Time-varying boundaries for diffusion models of decision making and response time

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Mathematical Psychology

Background:

• Diffusion models are standard for modeling two-choice decision making, typically assuming fixed evidence thresholds (boundaries).
• Theoretical statistics link decision and response time distributions to diffusion models with time-varying boundaries.

Purpose of the Study:

• To summarize theoretical results on time-varying boundaries in diffusion models.
• To develop and apply a computational method for inferring time-varying boundaries from empirical data.
• To explore the implications of time-varying boundaries for decision-making models.

Main Methods:

• Summarized theoretical statistical results relating decision/response time distributions to diffusion models with time-varying boundaries.
• Developed a novel computational method for inferring time-varying boundaries from empirical data.
• Applied the method to equate diffusion models with accumulator models and to fit individual-level perceptual decision data.

Main Results:

• Demonstrated how time-varying boundaries can make diffusion models equivalent to accumulator models.
• Showed that individual-level time-varying boundaries can best fit empirical data for perceptual tasks with ambiguous evidence.
• Provided a framework for analyzing decision-making with flexible evidence thresholds.

Conclusions:

• Time-varying boundaries offer a more flexible and potentially more accurate approach to diffusion modeling in decision making.
• The developed computational method enables the inference of these dynamic boundaries from empirical data.
• This work opens new avenues for understanding the temporal dynamics of evidence accumulation in decision processes.