Two-Step Task
Sequential decision task probing model-based vs model-free reinforcement learning.
Frequently Asked Questions
Have another question? Let us know.
What does it assess?
It provides a window into model-based vs model-free tendencies — the balance between goal-directed and habitual control. With 100 trials, the estimate is only indicative.
What is the difference between model-based and model-free control?
Model-free control repeats whatever paid off last time; model-based control uses knowledge of the task's transition structure to plan. The two-step task (Daw and colleagues, 2011) was designed to separate them behaviorally.
How does a model-based choice show up?
As an interaction with transition type: after a rare transition, model-based agents adjust their first-stage choice more, because they replay the structure rather than just the reward.
Why has this task become a standard?
It offers a clean behavioral readout of planning versus habit, which has made it a workhorse of computational psychiatry research.
Show Details
About the Two-Step Task
Computational Psychiatry
The Two-Step Task, developed by Daw and colleagues, is a landmark paradigm in computational psychiatry. It separates model-based (goal-directed) from model-free (habitual) learning by having choices lead to rewards through different state transitions.
How It Works
In Stage 1, you choose between two options. Each leads to Stage 2 with different probabilities (common vs rare transitions). The key measure is how stay probabilities depend on the previous reward and transition type: the reward×transition interaction provides a window into model-based vs model-free tendencies. With 100 trials the estimate is only indicative.
Clinical Applications
Reduced model-based control is observed in addiction, OCD, and depression. The task helps researchers understand how psychiatric conditions affect the balance between flexible goal-directed behavior and automatic habitual responses.
References
- Daw, N. D., Gershman, S. J., Seymour, B., Dayan, P., & Dolan, R. J. (2011). Model-based influences on humans' choices and striatal prediction errors. Neuron, 69(6), 1204-1215. 10.1016/j.neuron.2011.02.027
- Sochat, V. V., Eisenberg, I. W., Enkavi, A. Z., Li, J., Bissett, P. G., & Poldrack, R. A. (2016). The Experiment Factory: Standardizing behavioral experiments. Frontiers in Psychology, 7, 610. 10.3389/fpsyg.2016.00610
- Daw, N. D., Niv, Y., & Dayan, P. (2005). Uncertainty-based competition between prefrontal and dorsolateral striatal systems for behavioral control. Nature Neuroscience, 8(12), 1704-1711. 10.1038/nn1560
Related Experiments
Wisconsin Card Sorting Test (WCST)
FeaturedAssesses cognitive flexibility and abstract thinking through rule-switching tasks.
Iowa Gambling Task (IGT)
FeaturedSimulates decision-making under risk to study emotional and cognitive influences.
Tower of Hanoi
FeaturedPractice planning and move optimization in the classic three-peg puzzle; your move count is compared against the optimal solution.
Learn More
APM Test - Actions Per Minute
FeaturedMeasure your clicking speed and visual reaction time. Click numbers from 50 to 1 as fast as possible!
Free IQ Test
FeaturedTest your intelligence quotient with this free IQ test covering pattern recognition, logical reasoning, and spatial skills.
Cognitive Processing and Performance Mindset Scale
FeaturedAssess cognitive processing and performance mindset across standard dimensions.
About This Experiment
This experiment is an educational adaptation for demonstration purposes. It is not a clinical diagnostic tool. Results reflect your performance in this specific session and should not be interpreted as an assessment of your cognitive abilities.
Data Privacy
Your experiment data is processed and stored locally in your browser. No data is uploaded to any server unless you explicitly export it.
