Drift-Diffusion Model (Claude Skill)

Guides Claude through evidence-accumulation modeling of two-choice response-time data — choosing a DDM variant and fitting tool, cleaning RTs, checking convergence, and validating that the recovered parameters mean anything.

   
Type Claude Skill
Supplier Awesome Cognitive and Neuroscience Skills (community OSS, MIT)
Availability GA — one of ~40 research skills in the collection (last upstream push 2026-07-13)
Pricing Free / OSS (MIT)
Capabilities Read/Write — methodology guidance; Claude writes and runs the fitting code locally
Verified works · 2026-08-03
Security cleared · 2026-08-03 — MIT confirmed, repo transferred to NeuroAIHub org (old path still redirects), no advisories

How to install

  • Claude Code — plugin marketplace (installs all skills in the collection): 
    /plugin marketplace add HaoxuanLiTHUAI/awesome_cognitive_and_neuroscience_skills
/plugin install awesome-cognitive-and-neuroscience-skills@awesome-cognitive-and-neuroscience-skills

    Restart Claude Code afterwards. The skills are description-activated — there is no slash command; ask a reaction-time modeling question and Claude loads the skill.

  • Claude Code — single-skill alternative. This skill declares a required dependency on the collection’s research-literacy skill, so copy both: 
    git clone https://github.com/HaoxuanLiTHUAI/awesome_cognitive_and_neuroscience_skills
cp -r awesome_cognitive_and_neuroscience_skills/skills/drift-diffusion-model ~/.claude/skills/
cp -r awesome_cognitive_and_neuroscience_skills/skills/research-literacy ~/.claude/skills/

    (Project-scoped alternative: copy into .claude/skills/ instead.)

  • Underlying software — the skill is method guidance, not a fitting engine: install whichever tool you select. PyDDM (pip install pyddm) and HDDM (pip install hddm, hierarchical Bayesian, needs a working PyMC/Theano-class backend) are Python; fast-dm is a compiled C program from the Voss group; EZ-diffusion needs no software at all — it is a closed-form calculation.

What it does

Runs an eight-stage protocol:

  1. Research planning — state the question, justify DDM over simpler RT analyses, and declare expected parameter effects before fitting.
  2. Data preparation — drop RTs below ~200 ms and above a 3,000–5,000 ms cutoff.
  3. Model specification — pick the variant (classic vs. full DDM) and decide which of drift rate, boundary separation, starting point and non-decision time are free versus fixed.
  4. Estimation — fit with the chosen tool.
  5. Convergence — verify the optimizer succeeds from multiple starting points (or, for HDDM, that chains have mixed).
  6. Model comparison — BIC for maximum-likelihood fits, DIC/WAIC for Bayesian fits.
  7. Posterior predictive checks — compare simulated against observed RT quantiles, not just means.
  8. Parameter recovery — refit simulated data with known parameters to confirm the pipeline can recover them.

Tool selection guidance: EZ-diffusion for a fast three-parameter summary, fast-dm or PyDDM for maximum-likelihood fits of the full model, HDDM when trial counts per subject are low enough that hierarchical shrinkage helps.

Primary use cases: decision-making and speed–accuracy tradeoff studies, clinical/individual-difference RT analyses, model-based fMRI/EEG regressors.

Notes

AI-generated content — verify before use. The collection’s README states that all skills “were generated by AI from academic papers and other publicly available sources” and that “the content has not been individually verified by human domain experts.” This skill’s front-matter carries review_status: ai-generated. It cites Ratcliff 1978, Ratcliff & McKoon 2008, Wagenmakers et al. 2007, Voss et al. 2013 and Wiecki et al. 2013 — check the RT cutoffs and comparison criteria against those sources; no package version numbers are pinned upstream.

Related catalogued skills in the same collection: Signal Detection Analysis for accuracy-only sensitivity/bias decomposition.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.