Incorporating Procedural Fairness in Flag Submissions on Social Media Platforms
Yunhee Shim, Shagun Jhaver · 2026 · ACM Transactions on Social Computing · doi:10.1145/3797820
Summary
This paper examines how the design choices embedded in social media flagging mechanisms shape users' perceptions of procedural fairness. Flagging — the ability for users to report posts or accounts that violate community guidelines — is a critical component of platform governance, yet the design space of these mechanisms has received limited empirical attention. The authors conducted a large-scale mixed-methods survey experiment with 2,936 US participants, each randomly assigned to one of 54 distinct flagging interface designs. The interfaces varied across four components: rule violation classification scheme (none, simple, detailed), posting guidelines (none, simple descriptions, detailed with example violations), an open-ended text box for elaborating objections (present/absent), and moderator type (human, bot, or unspecified). After flagging, participants rated procedural fairness along three dimensions: consistency (equal treatment regardless of flagger identity), transparency (visibility into how reviews are conducted), and voice (ability to fully express objections). The study also included open-ended questions inviting improvement suggestions, generating 1,741 qualitative responses analyzed inductively. The paper is grounded in procedural fairness theory and platform governance literature, and connects to prior HCI work on content moderation appeals and personal moderation tools. It contributes empirically informed design recommendations for fairer flagging systems that balance expressivity with cognitive burden.
Key findings
Of 12 hypotheses tested, only two were fully supported. Displaying posting guidelines — even simple ones — significantly increased transparency perceptions compared to showing none (F(2, 1952.81) = 11.36, p < .001), with detailed guidelines producing the strongest effect. Providing an open-ended text box significantly increased perceived voice (M = 5.67 with box vs. M = 4.94 without, t(2859.50) = −13.32, p < .001). Counterintuitively, classification scheme granularity and moderator type (human, bot, unknown) had no significant effect on any fairness dimension. The text box did increase cognitive burden, suggesting a usability trade-off. An interaction analysis found that when no text box was available, providing any classification scheme significantly enhanced voice perceptions compared to no scheme. Qualitative analysis surfaced five themes: users wanted qualified unbiased reviewers (11%); greater expressivity including multiple category selection, severity ratings, and post-highlighting (24%); timely outcome notifications and status tracking (29%); transparency about review procedures and aggregate statistics (24%); and protections against false and coordinated flagging (12%). Users also explicitly raised needs for improved accessibility and plain-language descriptions of the flagging process.
Relevance
This paper has direct relevance to digital accessibility and platform equity. The authors note that BIPOC, LGBTQ+, disabled, plus-sized, and sex worker communities disproportionately face content removal sanctions under the same guidelines — suggesting that biases in the flagging and review pipeline specifically harm marginalized groups. Improving flagging accessibility through plain language, diverse reporting channels (chat, email, phone), and simplified processes directly benefits users with cognitive and technological differences. The finding that posting guidelines improve transparency without increasing cognitive burden is immediately actionable for accessibility practitioners advising platform teams. For disability communities, false flagging of disability-related content — such as feeding tube discussions flagged as self-harm promotion — is a documented harm, and the design recommendations here, including false-flag safeguards and transparent review criteria, are directly relevant. The study also establishes a methodological framework for evaluating fairness in moderation affordances that can be extended to test accessibility-specific flagging scenarios.
Tags: content moderation · platform governance · procedural fairness · flagging · social media · user experience · accessibility · algorithmic moderation · marginalized communities · transparency