Preference reversal and expert
Preference reversal and expertSubjects in gambling tasks that involve both choice and pricing show a pattern of responses known as preference reversal. That is, although subjects in a choice condition generally will give higher preference ratings to “safe”, high-probability/low-payoff, bets than to “longshot”, low-probability/high-payoff, bets, when they are asked in a pricing condition to generate an amount of money that they would accept to avoid the gamble altogether they tend to give higher values for longshots over safer bets. Tversky, Slovic, and Kahneman (1990) demonstrate that among the several possible actions that subjects could be taking to produce this pattern, the critical factor appears to be the overpricing of the longshot bets. If subjects are actually offered a monetary figure (hypothetically) by the experimenter to replace the gamble, they will accept this figure even though it is lower than the figure that they generated in the pricing condition. Tversky et al. (1990) further showed that this overpricing is largely due to a phenomena known as scale compatibility, which involves certain biases when the response required by the subject is in the same units as the factors influencing the decision. Since the payoffs of the bets and the buy-out prices assigned to them are both monetary values, this leads people to give greater weight to the payoff value of the bets when asked to price them (a situation of compatibility) than when asked to choose between them (a situation of non-compatibility).
The development of expertise in avoiding preference reversal, then, would have to involve the circumvention of the compatibility effect. One possible way in which this could occur would involve subjects consistently selecting either payoff or probability as the critical factor in both choice and pricing conditions. By adopting a strategy of maximizing the chance of any payoff in both the choice and pricing condition and giving that option the higher rating on both scales, preference reversal would be avoided. Conversely, considering only the greatest potential for gain in each condition would have the same effect.
This strategy, however, would be susceptible to preference reversals in the other direction. In the first case of maximizing the chance of payoff, the safe bet (H) would be favored over the longshot (L) and the pricing would also favor the safe bet (Ch) over the longshot (Cl) (i.e. Ch Cl). Yet when any amount of money (X) is offered at a %100 probability, that option would be selected over both H and L. That is for Ch X Cl, X is preferred over H and X is preferred over L, leading to Ch X H, an overpricing of H. Subjects would accept less money than their given value of Ch, or even of Cl as long as that money was guaranteed. In the opposite situation where the potential payoff is the critical factor, L would be favored over H and Cl Ch. But here, the buyout X would never be accepted because both H and L will always have a greater potential payoff. This results in an underpricing of L and of H, since Cl and Ch will not be any different than L and H respectively.