On the generality of optimal versus objective classifier feedback effects on decision criterion learning in...

Area of Science:

• Cognitive Psychology
• Decision Science
• Machine Learning

Background:

• Human decision-making often deviates from reward maximization, prioritizing accuracy even when suboptimal.
• Biased payoff matrices can create conflicting goals between maximizing reward and accuracy.

Purpose of the Study:

• To compare the effectiveness of objective classifier feedback versus optimal classifier feedback on decision criterion learning.
• To investigate how category discriminability and cost of errors influence the advantage of optimal feedback.

Main Methods:

• Participants received feedback based on either the objective (correct) or optimal (best possible) classifier response.
• Category discriminability and cost of incorrect responses were manipulated.
• Model-based analyses assessed the weight given to accuracy versus reward.

Main Results:

• Optimal classifier feedback led to superior performance compared to objective feedback, particularly with low category discriminability.
• The advantage of optimal feedback was consistent across zero and negative error costs.
• Optimal classifier feedback reduced the weight placed on accuracy and this effect diminished with training.

Conclusions:

• Feedback based on the optimal classifier promotes better decision criterion learning than objective feedback.
• This suggests that providing guidance on optimal responses can improve human decision-making performance and learning.