Presentation Details
| From Anecdote to Evidence: Studying Rare Behavior with AI Shir Bar. Massachusetts Institute of Technology, Cambridge, MA, USA |
Abstract
Many behaviors central to ecology and evolution are brief, rare, and easy to miss. Using two case studies in larval fish, I argue that artificial intelligence (AI) can advance rare-behavior research in two complementary ways. One is targeted measurement: making hard-to-observe behaviors of interest measurable. In larval sea bream, AI detection of prey-capture strikes in video from large mesocosms enabled estimates of strike rates outside the laboratory, revealing lower rates than lab studies had suggested and variation with temperature but not with other environmental variables. The other is exploratory search: helping researchers find unusual events before they have been formalized as behavioral categories. In larval zebrafish, I used AI to identify unusual kinematic events in pose-derived trajectories and guide expert review, even when such events made up only 0.02% of the data. Together, these examples show that AI can expand the empirical study of rare behavior not simply through automation, but through interdisciplinary research that adapts computational tools to the biological system and integrates expert knowledge into how rare events are defined, identified, and evaluated.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
Content Locked. Log into a registered attendee account to access this presentation.