(1/7)First paper of the PhD out today! What’s it about? Most of the places we traverse and actions we take do not directly result in reward (as a PhD student, I understand this problem all too well…).

(7/7) Huge thanks to all co-authors and mentors, especially Josh Berke and @nathanieldaw for all the time given to training and discussion, and a special shoutout to my original mentor in the Berke Lab, @mohebial for being the best scientific role model I could ask for in a lab.

(6/7) By combining these credit-assignment strategies, the brain is able to integrate new information about the changing world, in order to generate reward predictions that adaptively guide behavior.

(5/7) One relied on direct experience, progressively propagating value over sequentially traversed places. The other was able to use knowledge of the maze’s structure to infer upcoming reward, even from places not directly followed by reward in the past.

(4/7) We then found that dopamine in the ventral striatum scales with predictions of future reward from each place in the maze. Leveraging this relationship, we identified two algorithms responsible for assigning credit over space and generating this place value signal.

(3/7) In a large part of my PhD work, I developed a novel decision-making task that emulates the formidable credit-assignment challenges posed by the natural world: unstable rewards, changing paths, and a multitude of possible states and actions to choose from.

(2/7) When we finally obtain our desired outcome, our brain needs to decide which states and actions deserve credit for getting us to this point. I wanted to better understand how brains accomplish this computationally challenging task.