Value construction through sequential sampling explains serial dependencies in decision making

Area of Science:

• Decision Neuroscience
• Cognitive Psychology
• Neuroeconomics

Background:

• Decision-making often assumes stable item values, with inconsistencies attributed to random noise.
• Bounded Evidence Accumulation (BEA) models perceptual decisions using temporally uncorrelated noise.
• This assumption may not hold for value-based decisions where internal states fluctuate.

Purpose of the Study:

• To investigate if subjective item values change during a decision-making task.
• To develop and apply a novel algorithm (Reval) to detect and quantify value changes.
• To assess if dynamic values improve models of choice and response time.

Main Methods:

• Reanalysis of existing snack choice data using the Reval algorithm.
• Comparison of Reval-derived values against explicitly stated values.
• Modeling choice and response times with dynamic versus static value assumptions.
• Correlation of value changes with brain activity (BOLD signal) in the ventromedial prefrontal cortex.

Main Results:

• Subjective item values were found to change significantly within a short experimental session.
• Reval-derived dynamic values provided a better explanation of choice and response times than static values.
• Dynamic value changes also better explained neural activity in value-related brain regions.
• A modified BEA model incorporating non-independent evidence samples supported the revaluation concept.

Conclusions:

• Subjective values are not static but dynamically revalued during preference choices.
• Revaluation is a key factor in value-based decision-making, influencing choices and neural representations.
• Existing models of decision-making may need to incorporate dynamic value construction for greater accuracy.