View in EDS HTML Full Text PDF Full Text

Training Humans to Detect Children's Lies through Their Facial Expressions

Saved in:
Bibliographic Details
Title: Training Humans to Detect Children's Lies through Their Facial Expressions
Language: English
Authors: Alison M. O'Connor, Jennifer Gongola, Kaila C. Bruer, Thomas D. Lyon, Angela D. Evans
Source: Applied Cognitive Psychology. 2025 39(1).
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: 7
Publication Date: 2025
Sponsoring Agency: Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) (DHHS/NIH)
Contract Number: HD101617
Document Type: Journal Articles
Reports - Research
Descriptors: Deception, Nonverbal Communication, Recognition (Psychology), Children, Interviews, Accuracy, Ethics, Artificial Intelligence, Comparative Analysis, Training, Video Technology, Adults
DOI: 10.1002/acp.70024
ISSN: 0888-4080
1099-0720
Abstract: The accurate detection of children's truthful and dishonest reports is essential as children can serve as important providers of information. Research using automated facial coding and machine learning found that children who were asked to lie about an event were more likely to look surprised when hearing the first question during an interview about said event. The present studies explored if humans can be trained to look for surprised expressions to detect children's deception. Participants made lie-detection judgments after seeing children's expressions in very brief clips. In Study 1, we compared performance across a training condition and control condition, and in Study 2 we modified the training. With training, adults could detect children's lies at above-chance levels by viewing their facial expressions. Detection accuracy was further improved with modified training (Study 2), but participants held a consistent lie bias. Challenges with using facial expressions to detect deceit are discussed.
Abstractor: As Provided
Entry Date: 2025
Accession Number: EJ1460825
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwFiA_L2o6lRsIxCgQQYkTZYAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDALB0ErViWxO22Yg2AIBEICBm7UugHFxvRlOaxZRQfh01s_3IHDexZUQEX5rsLN-Xvgh1oRahzTInuF6xRDu7OX6cYGRg24Ev4KISdObZtmTJjbo4KeHvAnOd4HpHIZ9Edc3QSkKMHneBwDsn5MSent9YcxNxzg-iA9MDNCUrgnHTcMxdNtDcxykjfzlB-kPHPDjSHcBe1A-YE46YlzuXMcvW6WlOhLugHlj9L7S
Text:
  Availability: 1
  Value: <anid>AN0183757202;bu801jan.25;2025Mar19.05:43;v2.2.500</anid> <title id="AN0183757202-1">Training Humans to Detect Children's Lies Through Their Facial Expressions </title> <p>The accurate detection of children's truthful and dishonest reports is essential as children can serve as important providers of information. Research using automated facial coding and machine learning found that children who were asked to lie about an event were more likely to look surprised when hearing the first question during an interview about said event. The present studies explored if humans can be trained to look for surprised expressions to detect children's deception. Participants made lie‐detection judgments after seeing children's expressions in very brief clips. In Study 1, we compared performance across a training condition and control condition, and in Study 2 we modified the training. With training, adults could detect children's lies at above‐chance levels by viewing their facial expressions. Detection accuracy was further improved with modified training (Study 2), but participants held a consistent lie bias. Challenges with using facial expressions to detect deceit are discussed.</p> <p>Keywords: child; cue; facial expression; lie detection; surprise</p> <hd id="AN0183757202-2">Introduction</hd> <p>Previous research has reliably established that adults are not highly skilled at discriminating between children's dishonest and honest reports (54% overall accuracy, 48% lie accuracy, 64% truth accuracy, with a truth bias; Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref1">8</reflink>] meta‐analysis). This unimpressive accuracy rate may be problematic in legal contexts as the inaccurate detection of truthful and dishonest reports can either result in a failure to protect child abuse victims or a false accusation (Bala et al. [<reflink idref="bib1" id="ref2">1</reflink>]; Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref3">8</reflink>]). Children can be called to serve as witnesses in cases of child abuse, family conflict, and incidents of bullying (Fansher and del Carmen [<reflink idref="bib6" id="ref4">6</reflink>]); therefore, it is important to enhance adults' accurate detection of children's true and false reports. Indeed, researchers have explored various techniques to improve human detection accuracy (e.g., Mac Giolla and Luke [<reflink idref="bib13" id="ref5">13</reflink>]; Porter et al. [<reflink idref="bib15" id="ref6">15</reflink>]; Vrij et al. [<reflink idref="bib19" id="ref7">19</reflink>]). For example, researchers have explored how to train humans to detect changes in physical movements, voice pitch, and verbal content as indicators of potential deception (Hauch et al. [<reflink idref="bib10" id="ref8">10</reflink>]; Hartwig and Bond Jr. [<reflink idref="bib9" id="ref9">9</reflink>]). Some training has resulted in small to medium improvements (when detecting adults' lies; Hauch et al. [<reflink idref="bib10" id="ref10">10</reflink>]), but limited research has explored training with children's lies.</p> <p>Moving beyond the ability of <emph>human</emph> lie detectors, recent research has explored if we can use automated facial expression coding to detect subtle changes in facial muscles to more accurately detect children's deceit (Zanette et al. [<reflink idref="bib23" id="ref11">23</reflink>]; Bruer et al. [<reflink idref="bib3" id="ref12">3</reflink>]). Various cognitive theories (e.g., Walczyk et al. [<reflink idref="bib22" id="ref13">22</reflink>], [<reflink idref="bib21" id="ref14">21</reflink>]) suggest that telling lies is cognitively taxing and requires activation of cognitive skills (e.g., memory, inhibitory control). As children are still developing these cognitive skills, they may be less able to conceal various deception cues (such as indicators in their facial expressions), and adults may be able to use such cues to identify deception. Children also have less control over their facial musculature, suggesting that changes in their facial expressions may be harder to control both cognitively and physically (Bruer et al. [<reflink idref="bib3" id="ref15">3</reflink>]; Ekman, Roper, and Hager [<reflink idref="bib5" id="ref16">5</reflink>]; Feldman, Jenkins, and Popoola [<reflink idref="bib7" id="ref17">7</reflink>]). For example, Bruer et al. ([<reflink idref="bib3" id="ref18">3</reflink>]) employed an automated facial action coding system (FACET; Bartlett et al. [<reflink idref="bib2" id="ref19">2</reflink>]; iMotions [<reflink idref="bib11" id="ref20">11</reflink>]) on videos of child interviews where children were questioned about an event with a confederate (playing with toys). For some of the children, the toys broke, and children were asked to not tell anyone (i.e., the lie‐tellers) while for others, the toys did not break and so they truthfully reported the event. FACET provided a frame‐by‐frame analysis of facial muscle movements in response to the first interview question ("tell me everything that happened...") and generated 10 emotion expression scores (joy, sadness, anger, disgust, surprise, fear, confusion, frustration, contempt, and neutral). The researchers used machine learning to examine if these facial expressions could distinguish the liars and truth‐tellers. Together, all emotion scores could distinguish liars from truth‐tellers with a 73% accuracy rate. However, when exploring which specific emotions were helpful in detecting lies, surprise and fear expressions could predict which children were lying, with surprise expressions serving as a stronger predictor of deception than fear. There was a 30% chance of liars showing a surprised expression but only a 3% chance of a truth‐teller showing a surprised expression. Results from this study suggest that children who concealed a transgression (breaking toys) showed more surprised expressions upon being asked the very first question about this event. Truth‐tellers, who had nothing to hide about the event, were much less likely to show this surprised expression at the start of their interview.</p> <p>Given that Bruer and colleagues identified <emph>surprise</emph> as an important differentiator between truth and lie‐tellers, and that machine learning could use this to systematically predict whether children were lying, the present study examined whether this surprise cue can also help humans detect children's lies. Considering that most adults do not have access to the required technology for automated facial coding, it is important to explore if this technique can be useful for the human detection of children's lies to expand the application of this detection strategy. Adults can struggle to accurately detect the presence of adults' automated facial expressions (Döllinger et al. [<reflink idref="bib4" id="ref21">4</reflink>]; Monaro et al. [<reflink idref="bib14" id="ref22">14</reflink>]); therefore, adults may also struggle to detect children's expressions, albeit it may be easier given children's underdeveloped cognitive and physical control systems. Moreover, adults may struggle to be accurate lie‐detectors because of the high cognitive demands of assessing many potential deception cues (Monaro et al. [<reflink idref="bib14" id="ref23">14</reflink>]; Vrij et al. [<reflink idref="bib20" id="ref24">20</reflink>]); therefore, we focused our training on one specific cue (surprise) to try and reduce the cognitive demands.</p> <hd id="AN0183757202-3">The Present Research</hd> <p>The present study sought to build upon the work by Bruer et al. ([<reflink idref="bib3" id="ref25">3</reflink>]) by exploring if humans can be trained to use the surprised expression to accurately detect children's lies. Across two studies, adults watched videos of child interviews (a subset from Bruer et al. [<reflink idref="bib3" id="ref26">3</reflink>]) and were asked to judge if the child was lying or telling the truth. Prior to watching the videos, some participants were trained to identify the surprise cue as a cue to dishonesty. In Study 1, we compared performance across two conditions: a control condition with no training and a training condition where participants were trained to identify surprise as a cue to deception. In Study 2, we provided a modified training procedure in an attempt to further improve performance. We predicted that adults' accuracy in the training conditions would be greater than chance given the specific training provided, and as is typical in child lie‐detection research (Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref27">8</reflink>]), that participants would hold a truth bias. We also predicted that lie‐detection performance would be higher in the training group than the control group in Study 1, and that performance would be further enhanced with the modified training in Study 2 compared to the training group in Study 1.</p> <hd id="AN0183757202-4">Study 1</hd> <p></p> <hd id="AN0183757202-5">Method</hd> <p></p> <hd id="AN0183757202-6">Participants</hd> <p>A total of 426 adults participated in Study 1 (<emph>M</emph><subs>age</subs> = 38.27, SD = 12.16, 53% male) from an online participant pool (Amazon's Mechanical Turk; MTurk). All participants were American citizens. Thirty‐five participants were excluded for failing the attention check in the survey. Thus, the analyses were conducted on the remaining 391 participants (<emph>M</emph><subs>age</subs> = 38.30, SD = 11.98, 52% male). Sample size was determined by recommendations (Levine, Daiku, and Masip [<reflink idref="bib12" id="ref28">12</reflink>]) and a power analysis in G*Power that determined a minimum sample of 176 per group to detect a small effect in a <emph>t</emph>‐test (alpha = 0.05, power = 0.80). Participants were randomly assigned to a control condition (<emph>N</emph> = 190, <emph>M</emph><subs>age</subs> = 38.31, SD = 12.69, 53% male) or a training condition (<emph>N</emph> = 201, <emph>M</emph><subs>age</subs> = 38.29, SD = 11.22, 51% male), Participants were 70% White, 18% Black, 6% multiracial, and less than 2% were each Latinx, East Asian, South Asian, Native American, South East Asian, and West Asian. Approximately 6% completed high school as their highest education level, 12% completed some college or university, 44% completed an undergraduate degree, and 29% completed graduate school or a professional degree. The majority of participants (73%) did not have children.</p> <hd id="AN0183757202-7">Materials</hd> <p></p> <hd id="AN0183757202-8">Child Interview Videos</hd> <p>Parental consent was provided for the use of the child's video in future studies. A subset of 10 videos (following recommendations from Levine, Daiku, and Masip [<reflink idref="bib12" id="ref29">12</reflink>]) was selected from the videos used by Bruer et al. ([<reflink idref="bib3" id="ref30">3</reflink>]). The videos depicted children aged 5–9 years being interviewed after a play session with an adult confederate. During the play session, children either experienced the toys breaking and were asked not to tell anyone (breakage condition) or a control condition where no breakage occurred (non‐breakage condition). Half of the videos selected included concealers (breakage condition) and half‐truth‐tellers (non‐breakage condition).</p> <p>Bruer et al. ([<reflink idref="bib3" id="ref31">3</reflink>]) found that the FACET software could best distinguish liars and truth‐tellers from facial expressions to the first interview question ("tell me everything that happened when the person came in while I was gone"); therefore, the video recordings in the present study were edited to show only the child's reaction to this first question. The videos ranged from 3 to 4 s (<emph>M</emph> = 3.8 s, SD = 0.42) and were clipped immediately following the first interview question (i.e., children's verbal responses to this question were not included). Given that adults struggle to make accurate veracity judgments, we attempted to maximize potential performance by selecting the videos that depicted concealers with the highest surprise expression scores averaged over all the frames present in the video and truth‐tellers with the lowest surprise expression scores (from FACET coding by Bruer et al. [<reflink idref="bib3" id="ref32">3</reflink>]). The order of the videos was counterbalanced across 10 order conditions in a Latin Square design. The videos were programmed to begin and end automatically, and participants could not pause or replay the videos.</p> <hd id="AN0183757202-9">Surprise Cue Training</hd> <p>Training was provided through text format, and the training instructions are available in Supporting Information. All participants were given a brief overview of the play session and were told that, in the videos, some children kept a secret about broken toys and some children told the truth and did not have a secret to keep. Participants were told that they would see the child's reaction to their first interview question and that they needed to decide whether the child would keep a secret or not based solely on their reaction to this first interview question. Participants in the control condition then proceeded to watch the videos and provide lie‐detection judgments.</p> <p>In addition to the control instructions, participants in the training condition were told that children who are keeping a secret are more likely to look surprised when they hear the first interview question. Participants were told that a surprise expression can include raised eyebrows, an open mouth, and the eyes can widen. Two generic photos of surprised expressions were provided (one of a child and one of an adult) but participants were reminded that the surprised expressions in the videos can be very brief and subtle and may not contain all features; therefore, participants should pay close attention to small changes in the child's expression. Following this information, participants in the training condition then proceeded to provide lie‐detection judgments.</p> <hd id="AN0183757202-10">Lie Detection Judgments</hd> <p>Following each video, participants were asked if they thought the child was hiding toy breakage on a scale from 1 (<emph>the child is definitely not hiding breakage</emph>) to 6 (<emph>the child is definitely hiding breakage</emph>). Responses were dichotomized such that scores 1–3 were categorized as a "truth" judgment (i.e., not hiding breakage) and scores 4–6 were categorized as a "lie" judgment (i.e., hiding breakage).</p> <hd id="AN0183757202-11">Procedure</hd> <p>Participants completed this study as an online experiment through Testable (testable.org). All participants provided informed consent before beginning the study. Participation took approximately 15 min and participants received $1 USD. The present research was approved by the University of Southern California research ethics board. Data and survey materials associated with this research is available upon reasonable request to the corresponding author.</p> <hd id="AN0183757202-12">Results</hd> <p>Detection accuracy was measured across a series of variables (overall accuracy across all trials, accuracy on lie trials, and accuracy on truth trials) and compared to chance. In addition, these accuracy scores were used for signal detection analyses to measure <emph>d</emph> prime (discrimination scores (<emph>d'</emph>); i.e., how well one could discriminate between truth and lie videos) and criterion c (bias scores; i.e., if one was biased to rate the children in the videos as truth‐tellers or liars). These scores were computed used the Excel scoring formulas in Stanislaw and Todorov ([<reflink idref="bib17" id="ref33">17</reflink>]). <emph>D′</emph> values were calculated by subtracting standardized false alarm rates (proportion of truth‐tellers misclassified as liars) from standardized hit rates (proportion of liars classified as liars). Higher <emph>d</emph>' scores indicate greater discriminatory ability. Criterion c scores were calculated by adding standardized hit and false alarm rates and multiplying by −0.5. Positive criterion c values indicate a truth bias, negative scores indicate a lie bias, and scores farther from zero indicate stronger biases. Any perfectly accurate (<reflink idref="bib1" id="ref34">1</reflink>) and inaccurate (0) rates were adjusted by.</p> <p>1‐1/(2 N) and 1/(2 N), respectively, where <emph>N</emph> is 5 (the number of truth or lie trials). See Stanislaw and Todorov ([<reflink idref="bib17" id="ref35">17</reflink>]) for further information on scoring and signal detection analyses.</p> <hd id="AN0183757202-13">Detection Accuracy and Chance Analyses</hd> <p>We examined if overall, lie, and truth accuracy rates were significantly different from chance (50%) across the different conditions. All means are available in Table 1. In the control condition (i.e., baseline performance with no training), overall accuracy was not significantly different from chance, <emph>t</emph>(<reflink idref="bib189" id="ref36">189</reflink>) = 2.00, <emph>p</emph> = 0.050, <emph>d</emph> = 0.14, lie accuracy was significantly above chance, <emph>t</emph>(<reflink idref="bib189" id="ref37">189</reflink>) = 3.49, <emph>p</emph> < 0.001, <emph>d</emph> = 0.25, and truth accuracy was significantly below chance, <emph>t</emph>(<reflink idref="bib189" id="ref38">189</reflink>) = 5.31, <emph>p</emph> < 0.001, <emph>d</emph> = 0.39. We also compared overall accuracy to 54% (the average overall accuracy rate in prior detection studies; Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref39">8</reflink>]). Overall accuracy in the control condition was significantly lower than average (54%), <emph>t</emph>(<reflink idref="bib189" id="ref40">189</reflink>) = 5.92, <emph>p</emph> < 0.001, <emph>d</emph> = 0.43.</p> <p>1 TABLE Average lie‐detection performance across conditions in Study 1 and Study 2.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left" /><th align="center">Condition</th><th align="center">Overall accuracy</th><th align="center">Lie accuracy</th><th align="center">Truth accuracy</th><th align="center"><italic>d'</italic></th><th align="center">Criterion c</th></tr></thead><tbody valign="top"><tr><td align="left">Study 1</td><td align="center">Control</td><td align="center">0.48 (0.14)</td><td align="center">0.56 (0.25)</td><td align="center">0.40 (0.27)</td><td align="center">−0.11 (0.75)</td><td align="center">−0.22 (0.58)</td></tr><tr><td align="left" /><td align="center">Training</td><td align="center">0.52 (0.14)</td><td align="center">0.58 (0.28)</td><td align="center">0.46 (0.31)</td><td align="center">0.12 (0.76)</td><td align="center">−0.15 (0.68)</td></tr><tr><td align="left">Study 2</td><td align="center">Modified Training</td><td align="center">0.56 (0.13)</td><td align="center">0.57 (0.21)</td><td align="center">0.56 (20)</td><td align="center">0.48 (0.69)</td><td align="center">−0.08 (0.46)</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note:</emph> Standard deviations are included in brackets.</p> <p>In the training condition, overall accuracy, <emph>t</emph>(<reflink idref="bib200" id="ref41">200</reflink>) = 2.14, <emph>p</emph> = 0.034, <emph>d</emph> = 0.15, and lie accuracy, <emph>t</emph>(<reflink idref="bib200" id="ref42">200</reflink>) = 4.02, <emph>p</emph> < 0.001, <emph>d</emph> = 0.28, were significantly above chance levels (50%). Truth accuracy was not significantly different from chance, <emph>t</emph>(<reflink idref="bib200" id="ref43">200</reflink>) = 1.73, <emph>p</emph> = 0.085, <emph>d</emph> = 0.12. Although overall accuracy was above chance levels, it did not significantly differ from average performance (54%), <emph>t</emph>(<reflink idref="bib200" id="ref44">200</reflink>) = 1.86, <emph>p</emph> = 0.064, <emph>d</emph> = 0.13.</p> <hd id="AN0183757202-14">Signal Detection Analyses</hd> <p>First, one‐sample t‐tests were conducted to examine whether <emph>d'</emph> and criterion c scores were significantly above zero in each condition. A score of zero indicates no discriminatory ability and no bias in responding. In the control condition, <emph>d'</emph> scores were not significantly different from zero (<emph>p = 0</emph>.050), and criterion c scores were significantly below zero (<emph>p</emph> < 0.001, <emph>d</emph> = 0.38), indicating a lie bias. In the training condition, <emph>d</emph>' scores were significantly above zero (<emph>p</emph> = 0.029, <emph>d</emph> = 0.16), showing an ability to discriminate between truth and lie trials. Criterion c scores were also significantly below zero (<emph>p</emph> = 0.001, <emph>d</emph> = 0.23), indicating a lie bias.</p> <p>Next, a series of independent samples <emph>t</emph>‐tests were conducted to examine whether discrimination and bias scores differed across conditions. Discrimination scores were significantly higher in the training condition compared to the control condition, <emph>t</emph>(<reflink idref="bib389" id="ref45">389</reflink>) = 2.94, <emph>p</emph> = 0.003, <emph>d</emph> = 0.76. Yet, there was no significant condition difference in criterion c scores, <emph>t</emph>(384.84) = 1.04, <emph>p</emph> = 0.299, <emph>d</emph> = 0.11, indicating that a similar lie bias was found regardless of surprise cue training.</p> <hd id="AN0183757202-15">Discussion</hd> <p>The results of Study 1 found that, with training to understand and look for children's surprised expressions, detection performance improved (relative to those without the training); however, accuracy rates still hovered just above chance levels, and participants held a lie bias regardless of training. These accuracy rates can be interpreted in two ways. On the one hand, this suggests that relying solely on the detection of surprised facial expressions to detect children's lies was a challenging task for humans, and this may not be a technique suitable for lay people. Yet, on the other hand, detection performance in the training condition was somewhat comparable to the typical human detection performance found when adults have access to full interviews (e.g., 52% overall accuracy in our training condition compared with 54% overall accuracy found in the Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref46">8</reflink>] meta‐analysis). This suggests detection performance can look quite similar when participants have access to full videos (in past research) and 3‐ to 4‐ second clips (in the present research).</p> <p>Although we found the typical slightly above‐chance accuracy rate in our training condition, we unexpectedly found that participants showed a lie bias across both conditions. Considering how challenging it can be to identify these subtle expressions, especially when the video is being played at regular speed, participants may have been sensitive to facial movement and interpreted other expressions or facial changes as a surprised expressions, resulting in more lie judgments and a lie bias. It is also possible that our training procedure emphasized the detection of the surprised expression (to detect liars) without equal emphasis on the detection of a neutral or non‐surprised expression (to detect the truth‐tellers). Thus, based on performance in Study 1, we implemented a modified training procedure in Study 2 that sought to help enhance detection performance overall and reduce the lie bias.</p> <hd id="AN0183757202-16">Study 2</hd> <p>In Study 2, we tested a new group of participants in a modified training condition to further enhance detection performance and reduce the lie bias. In the modified training condition in Study 2, participants were given practice trials (with feedback), and instructions were modified to emphasize the importance of both detecting the surprise cue (for liars) and the absence of the surprise cue (for truth‐tellers) to reduce the lie bias found in Study 1. We predicted that accuracy would improve, and the lie bias would decrease in this modified training procedure relative to the training condition from Study 1.</p> <hd id="AN0183757202-17">Method</hd> <p></p> <hd id="AN0183757202-18">Participants</hd> <p>A total of 210 new participants participated in Study 2 (<emph>M</emph><subs>age</subs> = 36.48, SD = 12.03, 49% male) from an online participant pool (Prolific). All participants were American citizens. Two participants were excluded for failing the attention check in the survey. Thus, the analyses were conducted on 208 participants (<emph>M</emph><subs>age</subs> = 36.49, SD = 12.08, 49% male). All participants received the modified training. Participants were 73% White, 9% multiracial, 8% Black, 6% multiracial, and 3% or less were Latinx, East Asian, South Asian, Native American, South East Asian, and West Asian. Less than 1% did not complete high school, 14% completed high school as their highest education level, 31% completed some college or university, 9% completed a college diploma, 34% completed an undergraduate degree, and 12% completed graduate school or a professional degree. The majority of participants (63%) did not have children.</p> <hd id="AN0183757202-19">Materials</hd> <p>The materials used in Study 2 were identical to those in Study 1 except for a modified training procedure.</p> <hd id="AN0183757202-20">Modified Training Procedure</hd> <p>The same training materials were used from Study 1, with a few modifications described below. In the instruction phase, the same two generic photos of surprised expressions were shown to participants, but a third photo (of a child's surprise expression from the toy breakage study who was not included as a video in the test trials) was also shown to increase the relevance of the example photos to the test stimuli. Next, the language used in this instruction phase was modified slightly to direct participants' attention to the fact that they would need to both detect the <emph>presence</emph> of a surprise expression (to detect those who are keeping a secret; liars) and the <emph>absence</emph> of a surprise expression (to detect the truth‐tellers). Given that participants in Study 1 were at chance for detecting truth‐tellers, this modification was made to enhance the salience of the truth‐teller (no surprise) detection process. Lastly, the modified training procedure also included a round of practice trials with feedback. Participants completed four practice trials (two truth‐tellers; two liars) presented in a randomized order. The practice trials depicted videos from the same data set (Bruer et al. [<reflink idref="bib3" id="ref47">3</reflink>]). In each practice trial, participants were first shown a still image (captured from the video) of the child showing surprise and not showing surprise. Specifically, using the frame‐by‐frame emotion coding from Bruer et al. ([<reflink idref="bib3" id="ref48">3</reflink>]) we identified the moment where these children had the highest and lowest surprise scores, and we captured images of these moments to present as examples for participants. Then, participants watched the video and indicated if they thought that they saw the surprise expression in the clip or not. Participants were then given feedback indicating if their response was correct or incorrect. On average, participants were 71% accurate across the practice trials. After this modified training procedure, participants began the test trials following an identical procedure to Study 1.</p> <hd id="AN0183757202-21">Results</hd> <p></p> <hd id="AN0183757202-22">Detection Accuracy and Chance Analyses</hd> <p>We examined whether overall, lie, and truth accuracy rates were significantly different from chance (50%). Overall accuracy, <emph>t</emph>(<reflink idref="bib207" id="ref49">207</reflink>) = 7.22, <emph>p</emph> < 0.001, <emph>d</emph> = 0.50, lie accuracy, <emph>t</emph>(<reflink idref="bib207" id="ref50">207</reflink>) = 4.87, <emph>p</emph> < 0.001, <emph>d</emph> = 0.30, and truth accuracy, <emph>t</emph>(<reflink idref="bib207" id="ref51">207</reflink>) = 4.36, <emph>p</emph> < 0.001, <emph>d</emph> = 0.34 were all significantly greater than chance levels. Overall accuracy was also significantly greater than average detection abilities in past studies (54%; Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref52">8</reflink>]), <emph>t</emph>(<reflink idref="bib207" id="ref53">207</reflink>) = 2.80, <emph>p</emph> = 0.006, <emph>d</emph> = 0.19.</p> <p>All means are available in Table 1.</p> <hd id="AN0183757202-23">Signal Detection Analyses</hd> <p>First, we examined whether discrimination scores (<emph>d</emph>') and bias scores (criterion <emph>c</emph>) were significantly different from zero. Discrimination scores were significantly above zero, <emph>t</emph>(<reflink idref="bib207" id="ref54">207</reflink>) = 9.98, <emph>p</emph> < 0.001, <emph>d</emph> = 0.69, indicating that participants could discriminate between the truth‐tellers and concealers. Bias scores were significantly below zero, <emph>t</emph>(<reflink idref="bib207" id="ref55">207</reflink>) = 2.44, <emph>p</emph> = 0.015, <emph>d</emph> = 0.17, indicating that participants held a lie bias. All means are available in Table 1.</p> <p>Next, we explored whether discrimination and bias scores in the modified training condition (Study 2) differed from the original training condition (Study 1). Discrimination scores were significantly higher in the modified training condition than in the original training condition, <emph>t</emph>(<reflink idref="bib407" id="ref56">407</reflink>) = 5.01, <emph>p</emph> < 0.001, <emph>d</emph> = 0.50. However, participants in both training conditions held a similar lie bias, <emph>t</emph>(352.42) = 1.32, <emph>p</emph> = 0.186.</p> <hd id="AN0183757202-24">Discussion</hd> <p>The results of Study 2 indicated that the modified training procedure did enhance detection performance relative to the training in Study 1. The practice trials likely helped participants to understand what a surprised expression could look like and how subtle this could be. The modified instructions to look for both the presence of surprise and the absence of surprise may have also helped to enhance detection performance, but contrary to our prediction, they did not significantly reduce participants' lie biases.</p> <hd id="AN0183757202-25">General Discussion</hd> <p>Across two studies, we sought to explore if humans could be trained to use children's facial expressions (and specifically the presence of a surprised expression) to accurately detect children's truthful and dishonest reports. Study 1 demonstrated that those who were trained on the importance of a surprised expression outperformed those without this knowledge, but even with training, performance was only slightly above chance levels and participants held a lie bias. The training was modified in Study 2, and participants' detection accuracy did improve (and surpassed average performance levels from past research), but the lie bias remained. Together, these results suggest that with training, adults can detect children's lies at above chance levels by only seeing their facial expressions to the first interview question. However, training to detect indicia of lying appeared to bias participants into believing that children are lying.</p> <p>First, it is noteworthy that participants could detect children's lies at above chance levels across the training conditions even though they were only assessing the child's facial expression during the first interview question. Moreover, the overall accuracy rate in the modified training condition was significantly greater than average performance from past research where participants had far more information to inform their judgments (Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref57">8</reflink>]). Yet, the fact that accuracy rates were still only slightly above chance suggests that this novel technique does not necessarily offer a competitive advantage when considering techniques to enhance detection performance <emph>substantially</emph> above chance levels. Even with training, humans in the present study did not perform at the same level as a computer (from Bruer et al. [<reflink idref="bib3" id="ref58">3</reflink>]; similar to Monaro et al. [<reflink idref="bib14" id="ref59">14</reflink>]). Nevertheless, the present results provide novel evidence that adults can use children's facial expressions to help them distinguish between truthful and dishonest reports. Considering adults often struggle to accurately detect lies, there is a need to further develop informed and reliable cues that might indicate deception, and the present study provides one additional cue (looking for a surprised expression) that may aid in adults' detection process. From a cognitive perspective, it has been argued that adults are poor lie detectors because there are many potential indicators of lying to assess at once (Monaro et al. [<reflink idref="bib14" id="ref60">14</reflink>]; Vrij et al. [<reflink idref="bib20" id="ref61">20</reflink>]). Thus, it is also possible that allowing adults to focus on only one deception cue reduced the cognitive demands of the task and increased accuracy.</p> <p>Across our two studies, we also modified the training procedure, and the modifications made in Study 2 helped to improve detection accuracy relative to the training from Study 1. Modifications included creating more specific guidance for detecting lie‐ and truth‐tellers, as well as incorporating practice trials with feedback. Accurately interpreting facial expressions (without automated technology) can be challenging, and we believe the practice trials likely helped totrain participants on what a surprised expression could look like, thereby enhancing detection accuracy on the test trials. Notably, the modified training in Study 2 added emphasis on how to detect truth‐tellers (who did not show surprise) and this improved truth‐accuracy detection rates. This research adds to the growing literature on lie‐detection training (e.g., Hauch et al. [<reflink idref="bib10" id="ref62">10</reflink>]; Hartwig and Bond Jr. [<reflink idref="bib9" id="ref63">9</reflink>]; Monaro et al. [<reflink idref="bib14" id="ref64">14</reflink>]). In particular, prior studies have assessed the use of expressions as a lie‐detection tool with adult liars (e.g., Monaro et al. [<reflink idref="bib14" id="ref65">14</reflink>]; Shen et al. [<reflink idref="bib16" id="ref66">16</reflink>]); therefore, this study provides a novel contribution by exploring how to train adults to detect lies through children's facial expressions, as children may differ in their control over these expressions.</p> <p>We also saw an interesting pattern across all conditions where participants consistently held a lie bias when providing detection judgments. This contrasts with previous research that typically found that adults hold a truth bias when rendering lie‐detection judgments (Gongola, Scurich, and Quas [<reflink idref="bib8" id="ref67">8</reflink>]). It is possible that the process of detecting a specific facial expression (surprise) was quite challenging for our participants and that they misinterpreted other facial expressions as surprise, resulting in more lie judgments. It would be interesting to further modify the training to educate participants on various facial expressions to help them better recognize what surprise looks like compared to, for example, sadness. It is also possible that our training modified participants' base rate assumptions of how many truthful and deceptive videos they would see (similar to what was found in Street and Richardson [<reflink idref="bib18" id="ref68">18</reflink>]). In Study 1, when training emphasized the detection of liars, this may have inflated base rate perceptions of lie videos, and when training in Study 2 put similar emphasis on truth and liars, this may have evened out base rate perceptions. Biases can be problematic as truth biases can result in a false report or false accusation being believed, while lie biases can result in the dismissal of a child's report that was genuine. Thus, it is important to caution that relying solely on children's facial expressions may result in some honest reports being perceived as deceptive. Together, the results from Bruer and colleagues and the present study suggest that the detection of a surprised expression may be one additional cue that can aid lie‐detection performance, though relying solely on this cue would be an ineffective strategy.</p> <hd id="AN0183757202-26">Limitations and Future Directions</hd> <p>There are several limitations to consider with the present research. First, while all participants were instructed to detect children's lies based solely on the child's facial expressions, it is possible that participants still relied on other features (e.g., one's gut instinct or the child's appearance) to inform their decision. While we did have participants view 10 different child interviews, increasing the number of trials and number of children (i.e., senders) can help to increase the generalizability of the present results. In line with a limitation acknowledged by Bruer et al. ([<reflink idref="bib3" id="ref69">3</reflink>]), it is important to remember that this procedure only involved children's facial expressions upon hearing the very first interview question. While this is informative in narrowing the scope of when this might be most important, there are many additional factors (e.g., the child's actual report, other facial expressions, response latency) that may be active in the interview that complicate one's detection judgment. In other words, it may be challenging in practice to rely on this surprised expression when there are other potential factors at play. It is also important to note that we selected videos with the strongest surprised expression scores to show participants; therefore, detection accuracy may decrease when assessing a broader range of surprised expressions. There is also a need to test this lie‐detection paradigm with novel stimuli, as the present study used the same stimuli set as Bruer et al. ([<reflink idref="bib3" id="ref70">3</reflink>]).</p> <p>We tested a limited age range, and so it will be interesting for future research to explore if similar results are found with children of various ages and if the detection of children's lies through their facial expressions differs with age. Future research could also explore how accurate adults are at detecting children's facial expressions in face‐to‐face interactions, as this may more closely mimic the conditions under which some adults (e.g., parents) would be detecting deception. The present procedure may more closely mimic the conditions in a forensic context where one has a videotape that they can playback to code for facial expressions. A next step for this research could be to conduct this study with a sample of child forensic interviewers to explore their ability to detect surprised expressions and assess (dis)honesty. This could help to establish evidence‐based training procedures in facial expression detection that could help with the accurate detection of children's true and false reports.</p> <hd id="AN0183757202-27">Conclusion</hd> <p>The present study found that, with training, adults can detect children's lies at above chance levels based solely on their facial expression to the first interview question, though participants held a consistent lie bias across control and training conditions. These results provide an important extension of the work by Bruer et al. ([<reflink idref="bib3" id="ref71">3</reflink>]) by exploring if automated coding of facial expressions can be useful for the human lie‐detection process. Humans did appear capable of using knowledge on the surprise cue to detect deception, though accuracy rates still hovered just above chance. Considering that adults consistently struggle to accurately detect children's lies, the present study provides insight into the possibility and challenges of using facial expressions as a strategy to detect deception.</p> <hd id="AN0183757202-28">Author Contributions</hd> <p> <bold>Alison M. O'Connor:</bold> writing – original draft, methodology, formal analysis, data curation, conceptualization. <bold>Jennifer Gongola:</bold> conceptualization, methodology, writing – review and editing, data curation. <bold>Kaila C. Bruer:</bold> writing – review and editing, methodology. <bold>Thomas D. Lyon:</bold> conceptualization, funding acquisition, methodology, writing – review and editing, supervision, resources. <bold>Angela D. Evans:</bold> funding acquisition, methodology, writing – review and editing, supervision, resources, conceptualization.</p> <hd id="AN0183757202-29">Conflicts of Interest</hd> <p>The authors declare no conflicts of interest.</p> <hd id="AN0183757202-30">Data Availability Statement</hd> <p>The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.</p> <p>GRAPH: Data S1. Supporting Information.</p> <ref id="AN0183757202-31"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref2" type="bt">1</bibl> <bibtext> Funding: This work was supported by Social Sciences and Humanities Research Council of Canada grant was awarded to Angela D. Evans and National Institute of Child Health and Human Development, HD101617 to Thomas D. Lyon.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref19" type="bt">2</bibl> <bibtext> Alison O'Connor is currently at Mount Allison University.</bibtext> </blist> </ref> <ref id="AN0183757202-32"> <title> References </title> <blist> <bibtext> Bala, N., K. Ramakrishnan, R. Lindsay, and K. Lee. 2005. " Judicial Assessment of the Credibility of Child Witnesses." Alberta Law Review 42 : 995 – 1017. https://doi.org/10.29173/alr1270.</bibtext> </blist> <blist> <bibtext> Bartlett, M. S., G. C. Littlewort, M. G. Frank, C. Lainscsek, I. Fasel, and J. R. Movellan. 2006. " Automatic Recognition of Facial Actions in Spontaneous Expressions." Journal of Multimedia 1 : 22 – 35.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref12" type="bt">3</bibl> <bibtext> Bruer, K. C., S. Zanette, X. Ding, T. D. Lyon, and K. Lee. 2020. " Identifying Liars Through Automatic Decoding of Children's Facial Expressions." Child Development 91 : e995 – e1011. https://doi.org/10.1111/cdev.13336.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref21" type="bt">4</bibl> <bibtext> Döllinger, L., P. Laukka, L. B. Högman, et al. 2021. " Training Emotion Recognition Accuracy: Results for Multimodal Expressions and Facial Micro Expressions." Frontiers in Psychology 12 : 708867. https://doi.org/10.3389/fpsyg.2021.708867.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref16" type="bt">5</bibl> <bibtext> Ekman, P., G. Roper, and J. C. Hager. 1980. " Deliberate Facial Movement." Child Development 51 : 886 – 891. https://doi.org/10.2307/1129478.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref4" type="bt">6</bibl> <bibtext> Fansher, A., and R. V. del Carmen. 2016. " "The Child as Witness": Evaluating State Statutes on the Court's Most Vulnerable Population." Children's Legal Rights 36 : 1 – 44.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref17" type="bt">7</bibl> <bibtext> Feldman, R. S., L. Jenkins, and O. Popoola. 1979. " Detection of Deception in Adults and Children via Facial Expressions." Child Development 50 : 350 – 355. https://doi.org/10.2307/1129409.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref1" type="bt">8</bibl> <bibtext> Gongola, J., N. Scurich, and J. A. Quas. 2017. " Detecting Deception in Children: A Meta‐Analysis." Law and Human Behavior 41 : 44 – 54. https://doi.org/10.1037/lhb0000211.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref9" type="bt">9</bibl> <bibtext> Hartwig, M., and C. F. Bond Jr. 2014. " Lie Detection From Multiple Cues: A Meta‐Analysis." Applied Cognitive Psychology 28 : 661 – 676. https://doi.org/10.1002/acp.3052.</bibtext> </blist> <blist> <bibtext> Hauch, V., S. L. Sporer, S. W. Michael, and C. A. Meissner. 2016. " Does Training Improve the Detection of Deception? A Meta‐Analysis." Communication Research 43 : 283 – 343. https://doi.org/10.1177/0093650214534974.</bibtext> </blist> <blist> <bibtext> iMotions. 2018. iMotions Biometric Research Platform 7.1. Copenhagen, Denmark : iMotions A/S.</bibtext> </blist> <blist> <bibtext> Levine, T. R., Y. Daiku, and J. Masip. 2022. " The Number of Senders and Total Judgments Matter More Than Sample Size in Deception‐Detection Experiments." Perspectives on Psychological Science 17 : 191 – 204. https://doi.org/10.1177/1745691621990369.</bibtext> </blist> <blist> <bibtext> Mac Giolla, E., and T. J. Luke. 2021. " Does the Cognitive Approach to Lie Detection Improve the Accuracy of Human Observers? " Applied Cognitive Psychology 35 : 385 – 392. https://doi.org/10.1002/acp.3777.</bibtext> </blist> <blist> <bibtext> Monaro, M., S. Maldera, C. Scarpazza, G. Sartori, and N. Navarin. 2022. " Detecting Deception Through Facial Expressions in a Dataset of Videotaped Interviews: A Comparison Between Human Judges and Machine Learning Models." Computers in Human Behavior 127 : 107063. https://doi.org/10.1016/j.chb.2021.107063.</bibtext> </blist> <blist> <bibtext> Porter, S., M. Juodis, L. M. ten Brinke, R. Klein, and K. Wilson. 2010. " Evaluation of the Effectiveness of a Brief Deception Detection Training Program." Journal of Forensic Psychiatry & Psychology 21 : 66 – 76. https://doi.org/10.1080/14789940903174246.</bibtext> </blist> <blist> <bibtext> Shen, X., G. Fan, C. Niu, and Z. Chen. 2021. " Catching a Liar Through Facial Expression of Fear." Frontiers in Psychology 12 : 675097. https://doi.org/10.3389/fpsyg.2021.675097.</bibtext> </blist> <blist> <bibtext> Stanislaw, H., and N. Todorov. 1999. " Calculation of Signal Detection Theory Measures." Behavior Research Methods, Instruments, & Computers 31 : 137 – 141.</bibtext> </blist> <blist> <bibtext> Street, C. N. H., and D. C. Richardson. 2015. " Lies, Damn Lies, and Expectations: How Base Rates Inform Lie–Truth Judgments." Applied Cognitive Psychology 29 : 149 – 155. https://doi.org/10.1002/acp.3085.</bibtext> </blist> <blist> <bibtext> Vrij, A., P. A. Granhag, S. Mann, and S. Leal. 2011. " Outsmarting the Liars: Toward a Cognitive Lie Detection Approach." Current Directions in Psychological Science 20 : 28 – 32. https://doi.org/10.1177/0963721410391245.</bibtext> </blist> <blist> <bibtext> Vrij, A., R. Fisher, S. Mann, and S. Leal. 2008. " A Cognitive Load Approach to Lie Detection." Journal of Investigative Psychology and Offender Profiling 5 : 39 – 43. https://doi.org/10.1002/jip.82.</bibtext> </blist> <blist> <bibtext> Walczyk, J. J., F. P. Igou, A. P. Dixon, and T. Tcholakian. 2013. " Advancing Lie Detection by Inducing Cognitive Load on Liars: A Review of Relevant Theories and Techniques Guided by Lessons From Polygraph‐Based Approaches." Frontiers in Psychology 4 : 14. https://doi.org/10.3389/fpsyg.2013.00014.</bibtext> </blist> <blist> <bibtext> Walczyk, J. J., K. T. Mahoney, D. Doverspike, and D. A. Griffith‐Ross. 2009. " Cognitive Lie Detection: Response Time and Consistency of Answers as Cues to Deception." Journal of Business and Psychology 24 : 33 – 49. https://doi.org/10.1007/s10869‐009‐9090‐8.</bibtext> </blist> <blist> <bibtext> Zanette, S., X. Gao, M. Brunet, M. S. Bartlett, and K. Lee. 2016. " Automated Decoding of Facial Expressions Reveals Marked Differences in Children When Telling Antisocial Versus Prosocial Lies." Journal of Experimental Child Psychology 150 : 165 – 179. https://doi.org/10.1016/j.jecp.2016.05.007.</bibtext> </blist> </ref> <aug> <p>By Alison M. O'Connor; Jennifer Gongola; Kaila C. Bruer; Thomas D. Lyon and Angela D. Evans</p> <p>Reported by Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib13" firstref="ref5"></nolink> <nolink nlid="nl2" bibid="bib15" firstref="ref6"></nolink> <nolink nlid="nl3" bibid="bib19" firstref="ref7"></nolink> <nolink nlid="nl4" bibid="bib10" firstref="ref8"></nolink> <nolink nlid="nl5" bibid="bib23" firstref="ref11"></nolink> <nolink nlid="nl6" bibid="bib22" firstref="ref13"></nolink> <nolink nlid="nl7" bibid="bib21" firstref="ref14"></nolink> <nolink nlid="nl8" bibid="bib11" firstref="ref20"></nolink> <nolink nlid="nl9" bibid="bib14" firstref="ref22"></nolink> <nolink nlid="nl10" bibid="bib20" firstref="ref24"></nolink> <nolink nlid="nl11" bibid="bib12" firstref="ref28"></nolink> <nolink nlid="nl12" bibid="bib17" firstref="ref33"></nolink> <nolink nlid="nl13" bibid="bib189" firstref="ref36"></nolink> <nolink nlid="nl14" bibid="bib200" firstref="ref41"></nolink> <nolink nlid="nl15" bibid="bib389" firstref="ref45"></nolink> <nolink nlid="nl16" bibid="bib207" firstref="ref49"></nolink> <nolink nlid="nl17" bibid="bib407" firstref="ref56"></nolink> <nolink nlid="nl18" bibid="bib16" firstref="ref66"></nolink> <nolink nlid="nl19" bibid="bib18" firstref="ref68"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1460825
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Training Humans to Detect Children's Lies through Their Facial Expressions
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Alison+M%2E+O'Connor%22">Alison M. O'Connor</searchLink><br /><searchLink fieldCode="AR" term="%22Jennifer+Gongola%22">Jennifer Gongola</searchLink><br /><searchLink fieldCode="AR" term="%22Kaila+C%2E+Bruer%22">Kaila C. Bruer</searchLink><br /><searchLink fieldCode="AR" term="%22Thomas+D%2E+Lyon%22">Thomas D. Lyon</searchLink><br /><searchLink fieldCode="AR" term="%22Angela+D%2E+Evans%22">Angela D. Evans</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Applied+Cognitive+Psychology%22"><i>Applied Cognitive Psychology</i></searchLink>. 2025 39(1).
– 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: 7
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2025
– Name: SourceSuprt
  Label: Sponsoring Agency
  Group: SrcSuprt
  Data: Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) (DHHS/NIH)
– Name: NumberContract
  Label: Contract Number
  Group: NumCntrct
  Data: HD101617
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Deception%22">Deception</searchLink><br /><searchLink fieldCode="DE" term="%22Nonverbal+Communication%22">Nonverbal Communication</searchLink><br /><searchLink fieldCode="DE" term="%22Recognition+%28Psychology%29%22">Recognition (Psychology)</searchLink><br /><searchLink fieldCode="DE" term="%22Children%22">Children</searchLink><br /><searchLink fieldCode="DE" term="%22Interviews%22">Interviews</searchLink><br /><searchLink fieldCode="DE" term="%22Accuracy%22">Accuracy</searchLink><br /><searchLink fieldCode="DE" term="%22Ethics%22">Ethics</searchLink><br /><searchLink fieldCode="DE" term="%22Artificial+Intelligence%22">Artificial Intelligence</searchLink><br /><searchLink fieldCode="DE" term="%22Comparative+Analysis%22">Comparative Analysis</searchLink><br /><searchLink fieldCode="DE" term="%22Training%22">Training</searchLink><br /><searchLink fieldCode="DE" term="%22Video+Technology%22">Video Technology</searchLink><br /><searchLink fieldCode="DE" term="%22Adults%22">Adults</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1002/acp.70024
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0888-4080<br />1099-0720
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: The accurate detection of children's truthful and dishonest reports is essential as children can serve as important providers of information. Research using automated facial coding and machine learning found that children who were asked to lie about an event were more likely to look surprised when hearing the first question during an interview about said event. The present studies explored if humans can be trained to look for surprised expressions to detect children's deception. Participants made lie-detection judgments after seeing children's expressions in very brief clips. In Study 1, we compared performance across a training condition and control condition, and in Study 2 we modified the training. With training, adults could detect children's lies at above-chance levels by viewing their facial expressions. Detection accuracy was further improved with modified training (Study 2), but participants held a consistent lie bias. Challenges with using facial expressions to detect deceit are discussed.
– 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: EJ1460825
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1460825
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1002/acp.70024
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 7
    Subjects:
      – SubjectFull: Deception
        Type: general
      – SubjectFull: Nonverbal Communication
        Type: general
      – SubjectFull: Recognition (Psychology)
        Type: general
      – SubjectFull: Children
        Type: general
      – SubjectFull: Interviews
        Type: general
      – SubjectFull: Accuracy
        Type: general
      – SubjectFull: Ethics
        Type: general
      – SubjectFull: Artificial Intelligence
        Type: general
      – SubjectFull: Comparative Analysis
        Type: general
      – SubjectFull: Training
        Type: general
      – SubjectFull: Video Technology
        Type: general
      – SubjectFull: Adults
        Type: general
    Titles:
      – TitleFull: Training Humans to Detect Children's Lies through Their Facial Expressions
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Alison M. O'Connor
      – PersonEntity:
          Name:
            NameFull: Jennifer Gongola
      – PersonEntity:
          Name:
            NameFull: Kaila C. Bruer
      – PersonEntity:
          Name:
            NameFull: Thomas D. Lyon
      – PersonEntity:
          Name:
            NameFull: Angela D. Evans
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-print
              Value: 0888-4080
            – Type: issn-electronic
              Value: 1099-0720
          Numbering:
            – Type: volume
              Value: 39
            – Type: issue
              Value: 1
          Titles:
            – TitleFull: Applied Cognitive Psychology
              Type: main
ResultId 1