Virtual fMRI brown bag: September 11

Time: 12:00-1:00pm

Place: Zoom

Please join us for a talk given by Seongmin A. Park, a project scientist working with Erie Boorman at UC-Davis. Recently published work by Dr. Park includes this bioRxiv preprint, "Novel inferences in a multidimensional social environment use a grid-like code".

How does the brain construct and navigate a cognitive map of abstract relationships to guide novel decision-making?

Abstract:

Recent findings suggest the hippocampal-entorhinal (HC-EC) system may serve a general mechanism for representing and navigating cognitive maps of non-spatial relationships between abstract concepts. A powerful advantage of the cognitive map is the ability to make structural inferences from limited experiences that can dramatically accelerate learning and guide novel decisions never faced before. However, it has not been clear yet how the brain integrates experiences from different time points to construct a cognitive map even in the absence of continuous sensory feedback and how such a neural representation of abstract cognitive maps affords structural inferences. Here, we developed a novel task to address these questions. During behavioral training participants learned the relationship between individuals in two independent social hierarchy dimensions through a series of comparisons between pairs at neighboring ranks in one dimension at a time. Without seeing the overall graph structure, participants could infer the latent social hierarchy structure through transitive inferences. During fMRI participants were asked to make a novel decision of selecting a better ‘business partner’ between two for a given individual. Unlike the behavioral training, the decision-value for the partner selection task is not determined by the ranks of two candidates in the social hierarchy; instead, it depends on the area drawn by two individuals over the abstract 2-dimensional (2-D) space. We find that discretely sampled abstract relationships between people are reconstructed into a unitary cognitive map in that the pattern similarity of the HC and EC activity represented the relationship between individuals in the 2D space. Moreover, we find that the human brain utilizes a hexadirectional grid-like code in the EC, medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC), temporoparietal junction (TPJ) and superior temporal sulcus (STS) for inferred direct vectors between individuals in the reconstructed 2-D social hierarchy space. This grid-like signal is consistent between sessions acquired more than a week apart. Moreover, the neural grid-like codes in EC explain neural decision value computations in TPJ and mPFC during decisions. Finally, during decisions the EC and vmPFC activity also encodes the relative decision value. These findings shed light on the representational architecture of abstract cognitive maps and how decision value is constructed from a multidimensional cognitive map and compared to guide a novel decision. Taken together our findings provide evidence that grid-like codes are extended to encode trajectories in abstract task spaces to guide novel decision making, suggesting a general mechanism underpinning flexible decision making and generalization.