Many of the previous chapters have been necessarily technical in nature, describing specific techniques for estimating parameters and comparing models on their account of empirical data. Having estimated parameters of a model, or identified one model as being more likely than another, what have we learned about how people think and behave?

The last part of this book is dedicated to describing how models are typically used in psychology. This chapter aims to give some broad coverage of the use of models in psychology. We discuss the inferences psychologists can (and do) draw from modeling, and how we use models as tools to aid understanding. We finish the chapter with a discussion of reproducibility, and how our modeling tools can be shared so as to help others understand our models and how they apply to our data.

Broad Overview of the Steps in Modeling

To begin our discussion, it will help to step back and consider how the preceding chapters all fit together. Figure 12.1 shows a flowchart representing the relationships between models and data that were implicitly or explicitly covered in the preceding chapters. The boxes on the left represent psychological science as it is often conducted, without the aid of modeling. People generate data in experiments, and the data that people produce depend on the exact situation that we put them in (i.e., the experimental method). The data are only informative to the extent that they can be related to some theory. What Figure 12.1 does not show are the (usually vague) verbal theories that researchers hold when conducting experiments, and the predictions that might be loosely associated with those vague verbal theories.

The boxes on the right of the figure show the mo del pathway. In Chapter 2 we talked about how our theoretical ideas can be turned into a computational model (for a similar demonstration for the domain of verbal working memory, see Lewandowsky and Farrell, 2011). Note that we distinguish between what a researcher might consider “core” assumptions, and more auxiliary assumptions that are not an essential part of the model but which need to be made in order to derive quantitative predictions from the model.