1)    New paper is out! Very excited to share our findings in monkeys and humans playing Battleship. Joint work with @[email protected], @[email protected], and Daniel Salzman. Thread below… https://www.nature.com/articles/s41598-023-27662-9.pdf
2)    How do we learn and reason in the face of the messy complexity of the real world? Historically, research on learning regards simple conditional associations. More recent work has focused on learning latent features, but often only a small number at a time based on rewards from single choices.
3)    In the real world, however, latent features are often learned from gathering information over sequences of decisions. For example, searching on the internet, hunting down the best deals in the store, and (perhaps especially!) scientific research all involve gathering information over multiple decisions.
4)    We wanted a task that would contain all these features and that both humans and animals could perform. We landed on Battleship!
5)    In our task, humans and monkeys learn to reveal shapes during trials by turning over tiles on a grid. If they found a part of the shape (a hit), they received a reward; if they did not (a miss), then they advanced to the next choice. Trials ended once shapes were revealed.
6)    On each trial, there was a single shape, drawn from a small number, but the participants had to learn the shapes over multiple trials.
7)    Both species were excellent at the task, though (unsurprisingly) the humans learned the shapes much faster. So how did participants learn to reveal the shapes?
8)    In a shock to us all, participants foraged for information to learn the shapes! They tended to jump to different parts of the board when their recent searches dropped below the average information gathered about the shapes. This is a version of the well-known Marginal Value Theorem in optimal foraging theory. A similar explanation based on rewards did not describe their behavior as well.
9)    In addition, in a genuine fall-out-of-my-chair-in-the-lab moment, we discovered that how well their behavior could be described as following the marginal value rule predicted how quickly they learned the shapes. In short, better foragers learned shapes faster!
10) This work has been cultivated for nearly 7 years now. Many thanks to my co-authors for their help and to the NIH for my K99 for funding this work. I’m eager to begin building on these findings--I’m on the job market, so feel free to reach out to learn more!