Experience and prior probability in a complex decision task.

Abstract

Six experienced and six naive 5s evaluated probabilistic data, determined sources of data generation, and predicted subsequent data in a complex decision task. Experience and prior probability were combined factorially. Results indicated that experienced 5s (a) were less conservative data evaluators, (6) determined data sources on the basis of fewer data samples, (c) were more sensitive to prior-probability values, and (d) adopted a maximization strategy in prediction more consistently than did naive 5s. The importance of using trained personnel in the evaluation of realistic decision capabilities and the need for caution in generalizing from data obtained from naive 5s who serve in most laboratory studies of decision making were discussed.

Errors in probabilistic reasoning and judgment biases

“…41 The papers for which this is true are Peterson, Ulehla, Miller, Bourne, and Stilson (1965), Peterson, Schneider, and Miller (1965), , Phillips, Hays, and Edwards (1966), Beach (1968), Chinnis and Peterson (1968), Dale (1968), Peterson and Swensson (1968), Sanders (1968), Beach, Wise, and Barclay (1970), Edenborough (1975), Dave and Wolfe (2003), Kraemer and Weber (2004), and Sasaki and Kawagoe (2007). The one, partial exception is Strub (1969), who finds that while it is true for naïve experimental participants, participants with extensive training update Bayesianly.…”

“…Posteriors were closer to Bayesian at the end of the fourth sequence than at the end of the first sequence but remained not extreme enough. Strub (1969) ran a sequential-sample updating experiment among a group of naïve participants and a group of trained participants, undergraduates who had received 114 hours of lecture sessions, demonstrations, problem-solving sessions, and other training in dealing with probabilities, including prior participation in bookbag-and-poker-chip experiments. Relative to the naïve participants, the trained participants had final posteriors that were much closer to Bayesian on average across updating problems, but the results are not reported in enough detail to evaluate whether the trained participants had biased beliefs in the updating problems considered separately.…”

Errors in Probabilistic Reasoning and Judgment Biases

“…42 The papers for which this is true are Peterson, Ulehla, Miller, Bourne, and Stilson (1965), Peterson, Schneider, and Miller (1965), , Phillips, Hays, and Edwards (1966), Beach (1968), Chinnis and Peterson (1968), Dale (1968), Peterson and Swensson (1968), Sanders (1968), Beach, Wise, and Barclay (1970), Edenborough (1975), Dave and Wolfe (2003), Kraemer and Weber (2004), and Sasaki and Kawagoe (2007). The one, partial exception is Strub (1969), who finds that while it is true for naïve experimental participants, participants with extensive training update Bayesianly.…”

“…In addition to testing for deviations from Bayesian updating, he also studied the robustness of these deviations to incentives for correct answers and experience with the same updating problem. In earlier work from psychology, there was also some attention to the effect of incentives (e.g., and experience (e.g., Martin and Gettys, 1969;Strub, 1969).…”

Subjective Probability and Its Measurement

“…Cohen and Hansel (1958) submit that age, experience and the number and value of the alternatives offered are factors of importance. Later studies (Strub, 1969)support the use of trained personnel in complex decision tasks. They apparently processed information better, were more aware ofthe meaning of the prior probability values and adopted a maximization principle more consistently in that they tried to obtain the optimum return from each decision task.…”