Russell Poldrack

It is commonly argued that understanding (particularly regarded in terms of intelligible explanation) and prediction are distinct and possibly competing objectives in scientific theorizing and modeling. Here I leverage Wilkenfeld’s notion of “understanding as compression” along with the link between prediction and compression from algorithmic information theory to establish that prediction and understanding are instead intimately linked. I clarify the definition of understanding as compression to address unclear aspects of Wilkenfeld’s original definition, and I propose that the oft-discussed conflict may result in part from confusion between understanding proper and vehicles of understanding (such as statistical or computational models). I exemplify the unification of prediction and understanding through compression via a discussion of the role of large neural network models as vehicles for understanding brain function in neuroscience.

It is commonly argued that understanding (particularly regarded in terms of intelligible explanation) and prediction are distinct and possibly competing objectives in scientific theorizing and modeling. Here I leverage Wilkenfeld’s notion of “understanding as compression” along with the link between prediction and compression from algorithmic information theory to establish that prediction and understanding are instead intimately linked. I clarify the definition of understanding as compression to address unclear aspects of Wilkenfeld’s original definition, and I propose that the oft-discussed conflict may result in part from confusion between understanding proper and vehicles of understanding (such as statistical or computational models). I exemplify the unification of prediction and understanding through compression via a discussion of the role of large neural network models as vehicles for understanding brain function in neuroscience.

We discuss recent progress in the development of cognitive ontologies and summarize three challenges in the coordinated development and application of these resources. Challenge 1 is to adopt a standardized definition for cognitive processes. We describe three possibilities and recommend one that is consistent with the standard view in cognitive and biomedical sciences. Challenge 2 is harmonization. Gaps and conflicts in representation must be resolved so that these resources can be combined for mark-up and interpretation of multi-modal data. Finally, Challenge 3 is to test the utility of these resources for large-scale annotation of data, search and query, and knowledge discovery and integration. As term definitions are tested and revised, harmonization should enable coordinated updates across ontologies. However, the true test of these definitions will be in their community-wide adoption which will test whether they support valid inferences about psychological and neuroscientific data.

"Well-publicized research in psychology tells us that over half of our attempts to change habitual behavior fail within one year. Even without reading the research, most of us will intuitively sense the truth in this, as we have all tried and failed to rid ourselves of one bad habit or another. The human story of habits and the difficulty of change has been told in many books - most of which will make only a quick reference to dopamine or the "lizard brain" before moving on to practical tips and tricks for behavior change. In contrast, Stuck: The Neuroscience of Why Changing Our Behavior is So Hard will tell the brain's story about why behavior is so hard to change. Russell Poldrack offers an in-depth, yet entirely accessible, guide to the neuroscientific research on habits and habit change. Part I introduces the "anatomy of a habit," starting with the argument that the resilience of our habits stems largely from a mismatch between the environment in which our brains evolved and the one in which we now live, and continuing on to introduce current work on fear and anxiety, motivation, and cognitive control that bears on habit formation. Part II focuses on what neuroscience can tell us about breaking habits, introducing evidence-based strategies that give us the best possible chance to break cycles of bad behavior. Throughout the book, Poldrack offers a clear-eyed view of what neuroscience can tell us about habit change, and what it cannot - and importantly, how we know what we know"--

The concept of “representation” is used broadly and uncontroversially throughout neuroscience, in contrast to its highly controversial status within the philosophy of mind and cognitive science. In this paper I first discuss the way that the term is used within neuroscience, in particular describing the strategies by which representations are characterized empirically. I then relate the concept of representation within neuroscience to one that has developed within the field of machine learning. I argue that the recent success of artificial neural networks on certain tasks such as visual object recognition reflects the degree to which those systems exhibit inherent inductive biases that reflect the structure of the physical world. I further argue that any system that is going to behave intelligently in the world must contain representations that reflect the structure of the world; otherwise, the system must perform unconstrained function approximation which is destined to fail due to the curse of dimensionality, in which the number of possible states of the world grows exponentially with the number of dimensions in the space of possible inputs. An analysis of these concepts in light of philosophical debates regarding the ontological status of representations suggests that the representations identified within both biological and artificial neural networks qualify as legitimate representations in the philosophical sense.