Frederik Aust

@frederikaust.com

1469 Followers 481 Following

Tenured researcher @ University of Cologne, formerly University of Amsterdam. STAR editor at Psychological Science. Bayesian statistics, mathematical models of cognition, computational reproducibility, R package development, and coffee. ☕️

Quote: Jérôme Lecoq, a neuroscientist at the Allen Institute, used Claude Science to build a multi-agent “computational review template” comprising about 20 custom skills geared towards writing long-form reviews. The sub-agents read through thousands of papers, pulling the central claim and the key quantitative finding, and storing them in an evidence state database. Then the pipeline constructs a narrative arc, writing the review section by section and delegating each to its own specialized sub-agent. Within each section, dedicated agents generate quantitative cross-study figures directly from the evidence database. A key component of the workflow, enabled by Claude Science, is the use of actor-critic pairs: one agent creates content while a separate reviewer agent evaluates it for accuracy and citation fidelity. Before Claude Science, it could take Lecoq’s team as many as two years to write such a review. He now has about 10 reviews, many more than 100 pages, with citations that were checked over by reviewer agents. The team is now working with domain experts to further refine the AI-based critic agents.
6-panel plot showing adoption of Bayesian analyses for 6 psychology journals from different fields. Histograms show historical data from 2004 to 2024. Predictions from five-parameter logistic growth curve model with forecasts till 2044 are overlaid. Sample size for journals vary from 2009 to 3686. Use of Bayesian analyses is increasing in all journals and projected to level of at varying levels between 25% and 75%.
2D-pl,ot with variable DrivAge on the x-axis and ClaimNb on the y-axis. Points show binned expectations for values of DrivAge. Three colors distinguis between training data, posterior predictions, and hold-out data. The curves roughly show a u-shaped curve. In the background there is a grey-filled density curve that slightly right-skewed and shows the distribution of the predictor variable plotted on the x-axis.