A test of the optimal classifier's independence assumption in perceptual categorization

Area of Science:

• Cognitive psychology
• Decision science
• Perceptual learning

Background:

• Understanding how individuals make decisions under uncertainty is crucial in cognitive psychology.
• Previous research has explored the influence of base rates and payoffs on categorization tasks.
• The independence assumption in optimal classification models has been a point of theoretical debate.

Purpose of the Study:

• To investigate human performance in perceptual categorization tasks under various base-rate and payoff manipulation conditions.
• To compare the predictive power of the flat-maxima hypothesis against the independence assumption.
• To evaluate a hybrid model combining the flat-maxima and COBRA hypotheses.

Main Methods:

• Participants completed perceptual categorization tasks with separate, corresponding, and conflicting simultaneous base-rate/payoff manipulations.
• Performance metrics were analyzed in relation to different base-rate/payoff ratios (2:1 vs. 3:1).
• A hybrid model integrating the flat-maxima and competition between reward and accuracy maximization (COBRA) hypotheses was applied and compared to a baseline model.

Main Results:

• Performance was closer to optimal with 2:1 ratios compared to 3:1 ratios.
• Accuracy improved when base rates were manipulated versus when payoffs were manipulated.
• Data aligned better with the flat-maxima hypothesis than the independence assumption in simultaneous manipulation conditions.
• The hybrid model demonstrated superior performance compared to the independence assumption model.

Conclusions:

• Violations of the independence assumption in decision-making are expected and effectively modeled by the flat-maxima hypothesis.
• Observers' decision criteria adapt quickly towards reward maximization.
• Accuracy maximization is weighted more heavily in separate and corresponding simultaneous conditions than in conflicting ones.