University of Edinburgh
Zoom link emailed to symposium registrants
This talk examines the interplay between experimental practice and modelling techniques of idealisation and abstraction in computational cognitive neuroscience. Scientists measure neural activity during specific behavioural tasks and produce models that interpret the activity as carrying out certain computations, offering explanations of the brain’s involvement in the cognitive performances. The computational models are abstract and idealised representations of neural activity, but they also depend on simplifications introduced early on in the experiment-to-model pipeline.
Following Bogen and Woodward (1988), phenomena and not raw data are taken as the target of explanatory models. However, I argue that neuro-phenomena should not be thought of as regularities existing in nature independently of experiment and statistical processing. Instead, data generated via experimentally-introduced simplifications are further simplified through statistical methods to produce the regularities targeted by the computational model – the ideal pattern. I argue that the simplifications introduced in experiment and data-processing serve a crucial role in allowing the modeller to build a mathematical representation of the neural activity that is simple enough to be interpretable.