Supporting Informed Self-Disclosure: Design Recommendations for Presenting AI-Estimates of Privacy Risks to Users

TL;DR

The paper tackles how to present quantified privacy risks from online self-disclosures using population risk estimates (PREs). It employs speculative design and comic-boarding to evaluate five PRE concepts across four threat scenarios with $N=44$ Reddit users, revealing that PREs can raise risk awareness and motivate cautious disclosure, but can also induce anxiety and misinterpretation without scaffolding. The authors derive four design principles to improve PREs: provide explanations of exploitation, disclose calculation methods, offer de-risking guidance that preserves communicative intent, and present information in intuitive language. They discuss limitations such as recruitment context and the need for broader validation across platforms, and highlight the practical impact of designing user-facing privacy risk tools that support informed decision-making without unnecessary self-censorship.

Abstract

People candidly discuss sensitive topics online under the perceived safety of anonymity; yet, for many, this perceived safety is tenuous, as miscalibrated risk perceptions can lead to over-disclosure. Recent advances in Natural Language Processing (NLP) afford an unprecedented opportunity to present users with quantified disclosure-based re-identification risk (i.e., "population risk estimates", PREs). How can PREs be presented to users in a way that promotes informed decision-making, mitigating risk without encouraging unnecessary self-censorship? Using design fictions and comic-boarding, we story-boarded five design concepts for presenting PREs to users and evaluated them through an online survey with N = 44 Reddit users. We found participants had detailed conceptions of how PREs may impact risk awareness and motivation, but envisioned needing additional context and support to effectively interpret and act on risks. We distill our findings into four key design recommendations for how best to present users with quantified privacy risks to support informed disclosure decision-making.

Figures (24)

• Figure 1: This image depicts each of the narrative vignette scenarios we explored in the comic-boards, from interpersonal threats to the threat of ambiguous others.
• Figure 2: Image depicting our various population risk estimate designs, explored in the comic-boards.
• Figure 3: We use an adapted version of the comic-boarding method moraveji2007comicboardingkuo2023understandingjin2022exploringwu2025design to understand Reddit users perspectives on various population risk estimate designs (this image depicts the risk meter design inspired by Ur et. al's data-driven password meter from usable security & privacy literature) ur2017design. Users were randomly presented 3 out of 5 different population risk estimate designs, matched to 3 out of 4 different risk scenarios in order to elicit specific feedback and reactions around the design and deployment of this risk awareness technology.
• Figure 4: A stacked bar graph depicting global ranking preferences of all PRE designs. We used the Plackett-Luce method maystre2015fastjin2022exploringwu2025design to merge partial rankings into a global preferred order of 5 PRE designs. PRE designs are ranked in order of preference from left to right. A higher bar indicates a more preferred PRE design. The dashed red allows for comparison across all 5 PRE designs. Design #5 (Risk by disclosure) saw the highest popularity in rankings, closely followed by Design #2 (Re-identifiability meter). Design #1 (the Raw Anonymity score) was the least preferred in comparison to the other PRE designs.
• Figure 6: Scenario 1 x Design 1