Related Experiment Videos
Optimal classifier feedback improves cost-benefit but not base-rate decision criterion learning in perceptual
1Department of Psychology, 1 University Station A8000, University of Texas, Austin, TX 78712, USA. maddox@psy.utexas.edu
Memory & Cognition
|July 21, 2005
Summary
When payoffs are unequal, people prioritize accuracy over optimal rewards. This study found that optimal classifier feedback, not objective feedback, better aligns decision criteria with reward maximization, especially with unequal payoffs.
Area of Science:
- Cognitive psychology
- Decision-making research
- Machine learning
Background:
- Decision-making involves balancing accuracy and reward maximization.
- Unequal payoffs can lead to suboptimal emphasis on accuracy.
- Existing hypotheses include the COBRA model.
Purpose of the Study:
- To test the COBRA hypothesis against alternative models (COBRM, COBRE).
- To compare objective classifier feedback with optimal classifier feedback under unequal payoffs and base rates.
- To investigate the impact of feedback type on decision criteria and accuracy-reward trade-offs.
Main Methods:
- Experimental comparison of objective vs. optimal classifier feedback.
- Manipulation of payoff inequalities and base rates.
- Model-based analyses of decision criteria and feedback effects.
Main Results:
- The COBRA hypothesis was supported: optimal feedback improved decision criterion leaning with unequal payoffs, unlike with unequal base rates.
- Optimal classifier feedback reduced the weight placed on accuracy compared to objective feedback.
- Delayed feedback influenced the learning of reward-maximizing criteria.
Conclusions:
- Optimal classifier feedback facilitates better reward maximization than objective feedback when payoffs are unequal.
- Decision-making under uncertainty is sensitive to feedback mechanisms and payoff structures.
- Understanding these dynamics is crucial for designing effective learning and decision support systems.