Research /

Digital Juries


As concerns have grown regarding harmful content spread on social media, platform mechanisms for content moderation have become increasingly significant. However, many exist- ing platform governance structures lack formal processes for democratic participation by users of the platform. Drawing inspiration from constitutional jury trials in many legal systems, this paper proposes digital juries as a civics-oriented approach for adjudicating content moderation cases.


The paper was accepted to the 2020 ACM Conference on Human Factors in Computing Systems, and is available on the ACM digital library as well as here [no paywall].

About this project

Building on existing theoretical models of jury decision-making, we outline a 5-stage model characterizing the space of design considerations in a digital jury process. We implement two examples of jury designs involving blind-voting and deliberation. From users who participate in our jury implementations, we gather informed judgments of the democratic legitimacy of a jury process for content moderation. We find that digital juries are perceived as more procedurally just than existing common platform moderation practices, but also find disagreement over whether jury decisions should be enforced or used as recommendations.

Download the paper from the ACM digital library or the Digital Juries site [no paywall].

Run-through of the deliberation interface

The experiment takes users through 3 versions of a “jury” interface, including a blind voting scenario and discussion with 6 actual other participants to decide how to adjudicate a case.

The designs for the jury conditions are displayed here, with a “no action” control. The blind voting scenario allows users to drag a slider to set the toxicity score, while the discussion room creates a temporary chat room to discuss the case.

This 5-stage jury model displays the various parameters needed to translate a real-world jury system into an online environment. Factors in this model were varied to create an online experience that could be empirically tested.

As part of the final exploration of the evolution of harmful content in communities, I created an agent-based model simulating social clustering of toxic opinions, modeled and simulated in NetLogo.