Social Computing Compass Tool for Democratic Design Justice and Democratic Framework to Foster Self-Regulation

KER 3 is an easy-to-use and gamified resource directed towards software developers working in academia and industry. It is designed to encourage a deeper level of self-reflection about the impact of AI software design on civic and democratic participation.

The purpose and motivation behind the Social Computing Compass Tool are twofold:

• Enable software developers to make granular, actionable, culturally sensitive decisions about their coding practices, and help them incorporate these decisions into their work at an early phase (in line with KT4D Objective 1: Establish guidelines)
• Encourage the development of more integrated and truly human-centric innovation paradigms around the use of AI and big data in contexts related to democracy (in line with KT4D Objective 3: Create technology innovation aligned to democratic values)

The Social Computing Compass Tool is situated at the level of coding decisions, rather than in high-level ethical principles. Its goal is to advance awareness of cultural specificity, rather than focusing just on universal human values.

Machine Learning and AI-based predictions are possible because the world has regularities. However, culture often defies these regularities and instead is made of differences and variations which matter greatly for democracy. Our goal is thus to offer a tool that encourages developers to adopt a different stance to potential users, one that considers and respects cultural differences and specificities.

In other words, the goal of KER3 is for software developers to appreciate how their decisions regarding technical aspects and requirements are received differently depending on the cultural values and identities of the communities who adopt those systems.

Input from Use Cases

To respond to developers’ real needs and workflow, we are partly basing the design of our Social Computing Compass Tool on the findings of the first meeting for Use Case 4, which took place in October 2023 in Dublin. The workshop provided us with extremely important inputs, and it allowed us to test our preliminary assumptions and to reconsider our initial plan.

Several issues and ideas crucial emerged during the workshop, such as:

1. Communicating the need for ethics in software development to people who are not already receptive and knowledgeable of the issues at stake is extremely challenging. To minimize the risk of ‘preaching to choir,’ we concluded that the best targets for our tool are 3rd level instructors, and HR departments.
2. Many tools supporting ethics in AI do exist, some very useful, but they do not reach their intended audience. Indeed, of all the tools we presented during the workshop, participants didn’t recognise any of them.
3. Differently from what we imagined, checklists do not work, nor competitive games – though ‘red teaming’ was positively viewed by our participants. Instead, they suggested narrative and collaborative games, which are more nuanced and complex than gamified checklists.
4. Participants were quite comfortable considering and including intersectionality in their work (gender identity, racial identity, different professional and social roles of the users). By contrast, multiculturality was described as extremely difficult to understand and to embed in software development.

Beside these insightful remarks, during the UC workshop participants created two prototypes for an AI ethics tool which further provided directions and inspiration for KER3.

Findings & Next Steps

KER 3 is developed in connection with other WPs and Tasks of the KT4D project. More specifically, it builds upon the insights into the cultural impact of AI and Big Data developed in Module C of the toolkit (WP3); the knowledge on ethical practices in developing AI systems and technologies, such as Participatory ML and Algorithmic Accountability described in Module H of the toolkit (WP7); the reflections on critical digital literacy as defined in Module F of the toolkit (WP6).

KER3 will be submitted at the end of the project in M36. Currently, we are working in conjunction with the other WPs, consulting with developers outside of the project team, conducting research on digital narrative and interactive fiction, and surveying existing AI ethical tools and games.