NSF Convergence Accelerator Track F: Misinformation Judgments with Public Legitimacy

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

There have been increasingly frequent calls for communication platforms to take action to reduce the impact of misinformation, either by marking articles and posts as misleading or by reducing their distribution. But there is little agreement on what process should be used to determine which articles and posts should receive such enforcement actions. Platforms are reluctant to be fully transparent about their policies and procedures, in part because those procedures depend on human judgments whose public legitimacy is not established. What is needed is a clear conceptual foundation that allows platforms to describe policies in terms of a "subjective ground truth" and be publicly accountable for how well their procedures enact their policies. The project will develop two services. The first is a "golden set" service that, for a sample of items each month, convenes a jury to render judgments. A jury will consider evidence that has been assembled about an item and deliberate with each other. Each juror will then report out an individual judgment about whether the item is misinforming in a way that deserves enforcement action. Those individual judgments provide a partial view of the hypothetical subjective ground truth, which is defined as the distribution of judgments that would be made by all potential jurors if they all participated in a similar jury process. Platforms can then express policies transparently in terms of the actions they intend to take as a function of the subjective ground truth. The golden sets will provide a way to evaluate, after the fact, how well a platform's procedures implement its policies, creating accountability for the outcome of the procedures without requiring the platforms to make details of those procedures public. The second service is a forecasting API that can be used as part of a platform's decision procedures. It will provide, for any content item, a forecast of the subjective ground truth. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.