Cognitive control is an important ability that allows us to exert more self-control (e.g., refrain from
drinking and driving), and has been shown to correlate positively with various markers of success in
modern society (e.g., better exam scores, greater job success). Consequently, many psychologists
wonder whether cognitive control can be trained. Thus far, despite several attempts, training
studies have been disappointingly unsuccessful. Here, we argue that earlier studies focused on
mental concepts that were too ill-defined. To remedy this, we will develop a model-based cognitive
training paradigm. Specifically, we start from a well-studied cognitive process model of learning and
decision making, which contains two crucial parameters: learning rate (how quickly one can learn
novel information) and discount rate (how steeply one discounts future benefits). Both parameters
can be thought of as a window, into the past and into the future, respectively. By manipulating
these “windows” (learning rate and discount rate) using well-studied cognitive tasks, we will test
whether training induces transfer of learning rate and discount rate across task paradigms (e.g.,
learning about reward or task difficulty), and from past to future. Using this approach, we will also
investigate which similarities in neural ensembles (i.e., voxel patterns) can predict such transfer,
thereby unraveling the neural substrates of these potentially generalisable control processes.