Kaplan, R. 1, 2 , King, J. 1 , Koster, R. 1 , Penny, W. 1 , Burgess, N. . 1 & Friston, K. 1
1 University College London
2 Universitat Pompeu Fabra
The hippocampus and mPFC are crucial for planning optimal spatial trajectories, yet the computations behind this type of decision-making are unclear. We examined the neural processes that enable humans to infer the shortest route between a starting point and target location. Subjects viewed mazes offering routes that varied in hierarchical depth and path length during fMRI scanning. We used Shannon entropy to quantify uncertainty about the shortest path length for a given maze - and the implicit computational complexity of selecting the shortest path. Entropy was calculated by estimating the distribution of choice probabilities using a softmax function of path length. The inverse temperature parameter of the softmax function was optimized using reaction times from each subject. We constructed predictors of fMRI responses in terms of the entropy (uncertainty) of the choice probability in each trial. We found vmPFC and hippocampal responses to confident (low entropy/precise) choices - and dACC/pre-SMA responses to uncertain (high entropy/imprecise) choices, consistent with evidence from value guided decision-making. Additionally, we found an interaction between hierarchical depth and entropy between rostral and caudal portions of the mPFC. Our findings illustrate how distributed responses in the hippocampus and mPFC contribute to optimal decision-making during spatial forward planning.