What is going on? An evidence-frame framework for analyzing online rumors about election integrity

Summary

Starbird, Prochaska, and Yamron argue that the dominant framing of online misinformation as “false facts” misses where most of the work of misleading rumors actually happens: in the interaction between (often accurate) evidence and politically-charged frames. Adapting Klein’s data-frame theory of sensemaking and integrating it with framing theory from Entman and Benford & Snow, they develop an evidence-frame framework and apply it to a mixed-method analysis of 479 tweets sampled from a 1.8M-tweet corpus on the 2022 Arizona midterm election. Twitter’s quote-tweet and comment architecture, they show, is a particularly legible site for observing how shared evidence gets framed, reframed, escalated, or countered in collective rumoring.

Key Contributions

• Theoretical: Adapts data-frame sensemaking theory to collective online rumoring, bridging individual cognition and collective discourse via framing theory.
• Methodological: A replicable coding scheme operationalizing rumors along four dimensions (relevance, evidence, interpretations, framing actions), tailored to quoted-tweet/comment pairs, with a ten-stratum sampling design.
• Empirical: A detailed case study of how accurate reporting about Arizona voting machine problems was transformed into “rigged election” narratives.
• Conceptual: Reframes the misinformation problem from veracity of claims to evidence–frame interplay, with implications for fact-checking, platform design, and how researchers operationalize misinformation.

Methods

Grounded, interpretive mixed-method analysis combining qualitative coding with descriptive quantitative analysis. From a real-time Twitter Streaming API collection covering Nov 8–9, 2022, the authors derived a focused 1.8M-tweet subset and sampled 479 tweets across ten strata (top-20, random-top-100, random-top-500 quoted tweets plus comments, and a 60-tweet random baseline). Three trained coders applied the four-dimension scheme to quoted tweets and their comments separately; inter-rater reliability was assessed with Krippendorff’s Alpha, and quantitative comparisons used chi-square tests and one-way ANOVA.

Findings

• ~70% of voting-related tweets concerned election administration, with election integrity dominating discourse.
• Highly quoted tweets disproportionately carried asserted evidence (82% vs. 45% in the random sample) — evidence attracts amplification and reframing.
• 94% of election-administration tweets contained a frame, but only ~39% framed it explicitly; high-influence accounts framed less explicitly.
• The “poor election integrity” meta-frame dominated (78%), with election fraud most common; “robust election integrity” framing was rare (12%).
• 17% of comments escalated frames (e.g., from election mess to fraud); 27% countered the quoted frame.
• Quoted tweets with no explicit frame routinely attracted comments that uniformly supplied a poor-election-integrity frame — a “call and response” rumoring pattern.
• Engagement metrics (retweet-to-comment ratios) differed significantly across framing actions (added/escalated vs. countered/aligned).
• A specific “poll worker” video shared by Bowyer, Kirk, and Johnson became a focal piece of largely accurate evidence repeatedly reframed into fraud narratives.

Connections

This paper extends a CSCW lineage on rumoring and collective sensemaking and connects directly to Prochaska2025-ef, which shares an author and likely overlaps in its treatment of election rumoring. Its argument that influential accounts mobilize partisan audiences to supply frames resonates with work on hyperpartisan media and participatory disinformation such as Marwick2025-ov and Donovan2025-ws, and with empirical accounts of election-related false narratives like Ventura2026-yc, Ventura2025-sw, and Graham2025-gp. The conceptual move away from veracity-as-everything also speaks to Budak2024-ef and Gonzalez-Bailon2024-rq, which similarly complicate simple “false facts” framings of the misinformation problem.