Evidence integration and decision confidence are modulated by stimulus consistency

Area of Science:

• Cognitive Neuroscience
• Decision Science
• Computational Psychology

Background:

• Evidence integration models decision-making with noisy data, but non-integration heuristics and elusive decision boundaries pose challenges.
• Existing methods struggle to differentiate between integration and heuristic strategies in real-time decision processes.

Purpose of the Study:

• To develop a model-free method for extracting decision boundaries.
• To provide empirical support for evidence integration over non-integration heuristics.
• To identify and characterize biases in evidence integration, specifically a consistency bias.

Main Methods:

• Introduced the 'decision classification boundary' (DCB) method, a model-free approach to identify choice boundaries based on accumulated evidence.
• Applied the DCB method across four cross-domain experiments to analyze decision-making behavior.
• Measured choice accuracy and decision confidence in response to varying stimulus consistency.

Main Results:

• Demonstrated that decision boundaries can be effectively extracted using the DCB method.
• Provided direct evidence supporting evidence integration over non-integration heuristics.
• Identified a 'consistency bias' where incoming evidence is weighted based on its alignment with prior information.
• Observed that this consistency bias enhances decision accuracy and confidence, particularly under noisy conditions.

Conclusions:

• The DCB method offers a robust tool for investigating decision-making mechanisms and validating computational models.
• Evidence integration is a key mechanism, modulated by a consistency bias that improves performance by increasing robustness to noise.
• This consistency bias acts as a form of pre-decision confirmation bias, optimizing decision outcomes in complex environments.