Decisions on alternative options often involve risks and uncertainties. Should we buy flood insurance or would we rather save the money? Do we prefer a high-risk or a low-risk stock market investment? When do we choose surgery rather than chemotherapy for cancer treatment? Should we take a risky option and try to save all the people with a one-third chance of success or should we avoid the risk and save one-third of them for certain?
This research project examines how people automatically search, select, and use social and communicational cues under real world constraints to determine their risk behavior. Our previous findings have shown that people are sensitive to cues of kinship, anonymity, group size, group constitution, and hedonic verbal framing/phrasing of choice options and place different priorities upon these decision cues. These pieces of information embedded in a decision problem determine the minimally acceptable outcome, which is called the "minimum requirement" (MR) for a particular task. According to the proposed mean-variance model of risky choice, a decision maker tries to maximize the chance of reaching the MR. Suppose there are two choice options: a sure gain of $100 vs. a gamble with a 50% chance of wining $300 and a 50% chance of losing $100. That is, the two options have the same mean outcome but different variances.
The mean-variance model takes into account the relationship between the MR, the expected mean values of choice options, and the variances in those choices. When the expected mean value of the options is below the MR, a decision maker is expected to favor a high variance (gamble) option over its sure-thing equivalent, because higher variance increases the chance of reaching the MR. In contrast, when the mean value is above the MR, the participants are expected to be risk averse and favor the sure thing over the gamble to avoid possible harms (i.e., falling below the MR).
The model also predicts that the mean-variance heuristic will be reversed either when MR is higher than the mean plus variance or when MR is extremely low. The predictions will be tested in different task domains (e.g., life vs. money), with different expected values (e.g., 600 lives vs. 6 lives; $6000 vs. $600), different payoffs (e.g., the number of lives vs. life-span expectancy) and different samples of the participants (e.g., students, local residents, professionals), in different task conditions (e.g., feedback by computer displays, lottery games), under both positive and negative framing conditions, and cross-culturally.
