Deep storytelling: Collective sensemaking and layers of meaning in U.s. elections

Prochaska, S., Vera, J., Tan, D. L., Yamron, B., Venuto, S., Kejriwal, A., Chu, S., & Starbird, K. (2025). Deep storytelling: Collective sensemaking and layers of meaning in U.s. elections. Proceedings of the ACM on Human-Computer Interaction, 9, 1–43. https://doi.org/10.1145/3757576

View paper

Summary

This paper examines how false and misleading claims about U.S. elections in 2020 and 2022 were sustained on Twitter through collective sensemaking, arguing that the contextual storytelling around individual rumors is what gives them meaning. The authors extend Hochschild’s concept of the “deep story” to online, participatory settings, contending that the how of telling — style, allusion, and cueing — is as constitutive of meaning as the explicit content. Through large-scale qualitative coding of election-related tweets, they show that influencers, political elites, and audiences collaboratively perform an evolving deep story of voter fraud, and that this dynamic is largely invisible when analysts examine individual posts in isolation.

Key Contributions

• Theoretical: Extends “deep stories” by arguing that storytelling style and allusiveness — not just content — constitute meaning, and integrates the concept with collective sensemaking and participatory disinformation frameworks.
• Empirical: Provides a longitudinal, cross-cycle comparison of election rumoring, documenting a shift from explanation-heavy discourse in 2020 to allusion- and cue-heavy discourse in 2022 as the deep story matured.
• Methodological: Introduces a dual-codebook design pairing tweet-level and incident-level qualitative coding with quantitative temporal visualizations, offering a template for studying context-dependent misinformation that resists single-post measurement.

Methods

Interpretive, grounded analysis of Twitter discourse from the 2020 and 2022 U.S. elections, drawing on Election Integrity Partnership datasets (over a billion tweets, with millions sorted into “incidents”). The team purposively sampled the top five incidents per election year (ten total), refining 2020 incidents into event-specific subsets comparable to 2022. Each incident was sampled at two levels — a per-tweet sample (~100 tweets) and a larger per-incident sample (~350 tweets) — mixing top-retweeted, random, and quote tweets. Two iterative codebooks were developed: one applied at the tweet level (e.g., “Explains story specifics,” “Presents artifact/event without describing meaning”) and one at the incident level. Coding used a multi-coder process with outside arbitration and consensus, achieving Cohen’s kappa of .78 and .73 on key codes, combined with thematic analysis and case-study interpretation.

Findings

• Across 2020 incidents, “Explains story specifics” dominated, reflecting substantial work constructing the mechanics of alleged fraud.
• Across conservative-leaning 2022 incidents, “Presents artifact/event without describing meaning” was more prevalent, indicating reliance on audiences’ prior knowledge of the established deep story.
• In Maricopa County 2022, influencers shared events and footage with minimal explicit interpretation while audiences supplied fraud framings consistent with the 2020 narrative.
• In the USPS/DeJoy case, 2020 discourse built an intersubjective narrative of intentional sabotage, whereas 2022 discourse invoked it as common knowledge to mobilize calls to “Fire DeJoy.”
• Storytelling cues (e.g., “Box 3,” “Sharpie,” “Dominion,” “DeJoy”) functioned as allusive triggers that activated the deep story without requiring explicit claims.
• Deep-story dynamics were often invisible at the single-tweet level but legible at the incident level, motivating the dual-codebook approach.

Connections

This work builds directly on prior research into elite–influencer–audience collaboration in election denial discourse, most closely Starbird2025-jj, with which it shares authorship and its participatory disinformation framing. Its emphasis on narrative context over discrete claims resonates with broader debates about whether misinformation’s impact is overstated, engaged elsewhere in the literature on partisan information ecosystems.