Abstract
Abstract
Where humans have been found to detect lies or deception only at the rate of chance in offline face-to-face communication (F2F), computer-mediated communication (CMC) online can elicit higher rates of trust and sharing of personal information than F2F. How do levels of trust and empathetic personality traits like perspective taking (PT) relate to deception detection in real-time CMC compared to F2F? A between groups correlational design (N = 40) demonstrated that, through a paired deceptive conversation task with confederates, levels of participant trust could predict accurate detection online but not offline. Second, participant PT abilities could not predict accurate detection in either conversation medium. Finally, this study found that conversation medium also had no effect on deception detection. This study finds support for the effects of the Truth Bias and online disinhibition in deception, and further implications in law enforcement are discussed.
Introduction
T
To deceive is to purposely “create a false belief” in another individual. 15 (p518) Although deception detection (further referred to as detection) has been found to rely more on the skills of the deceiver over the detector, deception and CMC is found in everyday life4,6,16–23 and requires further investigation to understand its limits online and offline. For example, due to the disassociation and disinhibition of offline social norms in CMC, an individual may find it harder to perceive or gauge another's reactions online while being deceived. 24 Online users are “physically invisible,” as CMC does not require face-to-face communication (F2F), 15 which may lose or alter multiple cues of deception; visual, voice pitch, language,11–13 and time limits. 25
The presence or absence of these cues can enhance a disassociation with online behavior, secure anonymity, and hinder detection in online conversations,12,15,26 as well as attribute to the Media Richness Theory (MRT). The MRT suggests that individuals are less likely to lie in media-rich contexts, that is, F2F compared to CMC, 27 due to the increased visual cues offline needed for accurate judgment and higher quality communication. 28
Previous methods of research and synchronous communication
Previously, research required participants to detect lies in artificial or genuine videos of participants or criminals lying.29,30 Although videos allow direct observation of deception, videos do not take in to account synchronous real-time conversation spontaneity and dynamic interactions between communications required for deception.23,30 The asynchronous communication in videos may also appear artificial or “too practiced,” a deception indicator.10,31–33 Furthermore, previous research has mainly focused on one side of the interaction, deceiver methods and motivations.13,14 Therefore, further synchronous communication research is required to understand the detector and how to detect deception.14,31
Synchronous instant messaging studies have been carried out in short online text-based conditions, comparing synchronous CMC to F2F deceptive conditions.13–15,28 For example, participants discussed five topics in dyads with a confederate under instruction to lie about two of the five topics.14,25 Results did not show any significant effect of communication medium on detection, highlighting the varied methodological difficulties associated with detection research for results to be generalized to a wider population. 34 The question remains: Does detection differ between online and offline contexts and, if so, how and why?
Trust
The aforementioned nonsignificant detection results14,25 could be due to characteristics of CMC, like higher levels of trust found in text only communication. 35 Trust can be understood as a willingness to share information with another individual, risking vulnerability, but with the conviction that the individual will willingly reciprocate the behavior.36,37
A predisposition to trust other is known as the Truth Bias. 1 It is found higher online than offline and has a cyclical nature whereby individuals reciprocate trust and rates of self-disclosure perceived in others, enhanced by reduced social inhibitions online.3,38,39 However, individuals believe that others are their “honest” selves online, regardless of whether individuals know each other well or not.16,40,41 This belief could result in a Halo Effect with increased trust, likability, and general positive affect resulting in lower accurate detection rates.32,42,43
The Truth Bias exists to reduce the uncertainty in communication by the “uncertainty reduction theory.” 44 Individuals can ignore cues that could indicate deception as a heuristic shortcut for decision-making, to reduce cognitive load while understanding conflicting incoming information.45,46 The truth bias may be enhanced further online since trust is higher in CMC and individuals will be less likely to expend cognitive effort to find contradictory evidence, believing that his/her conversation partner is telling the truth online. 46 For example, 92 percent of text messages received by participants were rated as honest, despite those same participants rating their own honesty in sent text messages at 73 percent, indicating a strong truth bias in messaging. 21
Social awareness
While trust possibly decreases detection, certain social personality traits might increase it. Those with higher social relationships, grasp of social norms, and social awareness have been found to have higher detection abilities.8–10,33,47 A positive correlation was found between social personality types with quick insight, practicality, and adaptability (extrovert, intuitive, thinking and perceiving [ENTP]-typed individuals) and the increased likelihood for accurate detection. 12 Higher levels of trait empathy and social awareness could also help detect deception online 48 with more research required. 49
Research has suggested that PT can help understand both positive and negative social exchanges 50 and can help predict another individual's deceptive behaviors. 51 Social awareness and empathy could be operationalized through the ability of PT, a cognitive reaction and trait variable, which considers the thoughts, context, and feelings of others and where an individual is more aware of another's point of view.52,53
The present study
This research considers the lack of studies on synchronous communication in deception, the traits and abilities of a detector rather than the deceiver, and the relationship of trust and detection, both online and offline.
Methods
Participants
Forty participants, Partner A the detector (21 = male, 19 = female, ages: 18–24 [n = 27], 25–36 [n = 10], 36–49 [n = 3]), were randomly divided into online (N = 20) and offline (N = 20) conditions. Four confederates, Partner B the deceiver (male = 1, female = 3), were used across both conditions to complete 20 conversations in each condition to create a control for the deceiver role. Both participants and confederates were conveniently sampled from a third level Institute of Technology college, and confederates were recruited from the college drama society. The majority of confederates and participants were middle class, Caucasian, and familiar with communicating online.
Measures
The General Trust Scale, a six-item five-point Likert Scale from “Strongly Disagree” to “Strongly Agree,” was used to measure an individual's general trust. 43 Questions acknowledge the Truth Bias and the risk of vulnerability, 54 for example, “Most people are basically honest.” Higher scores indicated higher levels of trust. A Cronbach's alpha in previous samples of 0.87 (Japan), 0.90 (United States), 40 and 0.862 (Spain) were found, 43 and in this study sample, 0.788, suggesting good internal consistency for the scale.
The PT subsection of the Interpersonal Reactivity Index was used to measure a cognitive construct of trait empathy; an emotional response to emotions experienced by others 55 and has been used in multidisciplinary settings such as negotiation,56,57 law, 58 and social studies. 59 A seven-item section with a five-point Likert Scale from “Does not describe me well” to “Describes me very well” includes the following: “I sometimes find it difficult to see things from ‘the other guys’ point of view.” Higher scores indicate a higher PT ability with items 1 and 4 reverse scored. Test–retest reliabilities were found between 0.61 and 0.62 54 and a Cronbach's alpha of 0.795, suggesting good internal consistency for the scale in this sample.
Procedure
This study used a between groups correlational design to measure the relationship of correct detection of false statements during a conversation task (DV) and trust and PT levels (IV) in two conditions of communication (text-based chat online and face-to-face offline).
Participants were initially told that the study was about conversation styles lasting 30 minutes. Participants were asked to fill out Trust and PT scales and then informed that the study was about deception detection (not just conversation styles). Participants were reminded that they could withdraw at any time and were introduced to their conversation partner. Offline, each dyad sat face-to-face and online, dyads were in separate locations communicating through previously setup and linked Skype accounts, an online text chat-based service on laptops, but were instructed to carry out the same procedure.
All participants followed the same Partner A task across conditions to participate in a conversation limited to 15 minutes and to detect whether two possible lies had been communicated by Partner B. Partner B's task was to initiate the conversation and to deceive Partner A on two indicated randomized conversation topics—time of getting up that morning, a preferred holiday destination, the last time they went running, the last a film was seen, and a favorite piece of technology. Confederates were allowed to practice different topics with each other before testing to eliminate differences in practice effect between the first and last conversations.
Results
Descriptive statistics
Online, 22 false statements were correctly detected (male = 12, female = 10) and offline, 19 were detected (male = 11, female = 8). Total average trust M = 23.95 (SD = 4.78, males; M = 24.62, females; M = 23.21) and total average PT M = 23.85 (SD = 3.36, males; M = 23.33, females; M = 24.42) were found.
Hypothesis 1: online
A standard multiple regression analysis was carried out to test if levels of trust and PT in participants could predict accurate deception detection in an online conversation task. Predictors met all criteria for low multicollinearity present (tolerance = 0.885 and 0.885) for trust and PT online, respectively. The results of the regression indicated the two predictors explained 5.6 percent of the variance [R2 = 0.056, F(2, 17) = 0.509, p = 0.61]. Participants' predicted deception detection accuracy was equal to −0.251 (Trust) and 0.059 (PT). This indicates that trust made the largest contribution of 25.1 percent and significantly inversely predicted detection, that is, lower trust levels accelerated higher detection rates, but PT made a 5.9 percent contribution and did not significantly predict the model online.
Hypothesis 2: offline
A standard multiple regression analysis was carried out to test if levels of trust and PT in participants could predict accurate deception detection in an offline conversation task. Predictors met all criteria for low multicollinearity present (tolerance = 0.886 and 0.886) for trust and PT offline, respectively. The results of the regression indicated the two predictors explained 19.5 percent of the variance [R2 = 0.195, F(2, 17) = 2.064, p = 0.158]. Participants' predicted deception detection accuracy was equal to −0.394 (Trust) and −0.108 (PT). This indicates that trust made the largest contribution of 39.4 percent, but did not significantly predict detection, and PT made a 10.8 percent contribution and did not significantly predict the model offline either.
Hypothesis 3: communication medium
An independent samples t-test was conducted to compare the deception detection scores for participants between online and offline conversation tasks. There was no significant difference in scores for communication online (M = 1.1, SD = 0.447) and offline [M = 0.95, SD = 0.759; t(30.7) = 0.761, p = 0.452]. The magnitude of the differences in the means was very small (η2 = 0.015), suggesting communication medium does not alter detection rates.
Discussion
This study aimed to investigate whether deception detection was affected by communication medium and whether there was a relationship in participant levels of trust and the personality trait of PT with accurate detection. The dominant nonsignificant results in this study suggest that communication is experienced equally online and offline by communicators.
Trust
One aim of the study was to investigate the link between trust and detection online and offline. Trust significantly inversely predicted detection online, that is, higher trust was related to lower detection rates, but not offline. This reflected higher trust levels found online previously, 38 and therefore, perhaps the MRT and the Halo Effect lent to an increased Truth Bias in the offline condition. Participants may have believed that the confederates were unlikely to lie F2F as per the MRT. 28 Second, preference and likability biases like attraction, by having something in common (e.g., similar age, attending the same college) or by having previous experience with the confederate,30,60 could have increased trust in participants and decreased detection F2F.24,30 Higher awareness of the Truth Bias and the Halo Effect (believing you have something in common with a conversation partner whether you know them or not) should be encouraged in CMC users and investigated further.
Perspective taking
A second aim of this study was to understand the relationship of PT and accurate deception detection and it was found that PT did not significantly predict detection in either communication media. This suggests that social awareness is experienced equally in CMC to F2F with participants and confederates familiar with CMC use daily. Future investigations could be broadened to include different personality types and CMC use in detection. For example, detection has been linked to other social personality traits such as ENTP 12 and self-awareness. Self-awareness, in turn, has been linked to the cognitive ability of Theory of Mind, 52 an operationalized concept of PT, which shows an interconnecting relationship between these concepts61–63 and possibly with detection. Future research is warranted.
Communication medium
Finally, participants did not statistically differ in accurate detection between communication media reinforcing the finding that communication is experienced equally online and offline. There were opportunities for confederates to rehearse and to edit any information before sending it online, 13 which instead of appearing rehearsed or artificial may have lent to successful deception. Freedom of editing in CMC, over familiarization with topics and motivation to detect deception, should be investigated further.10,31
Strengths
This research aimed to bridge the gap of the role personality traits play in detection and CMC while taking into account rates of trust in communication. The results reflected detection research with a near 50 percent detection rate; 47 percent offline and 55 percent online in this study,8–10,64 with a method that aimed to create a naturalistic conversation task in a controlled setting where the researcher could match correct detections instead of self-reports of detection. 65 The method also benefited by comparing truthful and nontruthful statements,23,32 in a design including a procedure for confederates who acted as a control across conditions for detection.
Limitations
A previous meta-analysis found that successful detection relied on the deceptive abilities of the liar rather than the abilities of a detector. 66 For example, using drama society confederates may have affected deceiver confidence, nonverbal behavior, and language used during the study.11–13 However, it should be highlighted that most deception research has been focused on the viewpoint of the deceiver only.13,14 This study found low levels of variance for the predicted models, suggesting that deception is neither due to the sole abilities of the deceiver nor the detector and there may be more room for investigating the dynamic interaction between deceivers and detectors.13,67–69 Finally, the sample was from a small third level institution where it could not be ensured that participants and confederates did not know each other previously. Future control for these variables could discern what interactions and deception abilities play more significant roles in detection.
Future recommendations
Future research could focus on a previous link that was found between certain professions and increased detection abilities, for example, law enforcement. 9 This could be explained by increased exposure, experience, and training in deceptive environments.9,12 In addition, certain personalities have been found to be attracted to certain professions, for example, law enforcement.70,71 Could individuals in deceptive work environments, for example, law enforcement or teaching, have increased detection abilities online and offline, along with higher empathy or lower trust levels? The implications consist of finding individuals who have natural higher detection abilities for these professions instead of focusing on training, which has found only moderate increased success of 54 percent.9,72
Conclusion
A positive note about detection research indicates that although online disinhibition elicits higher trust levels that could attribute to deviant behaviors, disinhibition could also alternatively reduce social barriers to communication, allowing personal matters to be discussed more freely online than offline.24,27,30 However, based on this study, individuals are poor at detecting deception online and offline with individual differences and variances remaining in detection; more testing is required.29,49 Some lies are also socially accepted, for example, altruistic lies 33 with personal self-disclosure being accepted as a social norm online.16,40,41 Higher awareness should be made of the self-maintaining and reciprocal nature of trust online as communication can become increasingly more and more personal based on the perceived intimacy of the shared information. 73
Footnotes
Acknowledgment
With many thanks to Nicola Fox Hamilton, supervisor of this article.
Author Disclosure Statement
No competing financial interests exist.
