Naturalistic multiattribute choice.

Area of Science:

• Cognitive Psychology
• Decision Science
• Computational Social Science

Background:

• Understanding how individuals evaluate and combine attributes of complex items like movies or food is crucial.
• Semantic memory theories offer insights into object representation, but their application to large-scale choice data is underexplored.

Purpose of the Study:

• To develop and test multiattribute representations for common choice objects using crowd-sourced data.
• To evaluate the predictive accuracy of various decision rules for aggregating these representations.
• To identify factors influencing accurate prediction of human choice behavior.

Main Methods:

• Leveraged large-scale crowd-sourced data to derive multiattribute representations of choice objects.
• Applied theories from semantic memory research to model object attributes.
• Conducted standard choice experiments to test decision rules against observed behavior.

Main Results:

• Existing multiattribute decision rules accurately predicted participant choices when using crowd-sourced object representations.
• Weighted additive decision rules demonstrated superior out-of-sample predictive performance compared to heuristic rules.
• Decision rules incorporating rich object representations with numerous attributes achieved the highest accuracy.

Conclusions:

• Multiattribute decision rules, when trained on detailed crowd-sourced data, effectively predict human choice behavior.
• The superiority of weighted additive rules challenges previous findings favoring heuristic approaches in certain contexts.
• The complexity and richness of object representations are key determinants of decision-making accuracy.