This dissertation work examined (1) how people search, select, and use information (cues) in decision task contexts of public safety and medical diagnosis, and (2) how some well-known and newly developed choice strategies/heuristics match with the actual choices of participants in selecting and evaluating consumer products. We derived our hypotheses from a risk sensitive, Tri-Reference Point (TRP) model (Wang, 2008) of choice involving multiple decision cues/attributes and tested them in three studies. This project was intended to enrich the current understanding of information search and decision strategy use, and to contribute to the development of behavioral decision aids that are choice task-sensitive and reference point-dependent.
In Study 1 participants made inference and preference choices in a medical domain, determining which patient should receive drug treatment for a certain condition and selecting which specific drug to use, based on medical guidelines. In Study 2 participants made inference and preference choices in a public safety domain, determining which set of consequences are most likely to occur from a specific type of terrorist attack and selecting which Threat Condition on the Homeland Security Advisory System is most appropriate in response to the attack, based on summarized scenarios published by the US department of Homeland Security. In both studies the amount of available information and attribute variance (attribute value ranges) were systematically varied, and subjective task significance, and expertise were rated. The participants' information search was recorded via the decision process tracing tool Mouselab.
The findings from Studies 1 and 2 indicated that participants searched for information in an attribute-wise manner and were selective in the attributes and options they inspected. These search patterns suggest that participants applied frugal heuristics, such as 'Take-The-Best-based heuristics instead of comprehensive information-consuming search strategies. The attribute-wise search trend was weaker in the medical domain, suggesting that participants were likely to treat each option (a patient) as a whole information unit. The amount of available information and attribute variance did not have a significant effect on information search. We found only weak evidence for influences of subjective decision significance and domain expertise on information search, but significant gender differences existed in the search patterns.
In Study 3, participants rated how likely they would select and how satisfied they would be with certain popular electronic products. They also reported the product attributes/features they would consider in making a choice, their respective weights, and respective reference points minimum requirement (MR), goal, and status quo, and rated their subjective task significance and domain expertise. Based on the above information, individual choice outcomes were prescribed by twelve decision strategies including normative strategies Multi-Attribute Utility Theory (MAUT) and Additive-Difference (AD), and simple heuristics, such as the 'MR-based heuristic', 'Goal-based heuristic', 'Take-The-Best', and 'Take-The-Best' on your MR. The fits between strategy-prescribed outcomes and empirical preferences were analyzed.
The findings from Study 3 showed that the preference fits of strategies based on the reference-points MR and Goal were as good as or better than the fits of the reference-point independent strategies MAUT, AD, and TTB. In particular, the reference-point based 'Take-The-Best' on your MR and 'Take-The-Best' on your Goal showed better fits than the reference-point independent 'Take-The-Best'. Strategies based on the reference point Status Quo did not fit well with empirical choices. We found only weak evidence for influences of subjective decision significance and domain expertise on strategy use. In addition, significant gender differences existed in the outcome fits, where males' actual choices on average fit better than females' choices to the choice strategies tested in the study. Theoretical significance and practical implications of the aforementioned findings are discussed.