Martin G. Evans
Faculty of Management
University of Toronto
Toronto Ontario M5S 1V4
Canada
E-mail: evans@fmgmt.mgmt.utoronto.ca
Send e-mail
In this paper we apply the tenets of evolutionary psychology to the auditor. This recent
advance in psychology suggests that a variety of production mechanisms evolved in the
Pleistocene and remain with us to this day. One such mechanism is a built in algorithm for
detecting cheaters. Using a scenario describing a company pension plan we explore three
perspectives, two are unilateral involving either an employee or an employer. The third is
bilateral involving an auditor who is looking for violations of the pension plan's rules. No
support is found for the view that bilateral cheating detection is an evolutionary residue.
Auditors should be aware that their penchant for detecting error in the work of others has a strong evolutionary underpinning. One of the more interesting trends in psychological research is that associated with Leda Cosmides and her colleagues (Cosmides, 1989; Cosmides & Tooby, 1989, 1994). In a startling series of papers she has argued that humankind during its evolutionary period in the Pleistocene, some 20 million years, developed a built in "cheater detector."
The argument is that in order to survive in an environment in which the availability of food and water was "lumpy" one needed to cooperate with others and that this cooperation entailed the sharing of food and water. Someone, or some clan, that was successful on one occasion would share with others; the others were expected to reciprocate when they got lucky. In the situation of reciprocal altruism, reproductive fitness would be enhanced for those members of the species who could detect (and punish) defection. In other words, if I did you a good turn, I needed to be aware if you failed to reciprocate. Those with this detection facility were less likely to die young and hence more likely to reproduce than those lacking the "cheater detector."
The contemporary evidence for this remarkable assertion lies in a series of studies of that most mundane of psychological phenomena: logical reasoning. Since the development of the modern research program on reasoning (Johnson-Laird & Wason, 1970), there has been a good deal of controversy about the explanation for the context effect found in the Wason reasoning task. This task has been described best by Kirby (p 2):
In Wason's four card selection task subjects are presented with four cards that are constrained to have instances from the sets P or not-P on one side and Q or not-Q on the other side. A conditional statement describes an alleged relation between the fronts and the backs of the cards: if a P is on one side of the card, then a Q is on the other. Subjects' task is to select which of the cards should be turned over to determine whether the conditional relation is true or false. ... subjects are usually considered in error when they fail to select the P or not-Q, or when they select the not-P or Q cards, which are logically incapable of violating the conditional.It is usual to find that when the task is represented simply as a logical syllogism few subjects (about 10%) solve the problem. When the task is given the trappings of a familiar situation, hit rates go to about 25%. However this result is not always found, and in some situations (e.g., (Griggs & Cox, 1982) "bar scene2" the hit rate can be as high as 75%. It was Cosmides' (1989) insight that showed that high hit rates were only found when the descriptive story had the trappings of a social contract (i.e., not being allowed to drink until one attained a certain age). Support and refinement of this position has come from (Gigerenzer & Hug, 1992) who established that it is not enough for the phrasing to be in terms of a social contract, there must be an attempt to violate this contract, rather than merely establish the existence of such a contract.
It was one of the scenarios in (Gigerenzer & Hug, 1992) that motivated this study. They describe a situation in which an employer offers a pension plan to all those employees who have more than ten years service. In a clever reversal, (Gigerenzer & Hug, 1992) wrote two versions of the story, one in which the subject took the role of an employee who was trying to detect whether or not the employer was fulfilling his bargain by giving all ten-year veterans a pension; in the second version, the subject took the role of the employer who was examining whether or not some nine year veterans were getting undeserved pensions.
The formal rule presented was :
"if a previous employee gets a pension from a firm [p], then that person must have worked
for the firm for at least ten years [q]" (Gigerenzer & Hug, 1992). Hit rates (involving
different pairs of cards for the two versions) were high (about 65%). For the employer's
perspective one needs to check the p and not-q cards; from the
employee's perspective one needs to select the not-p and the q cards.
The symmetry of this switch between employer and employee immediately suggests a third perspective. Under Ontario law, pensions have to be audited. To the auditor, any violation of the rules is reprehensible3. An employee getting an undeserved pension is as bad as an employer failing to pay a pension to a deserving retiree. The question is whether this bilateral cheating detector is as powerful as the unilateral cheater detector. The study reported here was designed to explore this question.
As part of a larger study involving six Wason-type tasks, subjects were given one of three versions of the (Gigerenzer & Hug, 1992) "pension" task. The three versions are given in Appendix A. In each version a strong rationale for taking each perspective (employer, employee, auditor) was given. Each subject received just one of the "pension" scenarios embedded with another five tasks. Task order was randomized with one constraint, a task with no surrounding rationale came either first or second in the questionnaire. The responses (p, not-p, q, not-q) to each story were also randomized.
Subjects were 82 third year Commerce students at the University of Toronto. There were 44% men and 56% women in the sample. Very few had any formal training in logic. The average age was 23.3 years (s.d. 4.5).
Two dependent variables were measured. The first is the number of hits in each condition (a hit is defined on indicating the appropriate cards for further investigation:
| Employer: | pension | worked eight years | ||
| Employee: | no pension | worked more than ten years | ||
| Auditor: | all four cards. | |||
The second, used for the audit condition only was the total number of cards turned over. The higher the number of cards, the nearer the subject is to a "hit."
As a baseline, a formal logical task was performed. Only 2 (2.4%) of the students solved it correctly. Hit rates were much higher for the employer and employee versions of the "pension" scenario: Employer, 27%; Employee, 40%. These differences are not significant. These hit rates are much lower than the 80% observed by Gigerenzer & Hug (1992). The hit rate for the audit version was not high, 3%. Most subjects only turned over one or two cards (mean 1.78 cards). This 3% hit rate for bilateral detection is much lower than the hit rates for both forms of unilateral detection.
We are unable to explain why the hit rates for the unilateral (employer or employee) scenarios are so much lower than those of Gigerenzer & Hug (1992). Some of the difference is clearly due to the logic training instilled in the German gymnasium, Gigerenzer & Hug (1992) allude to this in their explanation for their much stronger results in comparable problems (e.g., the "bar scene") than US students4
We identified one interesting additional finding that has implications for the independence of the auditor. We computed scores to identify whether or not the auditor was more employee or employer oriented. Recall that for the employer to detect whether or not an employee was cheating, the employer would have to examine the "pension" and "worked eight years" cards, similarly the employee, to detect employer cheating would have to examine the "no pension" and "worked over ten years" cards. Two scores were computed for each subject in the audit condition: the number of selected cards in the employer frame and the number of selected cards in the employee frame. The subjects in the audit condition were biased toward the employer's perspective. They chose 1.06 cards from that pair and only 0.72 from the employee frame. This difference approaches significance (p<.10, n=32).
If the evolutionary cheater detector exists, it exists only in a unilateral form. The mechanism for a bilateral cheater detector has not evolved. Professors of Accounting and auditing may relax, their training is still required. Perhaps a first focus would be to insist on even handed auditing of employee pension plans.
Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31, 187-276.
Cosmides, L., & Tooby, J. (1989). Evolutionary psychology and the generation of culture, Part II Case Study: A computational theory of social exchange. Ethology and Sociobiology, 10, 51-97.
Cosmides, L., & Tooby, J. (1994). Beyond intuition and instinct blindness: towards an evolutionarily rigorous cognitive science. Cognition, 50, 41-77.
Evans, M. G. (1994). Information in the Wason task: A further look. Submitted for publication. Gigerenzer, G., & Hug, K. (1992). Domain-specific reasoning: Social contracts, cheating, and perspective change. Cognition, 43, 127-171.
Griggs, R. A., & Cox, J. R. (1982). The elusive thematics material effect in Wason's selection task. British Journal of Psychology, 73, 407-420.
Johnson-Laird, P. N., & Wason, P. C. (1970). A theoretical analysis of insight into a reasoning task. Cognitive Psychology, 1, 134-148.
Under the new Ontario law, it is necessary for auditors to audit pension plans to ensure that all the rules of the pension plan are followed fully. Your assignment is to carry out the audit of the pension plan of a small manufacturing plant that has been operating for about thirty years. Under the contract which the firm has signed with the union, it is specified that a retiree only receives a pension if he/she has worked for the firm for ten years or more. This incentive is important to keep people working at the firm for a long period as the work is not very interesting, however the work does require firm specific knowledge that only accumulates over time. The employer wants to have a good proportion of long service employees in the workforce.
From your previous experience you know that all kinds of plan violations can occur. You have pulled four files at random from the records of previous employees. Someone has made notes on the file covers. The notes on each cover are incomplete.
Indicate only those file(s) which you would definitely need to examine to see if there had been any violations of the pension plan's rules. Please circle.
| Worked for more than 10 years | Worked for 8 years | Received a Pension | No Pension |
You work as a clerk in the personnel office of a small manufacturing company which has been in operation for over thirty years. Under the contract which the firm has signed with the union, it is specified that a retiree only receives a pension if he/she has worked for the firm for ten years or more. This incentive is important to keep people working at the firm for a long period as the work is not very interesting, however the work does require firm specific knowledge that only accumulates over time. You therefore want to be sure that you get your pension if you put up with the tedium for ten years or more.
There have been rumours that, in the past, the employer has not lived up to his part of the bargain (i.e., by not counting time on layoff as contributing to years of service even though the contract clearly states that such time should be counted), so that some employees with more than ten years of service may not have received pensions. Before deciding whether to stay any longer with the firm, you want to check this out.
You have pulled four files at random from the records of previous employees. Someone has made notes on the file covers. The notes on each cover are incomplete.
Indicate only those file(s) which you would definitely need to examine to see if the employer was ripping off any of the previous employees.
| Received a Pension | Worked for more than 10 years | No Pension | Worked for 8 years |
You are the owner of a small manufacturing plant that has been operating for about thirty years. Under the contract which you have signed with the union, it is specified that a retiree only receives a pension if he/she has worked for the firm for ten years or more. This incentive is important to keep people working at the firm for a long period as the work is not very interesting, however the work does require firm specific knowledge that only accumulates over time. You therefore want to have a good proportion of long service employees in the workforce.
You have recently begun to suspect that the people running the payroll department have been giving their friends the benefit of the doubt in calculating length of service (i.e., counting time on layoff as employment time even though this is specified as not eligible time in the contract) so that it is possible that some people with less than ten year's service have received pensions.
You have pulled four files at random from the records of previous employees. Someone has made notes on the file covers. The notes on each cover are incomplete.
Indicate only those file(s) which you would definitely need to examine to see if any of the retirees were ripping off the company.
| Worked for more than 10 years | Received a Pension | Worked for 8 years | No Pension |