Hauke Licht

Hauke Licht, Rupak Sarkar, Patrick Y. Wu et al.

Many constructs that characterize language, like its complexity or emotionality, have a naturally continuous semantic structure; a public speech is not just "simple" or "complex," but exists on a continuum between extremes. Although large language models (LLMs) are an attractive tool for measuring scalar constructs, their idiosyncratic treatment of numerical outputs raises questions of how to best apply them. We address these questions with a comprehensive evaluation of LLM-based approaches to scalar construct measurement in social science. Using multiple datasets sourced from the political science literature, we evaluate four approaches: unweighted direct pointwise scoring, aggregation of pairwise comparisons, token-probability-weighted pointwise scoring, and finetuning. Our study finds that pairwise comparisons made by LLMs produce better measurements than simply prompting the LLM to directly output the scores, which suffers from bunching around arbitrary numbers. However, taking the weighted mean over the token probability of scores further improves the measurements over the two previous approaches. Finally, finetuning smaller models with as few as 1,000 training pairs can match or exceed the performance of prompted LLMs.

Hauke Licht

Research increasingly leverages audio-visual materials to analyze emotions in political communication. Multimodal large language models (mLLMs) promise to enable such analyses through in-context learning. However, we lack systematic evidence on whether these models can reliably measure emotions in real-world political settings. This paper evaluates leading mLLMs for video-based emotional arousal measurement using two complementary human-labeled video datasets: recordings created under laboratory conditions and real-world parliamentary debates. I find a critical lab-vs-field performance gap. In video created under laboratory conditions, mLLMs arousal scores approach human-level reliability with little to no demographic bias. However, in parliamentary debate recordings, all examined models' arousal scores correlate at best moderately with average human ratings and exhibit systematic bias by speaker gender and age. Neither relying on leading closed-source mLLMs nor computational noise mitigation strategies change this finding. Further, mLLMs underperform even in sentiment analysis when using video recordings instead of text transcripts of the same speeches. These findings reveal important limitations of current mLLMs for real-world political video analysis and establish a rigorous evaluation framework for tracking future developments.