PAPER REVIEW
PRANJAL SAXENA
Y9424
To catch a liar: The effects of truthful and
deceptive testimony on inferential learning
Robert Montague, Daniel J. Navarro, Amy Perfors, Russell Warner, & Patrick Shafto
Department of Psychological and Brain Sciences, University of Louisville
School of Psychology, University of Adelaide
INTRODUCTION AND AREA OF WORK:
Deception is an act of creating a false belief in the mind of someone else. This paper aims to study deception involved in conceptual learning processes where by individuals build up information solely on the basis of testimony provided by others. The informers can be split into two categories, helpful and deceptive. Deceptive informers either lie outright, or can be economical with the truth, providing misleading information. Learners have to rely on this information , however they possess the option of verifying testimony in both the cases.
The previous studies in this area have focussed on picking up verbal , non verbal and visual clues to identify deception. Acoording to Bond and DePaulo,2006, a meta analysis shows that we are more adept in catching audible than visible lies and people appear more deceptive when there is an external motivation to believe them. Ekman & O'Sullivan, 1991 conducted experiments involving police officers, judges and psychiatrists where they were asked to distinguish a truthful experience by a person from a false one. They showed that the ability to distinguish truthful statements from lies is purely based on chance, and people performing better, often were able to pick up different behavourial clues and possessed specific skill set. This paper though,deals with recognising deceptive interactions over time, where different visual and non visual clues are absent and the learners must solely rely on the testimony of the informers , learners try to infer whether the informer is truthful or deceptive through verification, and as a result, choose to continue asking for information or learn on their own. The authors primarily aim to answer three questions through their study:
1. "First, how does the information provided by helpful and deceptive informants differ?
2. Second, how often do learners verify the information provided, and does this differ based on whether the informant is deceptive or not?
3. Third, how effectively do people recognize when they are being deceived, given only the content of the testimony"
-questions are directly cited from the paper,page 2 column1, second last paragraph.
EXPERIMENT:
For the purpose of the study a game was designed where, pairs of participants were divided into 2 broad groups: cooperative and competitive. Within each subgroup, individuals were randomly assigned the role of an informer or a learner. An interface was built where the objective of the exercise was that the learners should identify a rectangular area in a broad white field space. Cooperative and competitive informers provided helpful or misleading hints respectively in the form of interior or exterior points of the rectangle. The learners also had the option to 'dig' where they could click on a point and find out if it is an interior or exterior point, which also gives the option of verifying the testimony. Learners had five chances to gather information for a particular field and twenty fields were tested with the same informer, thus giving them a time space to infer on credibility of information by the informer. Some sample images of the interface(directly cited from the paper,page 1 right column) are shown below:
RESULTS OBTAINED:
Broadly, the learners' rectangle differed from the informers' rectangle more in the competitive casethan the cooperative case: (graph cited from paper, page 3, column 1)
However as an answer to the first question, the cooperative hints aimed at providing the points of the edges, whereas competitive informers chose to mislead by providing points distant from the exterior or comfortably placed at the interior. This information, although being truthful dsnt help the learner to reproduce the rectangle. Therefore ,it is interesting to note that they chose to mislead the learner by saying the truth rather than lying outright.
For the second question, the results show that the learners chose to verify about 30% of the hints in both cases with no significance escalation in the competitive case. This might be because, learners might not want to lose on a chance by checking the hints, rather use all the opportunities to establish the rectangle by digging themselves
Finally, on the basis of a rating by the learners, it was seen that they could identify deceptive informers from the helpful ones even when there was no change in the number of verified points. The learners based their judgement on the nature of hints received, for example, corner points enabled faith on the informers, and they ended up asking for more hints.
DISCUSSION:
A sparkling conclusion that the paper provides is that deceptive informants more often than not choose to provide misleading truthful information rather than lying outright. For example, a murder suspect might get away by providing a mix of information to the interviewing officer, where he puts enough truthful testimony which the officer can verify , thus maintaining trust, however adding enough evasive information which does not enable the investigation to proceed on the right track. Based on the results of the conducted experiment, the learners identify deceptive informants on the judgement whether the testimony has proved helpful or not, instead of verification at every step. Therefore from the investing officer's perspective, it would be ideal for him to compare the testimony of the witness with perfect testimony : ideal information that would allow him to solve the case. Any discrepancies would indicate the fact that the suspect is trying to mislead and is involved in the crime.
Another aspect that the results highlight is that learners put more trust on those informers who intend to give hints that the learners have preconceived to be helpful. Therefore learners form an expectation related to helpful hints which might enable them to locate the rectangle. This is in accordence with the work of Bonawitz et al., 2009; Xu & Tenenbaum, 2007 who showed that learners tend to form a narrow outlook over purposefully sampled information and do not look at all the attributes which they would have for a randomly sampled information.