Cas Ludwig, our collaborator from the School of Psychological Science at the University of Bristol spent time the last six months in Berlin working with out lab and Ralf Engbert’s group at the University of Potsdam. He has been a long-standing visitor of Berlin since 2018when he began working with both groups.
His research while in Berlin has focused on two projects:
(i) Developing a computational model of fixation control in extended sequences of saccades (with Potsdam);
(ii) A “perspectives-style” paper on ‘grounding cognitive model parameters’.
The first project was really the main project, but working on it led him to think more broadly about computational modelling practices, which then turned into the second project. The Potsdam group provided a wealth of knowledge and experience with likelihood-free methods of parameter estimation and with complex models of time-ordered data. Their expertise was particularly important for the main project he was working on.
His highlights from the visits were coming out of the Covid-19 pandemic, since working had at home had felt isolating, given the change in working practices (not organised in big lab groups and many people have continued to work from home). He was happy to be part of two large research groups and to feel part of a vibrant intellectual community again. Often, he was not working on specific, collaborative projects, but just having interesting chats with people and discuss research was invaluable. It reminded him of the fun side of being an academic and the importance of what he calls “intellectual play”. We enjoyed having Cas around the lab and look forward to working with him in the future!
To conclude his research visit he gave a talk to our team on the opinion paper he had been working on titled “Grounding cognitive model parameters”. A short summary can be found below.
Computational models of cognition have parameters that govern their behaviour. It is common practice to estimate values for these parameters by fitting the model to empirically observed data. These parameters are then interpreted as latent factors in the mind of the subject(s) who generated the data. It is generally beyond the scope of the model (and the modeller) to identify where those parameter values came from and how they develop over time: why and how does the subject adopt the estimated values for these parameters? This is an important psychological question in its own right. I will identify four major themes that should be addressed by a research programme for grounding cognitive model parameters. These themes will be illustrated with the broad class of evidence accumulation models of decision making.