Civic empowerment in the development and deployment of AI systems

Abstract

Our Critiquing and Rethinking Accountability, Fairness and Transparency interactive workshop at ACM’s Fairness, Accountability and Transparency in Machine Learning on March 5, 2021, provided workshop attendees with information on what constitutes meaningful civic empowerment in artificial intelligence (AI) systems’ development and deployment. The workshop offered a primer on participatory theory from the social sciences and had speakers from nonprofits discuss the role of tech-enabled community activism and consider political power dynamics in using tech to influence AI development and policy.

We used digital whiteboards and breakout sessions to have attendees reflect on what they heard and supply their own definitions of actors, roles, and empowerment, which was summarized by one attendee’s provocative question, "who would be participating if we weren’t having this intentional conversation about participation?” In terms of the ‘who’, individuals and groups directly impacted by algorithms should definitely be considered participants. After that, the attendees offered a plethora of potentially impacted peoples, which ranged from future generations, journalists, and the undocumented to non-human actors like environmental ecosystems. Often the reflection on who was implicated and who was harmed by AI was as important as an invitation to a predetermined or convenient stakeholder to participate. In terms of the ‘what’, we asked for roles and responsibilities of all the members involved in the development and deployment of an AI system. Many attendees described the role of individuals, groups, and organizations as providing feedback and inputs on the development process and its outcomes, on the participants’ part as well as ensuring transparency, accountability and fairness from the AI developers.

Defining and operationalizing civic empowerment proved the most difficult for workshop attendees. Definitions often centred around self-empowerment as personal satisfaction and community quality of life. Empowerment often was circular: a willingness to participate more or greater representation by marginalized communities. When attached to political power, empowerment varied from providing feedback to giving participants the power to halt an algorithm or change laws and algorithmic decisions.

We sought to co-develop an assessment questionnaire that will allow AI developers, government, and tech-enabled civil society organizations to gauge the level of civic empowerment in their AI system(s) (see Section 4). We categorized the contributions from attendees in terms of existential, epistemic, process-based and a few ready-to-go assessment questions. Our hope was that this set of questions could eventually augment AI certification, audit, and risk assessment tools. At this stage, instead of generating a simple list of best practices, our assessment questionnaire serves to “open Pandora’s box,” explicating the diversity of positions, confronting the challenge in synthesis and implementation, and moving towards meaningful and not merely performative empowerment.

Keywords: artificial intelligence, civic empowerment, public participation, AI systems