TAIGR: Towards Modeling Influencer Content on Social Media via Structured, Pragmatic Inference

TL;DR

Health influencer discourse often blends anecdotes and rhetoric, making claim-centric verification insufficient. TAIGR introduces a structured, three-stage framework—Takeaway extraction, Argumentation Structure, and Trust Inference—augmented with external PubMed evidence and modeled as a factor graph to assess takeaway trustworthiness. On a dataset of 195 TikTok health videos with expert annotations, TAIGR outperforms baselines by up to 9.7 macro-F1 points and scales to 1,430 videos, revealing that credibility is more strongly shaped by rhetorical structure than by popularity. The work advances interpretable, evidence-grounded analysis of health misinformation in influencer content and provides a pathway for scalable, nuanced discourse understanding in social media health domains.

Abstract

Health influencers play a growing role in shaping public beliefs, yet their content is often conveyed through conversational narratives and rhetorical strategies rather than explicit factual claims. As a result, claim-centric verification methods struggle to capture the pragmatic meaning of influencer discourse. In this paper, we propose TAIGR (Takeaway Argumentation Inference with Grounded References), a structured framework designed to analyze influencer discourse, which operates in three stages: (1) identifying the core influencer recommendation--takeaway; (2) constructing an argumentation graph that captures influencer justification for the takeaway; (3) performing factor graph-based probabilistic inference to validate the takeaway. We evaluate TAIGR on a content validation task over influencer video transcripts on health, showing that accurate validation requires modeling the discourse's pragmatic and argumentative structure rather than treating transcripts as flat collections of claims.

Figures (6)

• Figure 1: Treating influencer content as a series of claims fails to account for the pragmatic Takeaway conveyed to the user, and may lead to factually misleading results. We contend that un-"check-worthy" anecdotes may still be valuable to model, especially in their role of supporting an influencer's implicit takeaway.
• Figure 2: TAIGR overview: models influencer discourse in three stages: (1) Takeaway extraction, identifying the core recommendation; (2) Argumentation Structure, constructing a graph of supporting claims; and (3) Trust Inference, validating the recommendation using external evidence.
• Figure 3: Argumentation graph for transcript-1 (Table \ref{['tab:takeaway_sample_with_type']}), with node types and directed support edges.
• Figure 4: Rhetorical distributions of TikTok influencers vs. experts: expert content relies on premises and achieves higher credibility, while influencer content emphasizes anecdotes.
• Figure 5: Is viral content more credible? We plot the correlation matrix of video features, showing that engagement metrics are decoupled from credibility (near-0 correlation).