Learning to allocate limited time to decisions with different expected outcomes

Area of Science:

• Cognitive psychology
• Decision science
• Computational neuroscience

Background:

• Human decision-making involves allocating limited time to gather information.
• Accuracy and speed trade-offs are crucial in decision tasks.
• Participants' decision thresholds influence overall outcomes.

Purpose of the Study:

• Investigate how humans allocate time for decisions with varying properties.
• Examine the learning and adjustment of decision thresholds.
• Compare computational models of decision threshold adjustment.

Main Methods:

• Two behavioral experiments involving information accumulation and decision-making.
• Manipulation of trial difficulty and reward structures.
• Development and Bayesian comparison of 10 computational models.

Main Results:

• Participants often failed to adopt optimal, simple strategies for decision threshold adjustment.
• A reinforcement learning model with time-varying parameters best explained participants' behavior.
• In the absence of cues, participants attempted to infer trial difficulty rather than use a time-decreasing threshold.

Conclusions:

• Human decision-making under time pressure is suboptimal.
• Reinforcement learning mechanisms play a role in adjusting decision thresholds.
• Participants' strategies deviate from optimal solutions, particularly in complex or uncertain environments.