An interactive tutorial for developing likelihood functions for ACT-R models
Christopher R. Fisher · 2022
Building upon the work in Fisher et al. [2022], we developed a set of tutorial notebooks explaining howto develop likelihood functions for models based upon the ACT-R cognitive architecture [Andersonet al., 2004]. The tutorial notebooks combine text, equations, plots, and annotated computer code tocreate an integrated learning experience. Each tutorial notebook is interactive, allowing the readeradjust model parameters with sliders, and observe the effects on model predictions and the likelihoodfunction. The introductory tutorial explains the benefit of using likelihood functions to evaluatemodels, such as the ability to use Bayesian parameter estimation, and avoiding potential problemswith averaging artifacts. The remainder of the tutorial is organized into two broad sections: (1)background information covering software, mathematical concepts, and statistical concepts, and(2) a series of models ranging from simple to moderately complex. Each model tutorial guidesreaders through the development of the likelihood function, and demonstrates how to estimateparameters with Bayesian parameter estimation. A repo containing the tutorials can be foundat https://github.com/itsdfish/ACTRTutorials.jl. Installation instructions are enclosed within theREADME file.