• No results found

Recruiters', students', and pupils' ability in lie detection

N/A
N/A
Protected

Academic year: 2021

Share "Recruiters', students', and pupils' ability in lie detection"

Copied!
30
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Recruiters’, Students’, and Pupils’ Ability in Lie Detection

Beyza Ozkul

10573089

MSc Business Economics – specialization Managerial Economics and Strategy

15 ECTS

ABSTRACT.

In this thesis the lie detection ability of recruiters and laypersons are measured experimentally. During the experiment participants are randomly shown 5 (out of 40) fragments from the Dutch television game show ‘Wie van de 3’. Afterwards they are asked to correctly identify two liars and one truth teller, and choose on which cues they based their answers. No significant differences are found between the groups. Besides profession two other strategies (experience and age) are used to measure if professionals are better lie detectors than students and pupils. Both show no significant correlation.

(2)

2

STATEMENT OF ORIGINALITY

This document is written by Beyza Ozkul who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 INDEX

1.INTRODUCTION ...4

2.LITERATURE REVIEW ...7

3.EXPERIMENTAL DESIGN AND HYPOTHESES ...12

3.1 Experimental design ...12 3.2 Hypotheses ...14 3.3Additional research ...15 4.RESULTS ...16 4.1 Experience ...16 4.2 Detection cues ...20 4.3Additional research ...21 5.DISCUSSION ...23

5.1 Recruiters’ accuracy rates and difficulty ...23

5.2 Students’ accuracy rates ...24

5.3 Cues ...25 5.4 Comparison ...25 5.5 Further research ...26 6.CONCLUSION ...27 REFERENCES ...28

(4)

4

1. INTRODUCTION

Lying is a part of everyday life. A diary study by DePaulo et al. (1996) find that people lie on average one to two times a day to, e.g., avoid embarrassment or make positive impressions. Deception becomes more frequent when individuals have an incentive to present themselves in a particular manner. Individuals portraying themselves as likeable, competent, or high in potential tell on average 1.75 lies during a 10-minutes period (Feldman et al., 2002).

Impression management is used by individuals to try to portray him- or herself in the best way. These techniques are used to increase the chance that individuals achieve their objectives. For example, an applicant attempts to influence the recruiter’s perception about him or her being high in potential to gain employment. Research shows that these impression management strategies are often used during job applications (e.g., Ellis et al., 2002; Levashina & Campion, 2007).

Job applicants are incentivized to lie during the job application process. The job interview has the highest weight in making hiring decisions, therefore verbal lies are expected (Gilmore, 1989; Kinicki et al., 1990). Weiss and Feldman (2006) find that candidates are indeed verbally deceptive in the job application process. The lies told were besides verbally in job interviews also in writing, e.g., in resumes and cover letters. The finding of deceptive behavior in employment interviews is not surprising since it involves a high stakes process for both the interviewer and job applicant (Weiss & Feldman, 2006). For the interviewer, efficient hiring decisions potentially yield increased productivity which may result in increases in revenue and client base. Furthermore, the work climate could be positively influenced which also could be beneficial for the productivity. For the job applicant, a successful job interview will yield income, security, and stability. Besides these physiological and safety needs, self-fulfillment needs will possibly be satisfied. The self-esteem will be improved after being hired (McGregor, 1957).

It can be concluded that the ability for recruiters to correctly detect deceit during job application procedures is very valuable, since it deals with a high stakes process.

(5)

5

Hiring an unsuited applicant will cost the firm more time and money during the training and settling in period, and the potential benefits are lower. Laying off unsuited employees is also costly. Costs of turnover amount to one-fifth of an employee’s salary to replace that employee, and increase with the importance of higher levels of education and specialized training. Typical costs include costs of off-boarding (severance, benefits continuation), costs of hiring new person (advertising, interviewing), costs of onboarding new employee (training, management time), and lost productivity (a new person may take up to 1 year to reach the productivity of an existing employee, even if the existing employee was unsuited) (Hinkin & Tracey, 2000). These costs can be prevented by decreasing the turnover rate and hiring qualified workforce initially. Detecting deception correctly by recruiters can contribute to reach this result.

Lying from the applicants’ perspective has drawn a lot more research attention compared to the detection of lies by recruiters (Schmid Mast et al., 2011).1 Therefore,

this thesis will focus on the lie detection ability of recruiters and people with a hiring authority. In other words: are professionals better in detecting lies than laypersons? The ability of detecting deceit of professionals and laypersons is measured many times. However, the vast majority of these papers deal with professionals in the field of law enforcement. From these papers is concluded that there exist small variations in lie detection accuracy rates, and that people tend to be more accurate in judging audible than visible lies. (Bond & DePaulo, 2006).

The experimental design and main findings are now described in some detail. To test whether professionals are better in detecting lies than students and pupils, an online experiment is done. The three groups were asked to take voluntarily part in the experiment to measure their deception detection ability. In this experiment short videos were shown from the Dutch television game show ‘Wie van de 3’. In this game show a panel of four Dutch celebrities have the objective to identify one person telling the truth

1 Detection of faking by recruiters has received little research attention, e.g. Lievens & Peeters, (2008)

(6)

6

out of three candidates. All three candidates are pretending to take on the same profession. The four panelists have each one minute to ask the candidates some questions regarding their job. Participants in this experiment are shown an intro in which the profession is described shortly and one round of questioning by one of the panelists. After each fragment the participants are asked to choose two liars and one truth teller, and on which cues they based their answers. After answering the questions, a picture of the candidate who was telling the truth is made visible.

Three strategies in line with Aamodt & Custer (2006) are used to find out if professionals are better lie detectors than laypersons. First, the lie detection accuracy rate of the three groups are compared to each other to find out if occupation influences the lie detection ability. Second, a correlation is calculated between years of job experience of professionals and their accuracy rates. Third, again a correlation is computed, but now to measure the relation between age and accuracy. With results of the first strategy can also be concluded if education influences the lie detection ability. The participants scored a mean accuracy rate of 30 percent and therefore did not score significantly worse than the chance level of 33 percent. Professionals, students, and pupils did not score differently from each other (accuracy rates of 31.90, 27.91, and 30.24 percent respectively). No significant correlation between years of experience and the accuracy rate of professionals is found. The same holds for the correlation between age and accuracy. Overall, professionals do not differ from laypersons in their lie detection ability. Furthermore, education does not influence the ability to detect deceit. It can be concluded that recruiters and people with a hiring authority are not better in detecting deception than students and high school pupils.

The remainder of this thesis is structured as follows. In section 2, a review of the literature is provided. The experimental design and hypotheses are described in section 3. Experimental results are presented in section 4, and discussed in section 5. A conclusion is given in section 6.

(7)

7

2. LITERATURE REVIEW

Research shows that people are not highly skilled at using communication cues to detect lies and are on average only slightly better than chance in detecting deception (Vrij, 2000). Moreover, Bond and DePaulo (2006) conclude that participants are more accurate in judging audible than visible lies. They also reveal that individuals view their interaction partners as honest.

Truth bias combined with visual bias, demeanor bias, and expectancy violation bias account for making mistakes in detecting deception. Truth bias concerns people’s tendency to overestimate the truthfulness of their conversation partner. Visual bias leads to leaving out vocal and linguistic information while the main focus is on the visual aspect. Demeanor bias results in judging some people more honest or deceptive due to their nonverbal expressive style. Finally, on account of expectancy violation bias unusual behavior during conversations is seen as deception (Burgoon et al., 2008). Due to bounded rationality, people rely on mental shortcuts to help process information in potentially deceptive situations. Therefore, judgements made are biased and cause an inaccurate separation of liars and truth tellers. Deceivers’ messages were judged as honest and complete; their behavior as more involved and dominant; and their overall demeanor as more credible (Burgoon et al., 2008).

Another possible reason for disappointing deception detection ability lies in people’s inappropriate beliefs regarding ‘lying behavior’ (Akehurst et al., 1996; Vrij, 1993; Vrij & Semin, 1996). Stereotypes about liars cause inconsistencies between beliefs of participants regarding deception and actual indicators of deception. These inconsistencies could be found in both verbal behavior and non-verbal behavior. For example, pauses, stuttering, and eye blinks were believed to increase in frequency during deception while they have actually been shown to decrease. The opposite was believed for a monotonous voice, making eye contact, and turning to the interviewer. These occurred more frequently during deception than believed (Akehurst et al., 1996).

DePaulo et al. (2003) analyzed the results of 158 cues of deception. Deception cues indicate possible deceit without revealing the nature of the information that is held

(8)

8

back. These cues are signs of mental effort and occur because it is more difficult to lie than to tell the truth. Telling a lie requires constructing a never-experienced story whereas telling the truth is simply telling the way it is (Buller & Burgoon, 1996). This requirement of mental effort may firstly result in less compelling stories which involve, for example, more word repetitions, and less gestures. However, increases in pauses and speech disturbances, such as “uh” and “uhm”, are not indicators of lying behavior. Second, liars are less forthcoming which is mainly visible in the smaller number of details provided, and the shorter talking time. Finally, liars are less positive, more tense, and their lies include fewer ordinary imperfections (DePaulo et al., 2003).

A meta-analysis by Aamodt and Custer (2006) shows that some groups of professionals are more accurate in detecting deception than students. The accuracy rate for students is 54.22% with local and federal law enforcement agencies having similar rates in detecting lies, while a rate of 50 percent was expected at chance. Criminals, secret service agents, psychologists, social workers, teachers, and judges do better with rates around 60 and 70 percent. Detectives and parole officers seem to be the worst in detecting deception (40 percent).

The focus of previous research regarding detecting deception was greatly on law enforcement agencies, while employees active in the field of human research are of even importance. Both occupational groups are in contact with other people on a daily basis and have the main objective to filter good from bad. Further, in both cases is dealt with high stakes for the interviewees, i.e., imprisonment and employment, and is making mistakes by the interviewers costly. Before starting a career as recruiter, a training must have been undergone in which recruiters are discouraged to trust their instincts to prevent biased decisions (Woehr & Huffcutt, 1994). During their career recruiters accumulate lots of experience, but often without any feedback. It can be concluded that having a high lie detection ability is important for this group. However, recruiters have received only little research attention.

Two articles in this field are now described in some detail. Wood, Schmidtke and Decker (2007) examine the impact of human resource (HR) management experience on

(9)

9

the extent to which misrepresentations on job applications are perceived as lies and the hiring intentions. This is done by comparing the results of students and HR professionals from a questionnaire. The questionnaire contained a short scenario containing a fictitious company, a job description, and 20 items set up as omission misrepresentations. It appears that HR professionals are less likely to perceive misrepresentations as lies and more likely to hire job applicants who included misrepresentations. Both results are significant but in the opposite direction of what was hypothesized. The paper argues two possible reasons for the unexpected outcomes. First, recent scandals are used as classroom examples in business schools and may have a positive effect on students towards unethical behavior. Second, HR professionals may become insensitive to lying on applications or resumes and are more willing to tolerate this behavior.

Schmid Mast, Bangerter, Bulliard and Aerni (2011) investigate whether recruiters are better assessors of personality, better in detecting deception, and a potential relation between these two skills. To assess the big five personality traits, recruiters and students watched six filmed target applicants presenting themselves for a mock job interview. Each trait was judged afterwards on a six-point scale, where a larger value indicated a stronger presence of the given trait. To measure lie detection ability, participants watched four additional target applicants presenting themselves once correctly and once fictively. Afterwards, participants were asked to indicate which version was truthful and which one was a lie. Results show that recruiters were better in assessing the personality trait openness, students were better in assessing extraversion and conscientiousness, and both were not able to assess the traits neuroticism and agreeableness. Additionally, Schmid Mast et al. (2011) indicate that recruiters are better in detecting lies than chance with an accuracy rate of 64 percent and are also significantly better in detecting deception than students (mean of 47% and a chance rate of 50 percent). Eventually, correctly assessing conscientiousness or neuroticism, and recruiter job experience are significantly positively correlated to deception detection.

(10)

10

Other studies investigate the same question in different fields: Are professionals better in detecting deception than laypersons? Vrij and Mann (2001) investigate this in a real-life high-stake context. 51 Dutch police officers are shown 8 fragments of press conferences in which people are asking for help to find their relatives or the murderers of their relatives. Five fragments showed people who lied and were eventually themselves found guilty of killing their own relatives, and the other three fragments were fillers. The mean accuracy rate found is equal to 50 percent. Ekman and O’Sullivan (1991) and Ekman, O’Sullivan and Frank (1999) research deception detection ability of different law enforcement and psychologist groups with use of 10 one-minute videotapes. U.S. Secret Service agents are the most accurate within the first experiment and psychologist with a special interest in deception have the most accurate performance in the second experiment. The paper concludes that accurate judgments are not bound to selected law enforcement groups and may be related to special interests in deception detection.

Besides profession, the correlation between other personal characteristics and of accuracy rate of detecting deception are examined. Aamodt and Custer (2006) conclude that individual differences as age, gender, education, law enforcement experience, neuroticism, extraversion, and confidence are not related to lie detection ability. However, little research is done on the relationship between education and deception detection ability. The conclusion made is, therefore, arguable and more research is desirable.

The great part of previous papers concerning lie detection have the focus on law enforcement professionals like police officers and judges. Lie detection ability for this group is measured many times and on that account in many ways, i.e., visually, audibly, and written. The experiment done by Vrij and Mann (2001) involved a high-stakes setting. This was possible, because the ground truth was available. The persons visible on the tapes showed real cues of deception and the lie detection ability of the participants of the experiment was really measured. This is in contrast to the experiments by Ekman and O’Sullivan (1991) and Ekman et al. (1999), where the

(11)

11

observers were told that the subjects on tape were highly motivated, but stakes were low. The same holds for the experiment by Schmid Mast et al. (2011). Subjects participated in an experiment, in which stakes were relatively low. Further, all participants are shown the same four fragments. This may have caused biased results, because the internal validity and reliability are both questionable. With a few fragments it is unsure if the right accuracy rate was hit (validity), and showing the same fragments to all participants could have resulted in consistently measuring the wrong accuracy rate (reliability). Wood et al. (2007) measured the written lie detection ability of HR professionals, which is useful for written job applications. As mentioned before, the job interview has the highest weight in the hiring process (Gilmore, 1989; Kinicki et al., 1990). Written lie detection ability is useful, but not the most important competence. Another downfall of this experiment is that people who claimed some HR experience are immediately involved in the experiment without any verification. The actual HR experienced workers may be under-represented in this biased sample, which may have caused some inconsistent results too.

Compared to previous research, this thesis makes a number of contributions. First of all, the focus is on the area of hirers instead of the largely investigated area of law enforcement workers. Besides, both recruiters and employers with hiring authority are taken into account. Since both are involved in hiring decisions, it is useful to investigate the deception detection ability of both. Second, a large visual experiment is done in which 5 out of 40 fragments are randomly shown to the participants instead of the same fragments to all of the participants. Third, the interviewees have relatively high stakes, since they are both monetarily incentivized as well as monetarily. They are non-monetarily incentivized because of their television appearance and competitiveness, which makes it more likely that they participate actively. Last, besides gender, age, and experience, differences in education are examined. Most recruiters and managers have a high education, which makes it hard to separate through this characteristic. A third group of high school pupils is added to distinguish.

(12)

12

3.EXPERIMENTAL DESIGN AND HYPOTHESES

3.1. Experimental design.

Television show. ‘Wie van de 3’ is a Dutch television game show in which a panel of

four celebrities has been commissioned to correctly identify a candidate based on a description of a profession and answers to some questions. The described candidate is accompanied by two impostors who pretend to take on the same profession. The TV show starts with the three contestants presenting themselves as the ‘central character’, followed by a description of the profession narrated by the presenter. The celebrity panelists have each one minute to question the three candidates. The ‘central character’ is not allowed to lie, but the two impostors are. Candidates have some unknown time to prepare their answers and expand their knowledge about the profession. Contestants have relatively high stakes to act desirably. The prize money of 1000 euros is divided equally among the three candidates if all panelists chose an impostor, each correct vote diminishes the prize money with 250 euros.

Fragments. Subjects were shown edited fragments which included the introduction of

the contestants, the description of the profession by the narrator, and one minute of questioning by one of the panelists. Five out of forty fragments were randomly presented. Each fragment displayed a different occupation. More fragments with the same occupation existed, but these included a questioning period of another panelist. The occupations of dentist, drive-in greengrocer, maternity nurse, detective, dog groomer, archaeologist, veterinarian, carpenter, theme park manager, chimney sweeper, tour leader, and butcher had three different fragments. The occupations of marathon speed skater and broker only had two different fragments.

Experiment. An online experiment with five video fragments was created to measure

the deception detection ability of the three samples. The experiment started with a short questionnaire containing questions about gender, age, and the specification in

(13)

13

which sample the participant belongs. Afterwards, the test started with a description of the task. Participants were shown five fragments from the TV show ‘Wie van de 3’ and are asked to determine one truth teller and two liars. Subjects are also asked on which cues their choice was based. Possible answers included nonverbal behavior, given answers, and physical appearance of the game contestant, or betting and gut feeling of the subject. After each fragment, the right answer is shown. At the end, participants are offered the chance to leave their email address to receive their score relative to others.

A few matters are programmed to ensure the experiment was taken correctly by the participants. First, the fragments started immediately when the participant entered the screen. This was done to make sure the participants would watch the fragment. For the same reason, a confirmation had to be given to leave and go on to the questions. Second, pictures of the candidates were displayed at the questions to prevent mistakes. Third, the only possible combination of answers to give was selecting two liars and one truth teller, otherwise continuing the experiment was not possible.

Participants. Participation in the online experiment happened privately on an

electronic device connected to the internet, e.g. laptop, tablet, or smartphone. Instructions were displayed on the screen of the electronic device. Participants were told that their answers would remain anonymous to the experimenter and others unless they filled in their email address at the last step to receive more information about their own score relative to other participants.

The online experiment was conducted in Dutch and given to three separate samples. The first sample consisted of recruiters and people with hiring authority, e.g. managers and employers. People studying at university and high school students formed the second and third sample respectively. Participation in the experiment happened voluntarily and none of the samples was monetarily incentivized. The online experiment was mailed to 179 professionals (sample one). 42 returned complete information for a 23.5% response rate. 69 percent of the subjects reported the profession

(14)

14

of recruiter, others claimed a hiring authority. A total of 43 completed experiments were collected from students (sample two), resulting in a response rate of 86 percent. This sample included undergraduates and graduates. The experiment was also mailed to 45 high school students (sample 3). A total of 41 pupils responded (91.1 percent). These students attend the fourth or fifth grade of the Dutch secondary education.

3.2. Hypotheses. Meta-analyses by Aamodt and Custer (2006) and Bond and DePaulo (2006) find accuracy rates around fifty percent (equal to the level of chance) and no evidence that professionals are superior to laypersons in distinguishing lies from truths. In line with these papers is the article by Wood et al. (2007). They find that HR professionals are worse in detecting written deception compared to students. However, Schmid Mast et al. (2011) find that recruiters are significantly better in detecting visual lies than students. This thesis considers audible and visible deceptive behavior. Therefore, the following hypothesis is stated.

H1: Recruiters and other individuals with a hiring authority are better skilled in

detecting deception than students and high school pupils. The accuracy rates will be around chance level.

In previous research subjects had to tell whether a person was lying or not. The chance level was equal to fifty percent and accuracy rates were between forty and seventy percent (Aamodt & Custer, 2006; Bond & DePaulo, 2006). In this thesis the setup is slightly different. Subjects have the objective to correctly identify the truth teller. This automatically implies finding two liars. The chance is equal to 1 to find one liar, and one third two find both liars. Therefore, accuracy rates in this experiment are expected to be around 33 percent.

Aamodt and Custer (2006) made use of three strategies to conclude if experienced lie catchers are more accurate than students. First, to look at correlations between lie detection accuracy and years of experience of professionals. Second, to look at

(15)

15

correlations between age and deception detection accuracy, assuming that with age more opportunities occur to detect and encounter lies. The third strategy is to compare accuracy rates of student and professionals, and therefore similar to hypothesis 1. The other two strategies result in the following hypotheses.

H2: The number of years of experience within the field of hiring employees is positively correlated to accuracy in detecting lies.

H3: The age of participants is positively correlated to the accuracy in detecting

deception.

It is expected that recruiters will perform better than laypersons. Aamodt and Custer (2006) conclude that education and cognitive ability do not appear to be related to deception detection ability. To exclude that differences in education is a possible reason for results found, a third group of pupils is added.

H4: High school students are equally abled in detecting lies as the other group of laypersons (students), but worse in detecting deceit than recruiters.

This thesis predicts significant differences in lie detection accuracy rates between recruiters and the two groups of laypersons. These rates are expected to be around the level of chance (33 percent). Moreover, deception detection ability is expected to be positively correlated with individual differences, i.e., years of experience and age, but not correlated to education.

3.3. Additional research. A meta-analysis by DePaulo et al. (2003) concluded that liars tell less compelling stories, are less forthcoming, less positive, more tense and their lies include fewer ordinary imperfections. For each conclusion lots of cues were analyzed. In this thesis only the cues that were significantly more or less present during deception were analyzed. Besides, the cues must be clearly visible or intelligible in the fragments. Firstly, to make sure the participants were able to observe these cues. Secondly, to

(16)

16

ascertain the cues are rightly coded. For example, plausibility, pupil dilation, and vocal tension are hard to code without any specific knowledge or special attributes.

Coding is done by one person, and counted the occurrence of a cue of the truth teller and the mean occurrence of the two liars in one fragment. The cues time per answer, number of details per answer, number of lip presses per 10 seconds, number of chin raises per 10 seconds, number of repetitions per 10 seconds, and number of gestures per 10 seconds are coded. To control the number of ah-speech disturbances per 10 seconds and the number of shrugs per 10 seconds are also counted. Longer talking time, details, and gestures are expected less during deception. Lip pressing, chin raising2, and

repetition of words and phrases are expected more during deception. Finally, ah-speech disturbances and shrugs are not expected to be significantly different.

4.RESULTS

4.1. Experience. Summary statistics are reported in Table 1. The accuracy rate is calculated as the percentage of correct answered fragments (out of 5). The mean accuracy rate for detecting deception is 30 percent (28.89% for females and 31.11% for males). The participants are almost equally divided across the three subsamples and half of them was female. The largest part of the professionals (sample 1) had only a few years of experience.

The mean accuracy score for detecting deception was 30 percent, while a percentage of 33.3 could be expected on the basis of chance alone. The accuracy rates obtained from the experiment are ordinal and not normally distributed, but the simple random sample consists of a reasonably large number of observations. Therefore, an independent t-test could be done. From this can be concluded that subjects did not score significantly lower than chance, t (125) = -1.7969, p = 0.075. This differs, however,

2 Zivin (1982) found that liars more often than truth tellers raised their chins as part of a particular

facial constellation during conflict situations. This plus face consists of a raised chin, direct eye contact, and medially raised brows.

(17)

17

when the t-tests are computed separately for each group. Undergraduates and graduates (sample 2) performed significantly worse than chance, t (41) = -2.0294, p = 0.049.

Table 1: Summary statistics

Mean Std. Error Min. Max. Test Accuracy rate Background characteristics 30.00 20.82 0 100 Age 24.33 9.72 15 57 Female 0.50 Sample

Recruiters / Hiring authority 0.33

Students 0.34 Pupils 0.33 Years of experience 0 – 2 0.43 2 – 5 0.17 5 – 10 0.19 10 or more 0.21 N 126

The hypotheses stated a better lie detection ability for recruiters than laypersons. Recruiters and people with a hiring authority indeed scored higher in this experiment (Table 2). However, no conclusions can be drawn since both the fractions of females and ages are significantly different in each occupational group.3 To control for these

group characteristics a regression is performed. An ordered logit regression is used, because the dependent variable (accuracy rate) is ordinal and between a range of 0-100

3Fisher-Freeman-Holt exact tests for the fraction of gender and age resulted in p-values of 0.039 and 0.000 respectively. A Fisher-Freeman-Holt exact test is used, because frequencies for both age and gender are in 25 percent of the cases lower than 5, and both variables are not continuous.

(18)

18

percent. Furthermore, the underlying sample is not normally distributed which makes a logistic model more suitable than a probit model.

Table 3 shows the results. The coefficient for recruiter has a positive sign, but is not significant z = 1.21, p = 0.226. Participants who practice the profession of recruiter or have a hiring authority do not perform better in this experiment. In addition, the coefficient for high school pupils does not reach significance z = 0.03, p = 0.980. Consequently, education does not influence the lie detection ability. Recruiters and high school pupils did not differ in their lie detection ability compared to students.

Aamodt and Custer (2006) provide two other strategies to examine if experienced lie detectors are more accurate than laypersons. First, a correlation is calculated between the accuracy rate and years of job experience of the participants with a hiring authority.4 The variables are discrete and ordinal respectively, which makes a

Spearman’s rank correlation the most accurate. The correlation found did not significantly differ from zero, 𝜌" = -0.022, p = 0.888. Participants with more years of

job experience do not significantly perform better than participants with less job experience. Second, a correlation was computed across all three groups between age and deception detection accuracy. This is done, because more experience in detecting deception is expected when participants become older. The variables are again not

4Years of experience is not included in the ordered logistic regression because of perfect multicollinearity. With the Spearman’s rho calculated, the correlation between accuracy rate and years of experience within the group of recruiters is examined.

Table 2: Sample size, gender, age, and accuracy rates per group

Age Accuracy rate

N Female Mean Std. Error Mean Std. Error Sample

Recruiters / Hiring authority 42 0.64 34.88 9.83 31.90 23.39

Students 43 0.49 21.74 2.04 27.91 17.53

(19)

19

continuous, which makes the Spearman’s rank correlation the most accurate. However, no significant Spearman’s rho was found, 𝜌" = -0.029, p = 0.750. Correlations between

age and deception detection accuracy were also computed separately in each subsample. Again, none of these correlations approached significance, (𝜌#$ = -0.065,

p = 0.681; 𝜌#% = -0.134, p = 0.393; 𝜌#& = -0.004, p = 0.979). The same conclusion can be

drawn from the ordered logit regression (Table 3). The overall coefficient for age is negative (𝛽()*= -0.034) but not significant, 𝑧 = -1.13, p = 0.258.

Table 3: Logit regression for accuracy rate

Age -0.034 (0.030) Female -0.223 (0.337) Recruiter 0.669 (0.553) Pupil 0.011 (0.434) Threshold parameter 𝜇$ -2.342 𝜇% -0.644 𝜇& 1.245 𝜇- 2.314 𝜇. 4.145 pseudo 𝑅% N 0.0056 126

Note. The standard errors are given between brackets. ∗p < .1; ∗∗p < .05; ∗∗∗p < .01

To examine the differences in lie detection ability between males and females a Fisher-Freeman-Halton exact test is done. This test is used instead of a regular chi-squared because of the low frequency (5 or less) in half of the categories. The test across

(20)

20

all groups resulted in a p-value of 0.691. Consequently, there is no significant difference in gender for lie detection accuracy rates. This was also the case when tests were done separately in each subsample (𝑝$ = 0.383; 𝑝% = 0.885; 𝑝& = 0.644). Hence, males and

females do not differ significantly in their ability to detect deceit.

4.2. Detection cues. Percentages of fragments answered correctly ranged from 6 to 59 percent. These percentages are normally distributed resulted from the Shapiro-Wilk test, 𝑊 = 0.9815, 𝑝 = 0.7337. t-tests were computed to conclude if some fragments were

answered significantly more or less correctly. 4 out of 40 fragments were answered significantly worse than the chance level of 33 percent, and 2 out of 40 fragments were answered significantly better.

Different occupational groups based their answers on different cues. To test the significance of these differences, chi-squared tests are computed for each possible reason. Chi-squared tests are used to compare differences in large number of observations (N = 630) between the three occupational groups. Cues involving “given answers”, “body language”, and “physical appearance” reached significance, (𝜒345*6 86#9*:#% = 11.473, p = 0.003; 𝜒

;<=> ?86)@8)*% = 13.872, p = 0.001; 𝜒AB>#4C8? 8DD.% =

9.808, p = 0.007). In figures 2, 3, and 4 the distributions of these cues are shown for each individual group. Together with the chi-squared values, it can be concluded that different subsamples do not base their answers on the same cues.

From the figures can be seen that students detect lies relatively more with help of physical and verbal cues. However, both do not accurately hint to deception. Attractiveness is not significantly related to deception, and no evidence is found for stereotyping either (DePaulo et al., 2003). Liars are commonly judged as honest due to truth bias, and given answers do not point out to deception (Burgoon et al., 2008). Some parts of body language could be true cues of deception and could argue why recruiters and people with a hiring authority performed better. Some of these cues are examined in the next sections.

(21)

21

4.3. Additional research. Summary statistics for deception cues are given in Table 4. Means are shown for the truth teller group and the liar group. None of the cues were normally distributed, as resulted from Shapiro-Wilk tests. Due to the non-normality, continuity, and independence of the independent variable Mann-Whitney U tests are used to compare the means conditional on whether the participant is a liar or not.

Time in seconds per answer, details per answer, and gestures per 10 seconds are hypothesized to be lower in case of deception. Means of time and details are not significantly different from each other, 𝑝F4G* = 0.071; 𝑝=*F84?# = 0.636. However, liars are observed to be using more gestures, 𝑝)*#F@:*# = 0.000. Number of lip presses, chin

raises, and repetitions are expected to increase during deception. All three cues are indeed significantly higher in case of deceit, 𝑝?4D D:*## = 0.000; 𝑝CB46 :84#* = 0.000;

𝑝:*D*F4F4<6 = 0.000. The number of ah-speech disturbances and shrugs per 10 seconds

are involved as control and expected to be equal during deception and telling the truth. The number of filled speech disturbances were not significantly different in case of

0% 20% 40% 60% 80% 100% Figure 2: "Given answers" is given as cue Yes No 0% 20% 40% 60% 80% 100% Figure 3: "Body language" is given as cue Yes No 0% 20% 40% 60% 80% 100% Figure 4: "Physical appearance" is given as cue Yes No

(22)

22

veracity and deception, 𝑝#D**CB =4#F@:H86C* = 0.985. In contrast to number of shrugs, which was higher in case of lying, 𝑝#B:@) = 0.026.

Talking time, use of details, and presses of lips are cues to conclude if liars are less forthcoming. Only the number of lip presses reaches significance and is a sign of holding back. However, the other two cues are also in the predicted direction. Therefore, these findings are in line with the previous findings of liars being less forthcoming.

Gestures, chin raises, repetitions, shrugs and ah-speech disturbances are related to telling compelling tales. More gestures imply more engagement of the sender, which is seen more often by liars in the fragments of ‘Wie van de 3’. Chin raises and shrugs are associated with uncertainty and were both more present during deception. Repetitions and ah-speech disturbances make a tale less fluent and therefore less compelling. Repetitions are used more frequently by liars, whereas the number of ah-speech disturbances are equal. It can be concluded that liars are less compelling from these cues.

Table 4: Summary statistics for deception cues

Truth teller Liar

Mean Std. Error Mean Std. Error Cue

Time per answer (in seconds) 4.495 3.032 3.515 1.771 Number of details per answer 0.288 0.417 0.181 0.189 Gestures per 10 seconds 0.466 1.088 1.218 0.991 Chin raise per 10 seconds 0.069 0.249 0.705 0.697 Lip press per 10 seconds 0.517 1.110 1.355 1.520 Repetitions per 10 seconds 0.450 0.885 1.021 0.910 Ah-speech disturbances per 10 seconds 1.980 1.866 1.802 1.255

Shrugs per 10 seconds 0.252 0.648 0.505 0.779

(23)

23 5.DISCUSSION

Results are greatly consistent with the general hypothesis. The general hypothesis expects accuracy rates around chance level and no difference in lie detection ability between professionals and laypersons. Participants scored a mean accuracy rate of 30 percent, but did not score significantly worse than the expected chance level of 33 percent. When comparing the groups separately to the chance level, the group of students did score significantly worse with an accuracy rate of 27.91 percent. Furthermore, the coefficients in the ordered logit regression show that recruiters are indeed not better lie detectors than students and pupils. The accuracy rates are also not correlated to individual differences such as age and education. These results suggest that people are in general equally likely in detecting lies.

Individual scores showed that the lie detection task was difficult. Only one participant answered all five fragments correctly and five obtained an accuracy score of 80 percent, which corresponds to 4 out of 5 fragments correct. 1 out of 6 participants did not find any of the correct truth tellers.In the next subsection some limitations are addressed.

5.1. Recruiters’ accuracy rates and difficulty. Three reasons are suggested for the findings. First, stakes for contestants were relatively high in the TV show ‘Wie van de 3’ compared to other laboratory studies. However, these stakes may have been too low to strongly arouse emotions that could betray a lie. Coding concluded that some cues of deception were present during the fragments. These could have been not detected, because people are not aware of these cues or still hold on to old inappropriate beliefs, for example forms of speech disturbances and gaze aversion. Besides, fragments only showed a candidate lying or telling the truth which possibly made it hard to distinguish veracity from deception. Even more higher stakes could have been found in a setting with field data, in which applicants were applying for a job. A downside of field data is that it is hard to retrieve the ground truth since most methods, such as the polygraph and brain fingerprinting, are not a hundred percent accurate (Aamodt & Custer, 2006).

(24)

24

Second, none of the three groups of participants were monetarily incentivized, which could have caused lower motivation to detect liars. This could have been the case for the students and high school pupils groups, which are asked to do a favor by participating. The group of recruiters are also asked for a favor, but could easily ignore the invitation to participate in this experiment. Recruiters who did participate (response rate of 23.5 percent) are probably more interested in deception detection. Previous studies have shown that people with a special interest in deception are more accurate than others with the same occupation but no interest (Ekman & O’Sullivan, 1991; Ekman et al., 1999). To conclude, recruiters who participated are not expected to perform worse due to missing monetary motivation.

Third, recruiters could have been performing worse due to the unnatural setting in which they had to detect lies. In job interviews recruiters have the chance to ask questions in case of hesitation of the truthfulness of the applicant. Besides, fragments were fixed and maybe did not include questions the recruiters wanted to be answered. Both may have influenced the accuracy rates of the group of recruiters negatively. In contrast, a study by Buller et al. (1991) compared lie detection abilities of interviewers and observers of the interview. The results showed that the observers scored higher accuracy rates (49 percent) than the interviewers (29 percent). These findings suggest that interviewers are in disadvantage when detecting lies. Formulating, asking, and answering questions require cognitive energy which cannot be used for the deception detection task. Therefore, it is not obvious that the possibility to ask questions in the experiment may have resulted in higher accuracy rates. Comparing the accuracy scores of the panelists with the participants of the experiments is not possible in this setting since the panelists have seen 4 rounds of questioning. No accurate conclusion could be drawn from this comparison.

5.2. Students’ accuracy rates. The students group performed as only group worse than chance. This group based their answers significantly more on the cue “given answers” (Figure 2). A study by Strömwal et al. (2007) showed that adults were less accurate in

(25)

25

detecting prepared statements than unprepared statements. Furthermore, Bond and DePaulo (2006) concluded that planned messages appeared more truthful. Since candidates in the game show ‘Wie van de 3’ had time to prepare possible answers, detecting lies could have been more difficult. Students focused significantly more on verbal deception by the candidates compared to other factors. Therefore, they could have been misled by what they heard and missed some other cues to detect the real liars.

5.3. Cues. To see if the three groups based their answers on different cues, chi-squares test are computed. Figures 2, 3, and 4 show that this is the case with three out of five cues (the cue “other” not counted). Students focus relatively more often on given answers by the candidates. Recruiters and people with a hiring authority focus relatively less on physical appearance of the candidates. This could be due to recruiters sticking less to stereotypes and not influencing their choices on appearances alone.

Additional research showed that cues of deception were visible in the fragments. From this could be concluded that liars were less forthcoming and less compelling which is in favor of DePaulo et al. (2003). However, the cues were only counted when the candidates were visible on film and coding is done without any experience or specific knowledge. Off screen behavior is not analyzed and could result in other conclusions. Further, the number of gestures and shrugs were both more present during deception which is not in line with the meta-analysis of DePaulo et al. (2003). More precision and expertness in coding could possibly influence these results.

5.4. Comparison. Accuracy rates in this thesis cannot be immediately compared to accuracy rates in other studies, because the accuracy rate in this context does not only describe lie detection ability, but also the ability to detect veracity. Answering a fragment right corresponds to detecting two liars and one truth teller correctly. A wrong answer could point to identifying a truth teller as a liar and vice versa. In both cases the lie detection ability is not accurate. The first case is due to unjust suspicions,

(26)

26

while the second case this is due to truth bias. It is unknown how the choices are made by the subjects, and therefore not known which is the predominant bias.

5.5. Further research. Further research could focus more on cues used and training. The question about the cues used for answering could be less restricted and have a focus per candidate. This could help to have more insight in how participants answer the questions and the difference between lie detection ability and truth detection ability. Lying in job applications already got some attention, for example personality tests which involve some questions a few times to detect liars. This could be extended in the job interview. Experiments could investigate if detecting deception improves after some training of creating awareness, since some cues of lying behavior are present. Results could have a large impact for the job application process.

(27)

27 6.CONCLUSION

Lie detection accuracy rates of recruiters, people with hiring authority, students, and high school pupils are examined and are close to chance. Although recruiters and people with a hiring authority performed slightly better (31.90 percent) compared to the two groups of laypersons (27.91 and 30.24 percent), these differences were not significant. Other results concern individual characteristics. No correlation between lie detection ability and age or experience has been found. Hence, more experienced participants are not more accurate than novices. Furthermore, this thesis involved two groups of laypersons, pupils and students, to test if educated people are more accurate in detecting lies. Again, no correlation has been found and education does, as predicted, not appear to be related to accuracy in detecting deception.

It is unsure if lie detection ability is really measured in this thesis. First, both lie detection ability as truth detection ability are involved to answer a fragment correctly. Second, it is unsure that cues of deception were detectable and lie detection was truly possible. Other limitations deal with relatively low stakes for the game contestants, no monetary incentives for the participants, and no natural setting was used during the experiment. Participants saw the candidates for a short time and had no opportunity to ask questions. This may have resulted in lower accuracy rates for recruiters and people with a hiring authority. Further research could concentrate on cues participants use to detect deception. Since improvement in lie detection ability will benefit a lot of companies, it will also be interesting to investigate if training or awareness yields the desired outcome in future.

(28)

28 REFERENCES

Aamodt, M. G., & Custer, H. (2006). Who Can Best Catch a Liar? A Meta-Analysis of Individual Differences in Detecting Deception. Forensic Examiner, 15(1), 6-11. Akehurst, L., Köhnken, G., Vrij, A., & Bull, R. (1996). Lay persons’ and police officers’ beliefs regarding deceptive behaviour. Applied Cognitive Psychology, 10(6), 461-471.

Bond, C. F., & DePaulo, B. M. (2006). Accuracy of Deception Judgments. Personality

and Social Psychology Review, 10(3), 214-234.

Buller, D. B., Strzyzewski, K. D., & Hunsaker, F. G. (1991). Interpersonal deception II: The inferiority of conversational participants as deception detectors.

Communication Monographs, 58(1), 25-40.

Burgoon, J. K., Blair, J. P., & Strom, R. E. (2008). Cognitive Biases and Nonverbal Cue Availability in Detecting Deception. Human Communication Research, 34(4), 572-599.

DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, Charlton, K, & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129(1), 74-118.

Ekman, P., & O’Sullivan, M. (1991). Who can catch a liar? American Psychologist,

46(9), 913-920.

Ekman, P., O’Sullivan, M., & Frank, M. G. (1999). A few can catch a liar. Psychological

Science, 10(3), 263-266.

Ellis, A. P. J., West, B. J., Ryan, A.M., & Deshon, R. P. M. (2002). The Use of Impression Management Tactics in Structured Interviews: A Function of Question Type? Journal of Applied Psychology, 87(6), 1200-1208.

Gilmore, D. (1989). Applicant perceptions of simulated behavior description interviews.

Journal of Business and Psychology, 3(3), 279-288.

Hinkin, T. R., & Tracey, J. B. (2000). The cost of turnover: Putting a price on the learning curve. The Cornell Hotel and Restaurant Administration Quarterly,

41(3), 14-21.

Huffcutt, A. I., Van Iddekinge, C. H., & Roth, P. L. (2011). Understanding applicant behavior in employment interviews: A theoretical model of interviewee performance. Human Resource Management Review, 21(4), 353-368.

(29)

29

Kinicki, A. J., Lockwood, C. A., Hom, P. W., & Griffeth, R. W. (1990). Interviewer Predictions of Applicant Qualifications and Interviewer Validity: Aggregate and Individual Analyses. Journal of Applied Psychology, 75(5), 477-486.

Levashina, J., & Campion, M. A. (2007). Measuring Faking in the Employment Interview: Development and Validation of an Interview Faking Behavior Scale. Journal of Applied Psychology, 92(6), 1638-1656.

Lievens, F., & Peeters, H. (2008). Interviewers’ Sensitivity to Impression Management Tactics in Structured Interviews. European Journal of Psychological Assessment, 24(3), 174-180.

McGregor, D. M. (1957). The Human Side of Enterprise. Management Review, 46(11), 22-28.

Schmid Mast, M., Bangerter, A., Builliard, C., & Aerni, G. (2011). How Accurate are Recruiters’ First Impression of Applicants in Employment Interviews? International Journal of Selection and Assessment, 19(2), 198-208.

Strömwall, L. A., Granhag, P. A., & Landström, S. (2007). Children’s prepared and unprepared lies: Can adults see through their strategies? Applies Cognitive Psychology, 21(4), 457-471.

Van Iddekinge, C. H., Raymark, P. H., & Roth, P. L. (2005). Assessing Personality with a Structured Employment Interview: Construct-Related Validity and Susceptibility to Response Inflation. Journal of Applied Psychology, 90(3), 536-552.

Vrij, A. (1993). Credibility judgments of detectives: the impact of nonverbal behavior, social skills, and physical characteristics on impression formation. The Journal

of Social Psychology, 133(5), 601-610.

Vrij, A. (2000). Detecting lies and deceit: The psychology of lying and the

implications for professional practice. Chichester, England: John Wiley & Sons, Ltd.

Vrij, A., & Mann, S. (2001). Who killed my relative? Police officers’ ability to detect real- life high-stake lies. Psychology, Crime & Law, 7(1-4), 119-132.

Vrij, A., & Semin, G. R. (1996). Lie experts’ beliefs about nonverbal indicators of deception. Journal of Nonverbal Behavior, 20(1), 65-80.

Wood, J. L., Schmidtke, J. M., & Decker, D. L. (2007). Lying on Job Applications: The Effects of Job Relevance, Commission, and Human Resource Management Experience. Journal of Business and Psychology, 22, 1-9.

(30)

30

Weiss, B., & Feldman, R. S. (2006). Looking Good and Lying to Do It: Deception as an Impression Management Strategy in Job Interviews. Journal of Applied Social

Psychology, 36(4), 1070-1086.

Woehr, D. J., & Huffcutt, A. I. (1994). Rater training for performance appraisal: A quantitative review. Journal of Occupational and Organizational Psychology,

67(3), 189-205.

Zivin, G. (1982). Watching the sands shift: Conceptualizing the development of nonverbal mastery In R. S. Feldman (Ed.), Development of nonverbal behavior

in children, (pp. 63-98). New York: Springer-Verlag.

Referenties

GERELATEERDE DOCUMENTEN

Bij vermogensongelijkheid spelen verscheidene oorzaken een rol, onder andere rendement op vermogen dat hoger is dan economische groei, de eigenschappen van overerving, de toenemende

ktim in haar kis, skik haarself, en dan moet die bediendes die lang staanspieel bring en oor haar hou sodat sy kan sien hoe sy daarin lyk... D uskant Bethlehem is

Na deze instructie zal aan de studenten een korte vragen- lijst worden voorgelegd, waarin hun oordeel opnieuw wordt gevraagd over de practica massa-veer systeem 1 en 2 en

ANDANTEK differentieels serie SR kunnen worden gebruikt voor een groot aantal toepassingen.. Enkele voorbeelden zijn hieronder

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of

Even though the automated movement analysis for overall and specific body regions did not yield significant results between the experimental conditions, we did find a

Full body pose estimation: Comparison of body part detection rates and evaluation of different components of the model on the “Iterative Image Parsing” (IIP) dataset [15]

normen: goed betrouwbaarheid: goed begripsvaliditeit: goed criteriumvaliditeit: goed - vragenlijsten met uitslagformulier: €28,15 (20 stuks) - handleiding: €53,00 SRS