"Public(s)-in-the-Loop": Facilitating Deliberation of Algorithmic Decisions in Contentious Public Policy Domains

Multiple stakeholder groups must be involved, not just experts: The paper emphasizes treating "publics as plural political entities," meaning that algorithmic decisions affecting public policy should involve diverse communities and perspectives rather than relying solely on technical experts or a single unified "public."

Deliberation, not just consultation, is essential for legitimacy: The framework calls for "collective decision-making through deliberation," suggesting that meaningful public input requires structured dialogue and reasoning about algorithmic choices in contentious domains like recidivism prediction, rather than simply collecting feedback after decisions are made.

Who gets included in decision-making processes matters as much as how: The paper highlights "the construction of publics" as a central feature, meaning that researchers and designers must actively and thoughtfully determine which communities should participate in algorithmic governance—a choice that itself shapes outcomes and legitimacy.

Current approaches to AI governance in public policy are insufficient: By proposing this framework, the authors implicitly critique existing stakeholder participation methods, suggesting that contentious policy domains like criminal justice algorithms require more robust public deliberation mechanisms than are typically implemented.

This is an emerging research agenda for the HCI community: The paper positions public deliberation in algorithmic decision-making as an underdeveloped area requiring further investigation, calling on human-computer interaction researchers to develop practical methods and tools to support these deliberative processes.