I just got back from Boston where I attended the American Psychological Association Meeting. 15,000 psychologists together in Boston.
I had to attend quite a few administrative meetings, but I also presented a poster. The poster was actually presented along with my coauthors, Kevin Mahoney, Jeffrey Walczyk, and Diana Griffiths-Ross of Louisiana Tech University. It dealt with the topic of detecting dishonesty during the employment interview and with an alternative to the polygraph.
People are not always truthful. During the employment interview, organizations are interested in detecting two different types of dishonesty. First, they are interested in identifying those individuals who might engage in questionable behavior on the job. Second, they are interested in identifying those people who lie about their background or abilities during the interview itself. Our paper, and the method described in the paper, dealt with the latter.
Basically, Jeff Walczyk has developed a cognitive lie detector which is based upon response time. This cognitive lie detector is called Time Restricted Integrity Confirmation or TRI-Con. TRI-Con is designed to maximize cognitive load on liars, minimize it on truth tellers, and expose load-reducing countermeasures. Bottom line, lying takes time and thus, in theory, you can detect lying by measuring how long it takes to respond to questions. In the past, we have demonstrated that response times can discriminate truth tellers from both rehearsed and unrehearsed liars with 75% to 90% accuracy. Unfortunately, as with many things, additional research with the method is required. Also, it probably takes even higher rates of accuracy to have sufficient confidence to label someone as dishonest.
{ 5 comments… read them below or add one }
First off, Dennis, did everybody know your name?
Second, as much as I think reaction time is a too-often overlooked secret weapon in assessing a lot of things (e.g., emotional intelligence), you have to wonder about the viability of using reaction time in real-world assessment situations, particularly high stakes testing. Keep in mind that, while interviewing may often only be done when one has whittled down a list of applicants to a smaller set of candidates on a short list, the likelihood of it being used in a consistent manner across candidates within an organization is shaky at best.
Still, as guys like Paul Ekman can tell you, you don’t need a boatload of equipment to learn HOW to detect lies.
Question: Does the response lag depend on the type of lie being told? For instance, is response time added only when one has to muster a lie for something that could be “checkable”, or only when the behaviour lied about is important?
Mark,
Thanks again for your comments. As you might guess, I try to stay incognito. I usually find that is for the best. That is why I am wearing sunglasses.
I agree with you on reaction time. We have been doing reaction time research as it relates to assessment since back in the early 1980s. The problem remains the same – it is hard to get reliable data at the individual level, or make that reliable overall individual scores. Reaction time data tends to perform much better in either 1) experimental studies where you are looking at groups or 2) laboratory studies of individual differences. So, not just for detecting lying but in most areas, the use of reaction time in real world assessment situations is limited by the reliability of the calculation of reaction time scores at the individual level.
As for your question about response time and type of lie, again the answer is yes. For example, the rehearsal of inaccurate responses leads to shorter responses time. Importance, the type of behavior, and the nature of the fabrication all impact the response time measure.
So, in the end, there is still work and research to do, although this may represent a promising area.
I agree that reaction time has problems. If does not work at the individual level, then it cannot be used in the field of lie detection. In fact, the present widely-accepted form of lie detection uses control questions similar to the general questions used in this research. Such questions are meant to control for individual differences, and such controlling cannot occur comparing only across individuals, as was the case in this research. The truth is that lying is unique to the individual, which is why “Mom can tell because she knows you well.”
Because the polygraph reportedly produces about an 80% hit rate without accounting for contradicting responses on an individual level by comparing responses to questions within-subject, it is important that further research with TRI-Con reach at least the same hit rate without the added effects of consistency of responses. Moreover, it is imperative that a within-subject approach, along with a between-subjects, be taken so that the results can be said to have produced a hit rate, rather than reaching a certain accuracy. This has yet to be done, as we have not tried to look at an individual’s data and predict from that whether the individual was lying or telling the truth. However, one study that we have planned under the funded NSF reserach that we are now conducting, does propose to try to do just that. With the equipment that has been added that will allow for the measurement of more than response time and the new techniques for assessing the data, the results should be interesting and will hopefully provide much support for The Activation-Decision-Construction Model of Lying that underlies TRI-Con.
I agree that reaction time has problems. If does not work at the individual level, then it cannot be used in the field of lie detection. In fact, the present widely-accepted form of lie detection uses control questions similar to the general questions used in this research. Such questions are meant to control for individual differences, and such controlling cannot occur comparing only across individuals, as was the case in this research. The truth is that lying is unique to the individual, which is why “Mom can tell because she knows you well.”
Because the polygraph reportedly produces about an 80% hit rate without accounting for contradicting responses on an individual level by comparing responses to questions within-subject, it is important that further research with TRI-Con reach at least the same hit rate without the added effects of consistency of responses. Moreover, it is imperative that a within-subject approach, along with a between-subjects, be taken so that the results can be said to have produced a hit rate, rather than reaching a certain accuracy. This has yet to be done, as we have not tried to look at an individual’s data and predict from that whether the individual was lying or telling the truth. However, one study that we have planned under the funded NSF reserach that we are now conducting, does propose to try to do just that. With the equipment that has been added that will allow for the measurement of more than response time and the new techniques for assessing the data, the results should be interesting and will hopefully provide much support for The Activation-Decision-Construction Model of Lying that underlies TRI-Con.
I agree that reaction time has problems. If it does not work at the individual level, then it cannot be used in the field of lie detection. In fact, the present widely-accepted form of lie detection uses control questions similar to the general questions used in this research. Such questions are meant to control for individual differences, and such controlling cannot occur comparing only across individuals, as was the case in this research. The truth is that lying is unique to the individual, which is why “Mom can tell because she knows you well.”
Because the polygraph reportedly produces about an 80% hit rate without accounting for contradicting responses on an individual level by comparing responses to questions within-subject, it is important that further research with TRI-Con reach at least the same hit rate without the added effects of consistency of responses. Moreover, it is imperative that a within-subject approach, along with a between-subjects, be taken so that the results can be said to have produced a hit rate, rather than reaching a certain accuracy. This has yet to be done, as we have not tried to look at an individual’s data and predict from that whether the individual was lying or telling the truth. However, one study that we have planned under the funded NSF reserach that we are now conducting, does propose to try to do just that. With the equipment that has been added that will allow for the measurement of more than response time and the new techniques for assessing the data, the results should be interesting and will hopefully provide much support for The Activation-Decision-Construction Model of Lying that underlies TRI-Con.