Can Large Language Models Transform Computational Social Science?
This work provides a roadmap for augmenting computational social science research by using LLMs as zero-shot data annotators and for bootstrapping creative generation tasks, though it is incremental in applying existing methods to new domains.
The paper investigates whether zero-shot large language models (LLMs) can reliably classify and explain social phenomena like persuasiveness and political ideology, finding that on taxonomic labeling tasks, LLMs achieve fair human agreement but do not outperform fine-tuned models, while on free-form coding tasks, their explanations often exceed crowdworker quality.
Large Language Models (LLMs) are capable of successfully performing many language processing tasks zero-shot (without training data). If zero-shot LLMs can also reliably classify and explain social phenomena like persuasiveness and political ideology, then LLMs could augment the Computational Social Science (CSS) pipeline in important ways. This work provides a road map for using LLMs as CSS tools. Towards this end, we contribute a set of prompting best practices and an extensive evaluation pipeline to measure the zero-shot performance of 13 language models on 25 representative English CSS benchmarks. On taxonomic labeling tasks (classification), LLMs fail to outperform the best fine-tuned models but still achieve fair levels of agreement with humans. On free-form coding tasks (generation), LLMs produce explanations that often exceed the quality of crowdworkers' gold references. We conclude that the performance of today's LLMs can augment the CSS research pipeline in two ways: (1) serving as zero-shot data annotators on human annotation teams, and (2) bootstrapping challenging creative generation tasks (e.g., explaining the underlying attributes of a text). In summary, LLMs are posed to meaningfully participate in social science analysis in partnership with humans.