Adaptable history biases in human perceptual decisions

Area of Science:

• Cognitive psychology
• Decision science
• Behavioral economics

Background:

• Past experience significantly influences choices under perceptual uncertainty.
• However, this experience can introduce decision-making biases, negatively impacting performance.
• Choice history biases, like switching after failure, are observed even in simple detection tasks.

Purpose of the Study:

• To investigate human adaptability to choice history biases in decision-making.
• To determine if individuals can overcome irrational biases when trial statistics change.
• To compare human adaptation with principles of reinforcement learning.

Main Methods:

• Standard two-alternative detection task with manipulated choice history.
• Logistic regression modeling to capture choice history biases.
• Experimental manipulation of trial order statistics to test adaptability.

Main Results:

• Human choices were consistently influenced by past success or failure, leading to poorer performance.
• These choice history biases were similar across different countries and well-modeled by logistic regression.
• Adaptability to changing trial statistics was asymmetric: stronger for confirmatory, weaker for contradictory evidence.

Conclusions:

• Humans exhibit persistent choice history biases that impact decision-making accuracy.
• While some adaptation occurs, existing biases are difficult to eradicate, particularly those aligning with default patterns.
• Adaptation mechanisms are more sensitive to statistics that confirm existing biases than those that contradict them.