Assessing Generalization for Subpopulation Representative Modeling via In-Context Learning

Gabriel Simmons, Vladislav Savinov


Abstract
This study evaluates the ability of Large Language Model (LLM)-based Subpopulation Representative Models (SRMs) to generalize from empirical data, utilizing in-context learning with data from the 2016 and 2020 American National Election Studies. We explore generalization across response variables and demographic subgroups. While conditioning with empirical data improves performance on the whole, the benefit of in-context learning varies considerably across demographics, sometimes hurting performance for one demographic while helping performance for others. The inequitable benefits of in-context learning for SRM present a challenge for practitioners implementing SRMs, and for decision-makers who might come to rely on them. Our work highlights a need for fine-grained benchmarks captured from diverse subpopulations that test not only fidelity but generalization.
Anthology ID:
2024.personalize-1.3
Volume:
Proceedings of the 1st Workshop on Personalization of Generative AI Systems (PERSONALIZE 2024)
Month:
March
Year:
2024
Address:
St. Julians, Malta
Editors:
Ameet Deshpande, EunJeong Hwang, Vishvak Murahari, Joon Sung Park, Diyi Yang, Ashish Sabharwal, Karthik Narasimhan, Ashwin Kalyan
Venues:
PERSONALIZE | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
18–35
Language:
URL:
https://aclanthology.org/2024.personalize-1.3
DOI:
Bibkey:
Cite (ACL):
Gabriel Simmons and Vladislav Savinov. 2024. Assessing Generalization for Subpopulation Representative Modeling via In-Context Learning. In Proceedings of the 1st Workshop on Personalization of Generative AI Systems (PERSONALIZE 2024), pages 18–35, St. Julians, Malta. Association for Computational Linguistics.
Cite (Informal):
Assessing Generalization for Subpopulation Representative Modeling via In-Context Learning (Simmons & Savinov, PERSONALIZE-WS 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.personalize-1.3.pdf