The Drift Diffusion Model (DDM) is a mathematical model of two-alternative forced-choice decision-making. It assumes that evidence accumulates over time as a noisy process (Wiener diffusion) drifting toward one of two decision boundaries. The model simultaneously accounts for both choice accuracy and response time distributions, making it a powerful tool for understanding the cognitive mechanisms underlying decisions.
Key components
The DDM is characterized by several parameters, each with a psychological interpretation:
- Drift rate: the average rate of evidence accumulation, reflecting task difficulty or stimulus quality.
- Boundary separation: the amount of evidence required before committing to a decision, reflecting the speed-accuracy trade-off.
- Starting point: the initial bias toward one alternative, reflecting prior expectations.
- Non-decision time: time consumed by processes other than evidence accumulation, such as stimulus encoding and motor response.
Connections to my research
My interest in the DDM stems from my work in mathematical psychology and neuroeconomics:
-
Decision-making under risk. During my master’s studies, I worked with Prof. Yung-Fong Hsu on behavioral and fMRI experiments investigating the role of emotion when people face dual-process induced cognitive conflicts under risky conditions. The DDM provides a natural framework for modeling the speed and accuracy of such decisions, and I was involved in developing estimation methods for utility function parameters that connect to drift rate parameterizations.
-
Financial decision-making. In Prof. Yu-Ping Chen’s lab, I analyzed behavioral and fMRI data from a stock market bubble experiment. Participants’ decision processes during competitive trading can be modeled using sequential sampling frameworks, where the drift rate may vary with perceived market conditions and social information from other players.
-
Connections to adaptive methods. The DDM shares deep mathematical connections with stochastic processes that underlie adaptive psychophysical methods. The evidence accumulation process in the DDM is a continuous-time analogue of the sequential probability ratio test (SPRT), and insights from one domain often inform the other.