Shifting Attention to Accuracy Can Reduce Misinformation Online

David Rand

In brief

Rand presents evidence that misinformation sharing is driven primarily by inattention to accuracy rather than disbelief or ideological preference. Simple "accuracy nudges"—priming attention to accuracy via brief prompts—substantially reduce sharing of false content in both lab experiments and real-world Twitter deployments, and the effect generalizes across 16 countries and diverse content types.

Key findings

Inattention is the dominant mechanism. In controlled experiments, people accurately judge headlines' veracity when asked directly, yet share false headlines at high rates when sharing intentions are measured separately—a dissociation driven by limited attention to accuracy during rapid social media scrolling. Decomposition of false-sharing behavior shows ~50% attributable to inattention, ~33% to confusion, and only ~17% to deliberate "post-truth" behavior.
Accuracy priming works. Asking people to rate the accuracy of a single unrelated headline at the study's outset—a minimal cognitive intervention—triples the discernment between true and false headlines in subsequent sharing tasks. This effect replicates across 20 experiments (2017–2020) with 26,000 US participants.
Field evidence confirms lab findings. In a naturalistic Twitter field experiment with 5,000+ known misinformation sharers, an accuracy-nudge direct message significantly increased the quality of shared links (measured by fact-checker trustworthiness ratings). The treatment effect was specific to retweets (where attention is limited) rather than original tweets, and proved more effective among high-follower users.
Global generalizability. A 34,000-person cross-cultural study across 16 countries found accuracy nudges increase sharing quality by ~20% on average, with larger effects in countries where baseline discernment was poorest.
Crowdsourcing complements nudges. Accuracy responses from lay crowds (15 raters per headline) correlate as well with professional fact-checkers as fact-checkers correlate with each other, suggesting the collective wisdom can inform platform algorithms.

How the intervention works

The limited attention model formalizes why nudges succeed: even when people prefer not to share inaccurate content, accuracy is excluded from decision-making due to attentional constraints during rapid scrolling. Priming the concept of accuracy ensures it enters the utility calculation for subsequent sharing decisions. Drift-diffusion modeling of reaction-time data supports this account, showing the treatment changes what people think about (prioritizing accuracy) rather than simply inducing slower deliberation.

Implications for platforms

Rand proposes accuracy nudges as a scalable, decentralized complement to machine-learning classifiers and professional fact-checking:

No centralized arbiter of truth required (unlike labels or downranking).
Distributed accountability relies on users' own accuracy preferences, which are remarkably aligned across political lines.
Minimal burden: platforms can vary prompts to prevent habituation (e.g., rotating different accuracy-related questions).
Deployable off-platform: nonprofits and individuals can drive ads containing nudges to high-risk sharers.

TikTok and Twitter have implemented variations. Jigsaw (Google) and Facebook are exploring adoption.

Roozenbeek & van der Linden (2019) — The "Bad News" Game — inoculation through shared-world game mechanics; Rand's attention-based nudges are complementary to this prebunking approach.
Allen et al. (2021) — Scaling fact-checking using the wisdom of crowds — crowdsourced fact-checking can achieve professional accuracy; supports Rand's findings that lay crowds can rate accuracy as reliably as experts.
Misinformation topic page — broader literature on spread, psychology, and interventions.

Notes

This talk synthesizes Rand's lab work and real-world deployments. The cross-country generalization is particularly valuable; most prior misinformation intervention research focuses on the US. The emphasis on limited attention as the primary lever distinguishes this from moral/political reframing approaches. One open question: long-term habituation—how quickly do users stop noticing accuracy prompts if repeated—though Rand discusses rotating intervention designs as a mitigation. The field experiment's focus on domain-level rather than headline-level quality (fact-checkers rate outlets' overall trustworthiness, not individual pieces) is pragmatic but loses some signal.

How it connects

Discussed in GUIDE by

Understanding and Reducing the Spread of Misinformation Online Rand · 2020