ESRC Digital Good Network

News

Digital Good Network leads new project on including public voices in Responsible AI

Copy Of Copy Of Evaluating The Digital Good Blog Post 1 1

The Digital Good Network team has recently been successful in securing a £850,000 grant from UKRI, as part of its £36m investment in responsible and trustworthy AI, for a project called Public Voices in AI.

Digital Good Network Director, Helen Kennedy, and Associate Director, Ros Williams, will lead the project, with our Policy and Public Engagement Lead, Reema Patel, and Senior Associate Researcher, Susan Oman, leading aspects of the work. Project collaborators include the Ada Lovelace Institute, the Alan Turing Institute and University College London

The project runs from April 2024 to March 2025. An online launch event will take place in early June. (Sign up to our newsletter here, to ensure you don’t miss details about this event).

Public Voices in AI aims to ensure that public voices are front and centre in artificial intelligence research, development and policy. A central aspect of responsible AI is ensuring that it takes account of public hopes, concerns and experiences. As concern about the societal impacts of AI grows and pressure for its effective regulation mounts, understanding and anticipating societal needs and values can inform responsible AI developments and deployments. Yet public voice is frequently missing from conversations about AI, an absence which inhibits progress in RAI. Addressing this gap is essential to enable ‘good AI’ – that is, AI RD&P which maximises benefits, prevent harms and ensures that AI works for everyone.

Project lead, Helen Kennedy, said: 

“Public voices need to inform AI research, development and policy much more than they currently do. This project represents a commitment from UKRI and RAI UK to ensuring that happens. It brings together some of the best public voice thinkers and practitioners in the UK, and we’re excited to work with them to realise the project’s aims. 

However, different groups benefit from and are affected by AI differently, and their hopes, concerns and experiences also vary. Because of structural inequities, some groups are more negatively impacted by AI deployments than others – for example, in welfare systems, at borders, in policing. Some groups have more resources and access to power to shape AI technologies than others. There is also a participation gap between those with the social capital to participate in shaping AI and those without.”

Reema Patel said:

“I’m delighted to work with such an interdisciplinary team to extend the work I’ve done on frameworks for participation to the emerging field of AI technology. Enabling people to increase their agency and power over AI technologies seems a critical endeavour if they are to truly work for people and society.”

Susan Oman said:

“Discussions about ‘responsible AI’ are happening at pace in powerful places, and this is a good thing. We’re delighted that the digital good network and The University of Sheffield is leading a project to ensure these discussions consider the experiences and perspectives of less powerful people from less powerful places, too.”

Public Voices in AI will therefore put those most impacted and underrepresented front and centre. Susan Oman will lead an evidence review which goes beyond traditional methods which tend to validate certain knowledges and invalidate others, legitimising knowledge which has already been legitimated, and excluding knowledge from certain actors. Working with postdoctoral researcher Sara Cannizzaro, this work will include criteria that centre equity and inclusion. Reema Patel will also lead work to ensure that equity and inclusion are embedded across all project work, as well as synthesising this work into an AI participation framework, building on her prior work on participatory data stewardship and the outcomes of this project. 

Another way Public Voices in AI will centre underrepresented groups is by distributing up to £195,000 of its funding to support participatory projects with negatively affected or underrepresented groups. The Public Voices in AI Fund will be launched on 31st May 2024, with a closing date for applications of 20th June 2024.

All of these aspects of the project align with the Digital Good Network’s equity societal challenge, through which we ask: how can we make sure digital technologies contribute towards a more equitable society? 

Public Voices in AI will share all project outputs on the RAI UK website, including an evidence database, an AI participation framework, original empirical research and methodological resources to enable various stakeholders to include public voices in AI research, development and policy. There will also be a small Flexible Fund, supporting qualitative, participatory research with people from negatively affected and underrepresented groups. Through these activities, the project’s aim is that members of the public, especially underrepresented groups, have more power in and inform AI research, development and policy.

To ensure you don’t miss the latest information about this project, including the launch of The Public Voices in AI Fund, sign up to the Digital Good Network newsletter here.