Modeling speed-accuracy trade-offs in the stopping rule for confidence judgments

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Decision Science

Background:

• Decision-making and confidence reporting are often explained by evidence accumulation models.
• The stopping rule for post-decision evidence accumulation, crucial for confidence, remains underspecified.
• Existing models struggle to fully account for the dynamics of confidence judgments.

Purpose of the Study:

• To compare five prominent models of confidence within evidence accumulation frameworks.
• To investigate the stopping rule for post-decision evidence accumulation.
• To examine how speed-accuracy instructions affect decision and confidence boundaries.

Main Methods:

• Collected behavioral data from two experiments with varying decision and confidence speed-accuracy instructions.
• Quantitatively and qualitatively compared five evidence accumulation models of confidence.
• Proposed and tested a novel flexible confidence boundary model.

Main Results:

• The flexible confidence boundary model provided the best account of the experimental data.
• Decision-making speed-accuracy instructions modulated decision boundary heights.
• Confidence judgment speed-accuracy instructions modulated confidence boundary heights.
• Post-decision evidence accumulation to a bound explains confidence judgment stopping rules.

Conclusions:

• Confidence judgments, like decisions, can be modeled as an accumulation-to-bound process.
• The height of confidence boundaries is strategically controlled based on task demands.
• The flexible confidence boundary model offers a more complete explanation of confidence reporting.