Presentation Details
| RAGs to Riches(?): Using AI Research Assistants to Generate Ethograms Peter E Midford1, Anne B Clark2. 1SRI International (retired), Richmond, VA, USA.2Binghamton University, Binghamton, NY, USA |
Abstract
Ethograms underpin many behavior studies ranging from simple observation to evolutionary or conservation questions, but assembling ethograms from existing literature can be a challenge. Many species have few or no published ethograms. AI’s search & synthesize capabilities should excel in finding defined behaviors within publications and constructing ethograms de novo. We tested the abilities of large language models (LLMs) and Retrieval Assisted Generation systems (RAGs) to construct and compare ethograms for species of ground-dwelling spiders and passerine birds. Both quantitative and qualitative results show real tradeoffs: LLMs use more sources (e.g., books), ignore copyrights and often misreport or hide sources; RAGs draw on recent, OA papers, report sources, but miss key older literature and anything outside of journals. Considering the broader, but untrustworthy sourcing of LLMs and the limited sources used by RAGs, we find neither suitable to the task of generating complete ethograms.
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.