r = 0.42, p < 0.01) were positively correlated with performance. Text mining analysis identified distinct interaction patterns, with 'strategic inquirers' demonstrating significantly higher performance than 'exploratory inquirers' through more sophisticated follow-up questioning. Qualitative findings revealed that while most students were first-time ChatGPT users who initially showed resistance, they developed growing acceptance. Still, students tended to use ChatGPT sparingly and, even then, as only a starting point for assignments. The study highlights the need for targeted guidance in prompt engineering and AI literacy training to help nontraditional higher education students leverage ChatGPT more effectively for higher-order thinking tasks. – Name: AbstractInfo Label: Abstractor Group: Ab Data: As Provided – Name: DateEntry Label: Entry Date Group: Date Data: 2025 – Name: AN Label: Accession Number Group: ID Data: EJ1480022 |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1480022 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1111/bjet.13588 Languages: – Text: English PhysicalDescription: Pagination: PageCount: 28 StartPage: 1973 Subjects: – SubjectFull: Nontraditional Students Type: general – SubjectFull: Artificial Intelligence Type: general – SubjectFull: Technology Uses in Education Type: general – SubjectFull: Man Machine Systems Type: general – SubjectFull: Interaction Type: general – SubjectFull: Learner Engagement Type: general – SubjectFull: Self Efficacy Type: general – SubjectFull: Academic Achievement Type: general – SubjectFull: Performance Type: general – SubjectFull: College Students Type: general – SubjectFull: Predictor Variables Type: general – SubjectFull: Prompting Type: general – SubjectFull: Novices Type: general – SubjectFull: Resistance (Psychology) Type: general – SubjectFull: Computer Attitudes Type: general – SubjectFull: Adoption (Ideas) Type: general – SubjectFull: Assignments Type: general – SubjectFull: Digital Literacy Type: general Titles: – TitleFull: Analysing Nontraditional Students' ChatGPT Interaction, Engagement, Self-Efficacy and Performance: A Mixed-Methods Approach Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Mohan Yang – PersonEntity: Name: NameFull: Shiyan Jiang – PersonEntity: Name: NameFull: Belle Li – PersonEntity: Name: NameFull: Kristin Herman – PersonEntity: Name: NameFull: Tian Luo – PersonEntity: Name: NameFull: Shanan Chappell Moots – PersonEntity: Name: NameFull: Nolan Lovett IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 09 Type: published Y: 2025 Identifiers: – Type: issn-print Value: 0007-1013 – Type: issn-electronic Value: 1467-8535 Numbering: – Type: volume Value: 56 – Type: issue Value: 5 Titles: – TitleFull: British Journal of Educational Technology Type: main |
| ResultId | 1 |
