Title: mcplR: multiple cue probability learning models in R

Author: Maarten Speekenbrink

Affiliation: University College London

Abstract: mcplR is a package to fit a variety of psychological learning models in
R. In a prototypical learning task, a participant is shown a stimulus
which varies on a number of dimensions (cues) and asked for a response
(e.g., a classification regarding the category the stimulus belongs to),
after which the participant receives feedback about the quality of the
response (e.g., whether the classification was correct). Over repeated
trials, the participant learns from the feedback and adapts his or her
responses to the task structure. The models in mcplR model this process
in terms of two sub-models: a learning model in which an internal
representation is formed about the stimulus-feedback relation, and a
response model in which the internal representation is used to produce a
response for a given stimulus. Widely used models such as the
Rescorla-Wagner learning model and the Generalized Context Model (GCM)
are implemented in mcplR. The package has an object-oriented design
(using the S4 system) allowing users to add their own learning and
response models and to flexibly combine these. In this presentation, I
will discuss the underlying structure of the package and how it relates
to the structure of psychological models of learning.