Abstract: . . .  sub-optimally because they do not really care about the outcomes of their decisions, then little is learned. If, however, they exhibit departures from normative theory when real money is on the line, something interesting may be going on. Good basic experimental protocols are too numerous to name and largely depend on the nature of the experimental task. We do not discuss these here. Below, we review some of the work in experimental psychology that has relied on compu- tational procedures from operations research. In particular, we focus on multi-stage decision problems that can be solved by dynamic programming. We do so for two reasons. First, this will provide a certain cohesion. Second, much of our own work involves experimental inves- tigations of behavior in dynamic decision problems. Hence, when discussing these problems, we can give an insider’s view of ways in which experimentalists think about the OR methods and how we think these methods can inform us about human cognition. Although . . .
. . .  programming. We do so for two reasons. First, this will provide a certain cohesion. Second, much of our own work involves experimental inves- tigations of behavior in dynamic decision problems. Hence, when discussing these problems, we can give an insider’s view of ways in which experimentalists think about the OR methods and how we think these methods can inform us about human cognition. Although we restrict this review to experimental studies of sequential observation and selection behavior, OR researchers should be aware that other areas of dynamic decision making have been brought into the laboratory. Toda [63] pioneered the study of multi- stage decision behavior more than forty years ago. He devised a one-person game called Page 3 Bearden and Rapoport: OR in Experimental Psychology INFORMS—New Orleans 2005, c 2005 INFORMS 3 the “fungus-eater” game, in which subjects were asked to control the sequential search of a robot that attempts to maximize the amount of a valuable resource (“uranium”) . . .
. . .  could earn up to $50 for the one hour session, whereas those in Experiment 2 could earn up to $40. As with our previous experiments on the GSP, we used a hiring cover story, and instructed the subjects that the attributes were uncorrelated (in language they could easily understand). Threshold representations of the optimal policies for the two MASPs examined in Exper- iments 1 and 2 are shown in Table 5. The cell entries for a pair of relative ranks ( r 1 ,r 2 ) Page 16 Bearden and Rapoport: OR in Experimental Psychology 16 INFORMS—New Orleans 2005, c 2005 INFORMS Table 3. MASP Example Problem. Payoff Values a 1 2 3 4 5 6 p 1 ( a ) 6 5 4 3 2 1 p 2 ( a ) 5 4 3 2 0 0 Example Applicant Sequence Applicant ( j ) 1 2 3 4 5 6 a 1 j 2 4 3 6 5 1 a 2 j 5 2 1 3 6 4 r 1 j 1 2 2 4 4 1 r 2 j 1 1 1 3 5 4 Optimal Policy and Payoffs Applicant ( j ) 1 2 3 4 5 6 V * j +1 7.82 7.67 7.37 6.83 5.83 — E (? j | r j ) 5.83 5.73 7.55 2.93 1.83 8.00 ? j 5.00 7.00 9.00 4.00 2.00 8.00 correspond to the applicant position . . .
--3000,3,500,3251,53584