Tim participated in a panel and lightning talk session at The National Lottery Communities Fund AI Funders Conference: “How do we build a better society in the age of AI?”.
Notes from Tim’s lightning talk are below.
My goal in this talk is two fold.
Firstly, I want to start with a big URRR: That is to demonstrate a range of practical approaches to engage with the voices of affected communities when funders and their grantees explore how they Use, Refuse, Respond to, or Re-shape, Artificial Intelligence.
Secondly, reflecting on some of the themes coming up in public engagement activities, I want to highlight some of the more intangible, yet transformative, impacts that participatory engagement can bring.
To preview both points:
- Effective public engagement on data and AI calls for models of engagement that are informed, deliberative and inclusive.
And;
- When we center lived experience and participation in our approach to AI, we have the potential not only to build or adopt better and more appropriate AI, but we also contribute to a centering of human relationships, individual and collective agency, and solidarity: providing an important counterbalance to some of the centralising, atomising and inhuman logics of large-scale AI systems.
I’m going to start by looking at an example from India, presented at PAIRS, the Participatory AI Research and Practice Symposium we hosted alongside the India AI Impact summit in February - where a team from Wadwani, an Indian AI for Good funder, reflected on their journey from building tools in the lab, to inviting users in the field to be more involved in helping them to understand what use would really look like.
Their Health Vaani platform they have been building is AI chatbot, designed to give community-based frontline health workers, particularly ‘Anganwadi Workers’, access to healthcare information.
In the work the team presented in February, they spoke of encountering questions about differential levels of use of their tools: raising fears that the tool as initially prototyped might not work equally well for different users.
That meant they had to move from funding not just a technology development team, but also a user-research team, who went out to the communities supposed to benefit from the tools being built, and sought to better understand how a chatbot answering healthcare questions fits into the lived reality of providing support at the village level.
The team described how they quickly switched from using live-demos on phones, to working with paper-prototypes and print-outs of application screens, to have better dialogue with healthcare workers - and to understand the ‘cognitive load’ put onto workers by having to shape their questions into the formats expected by prompt-driven mobile apps.
This kind of participatory-design work generated insights both for application design - but also for adoption: pointing towards the kind of interventions that would be needed alongside the introduction of technology if it was to have its intended impact.
In particular, through ethnographic work and co-design sessions, they found that for many users, when the application could not provide a clear answer straight away, this quickly led to disengagement, rather than attempts to craft more refined questions in the way designers had expected. This helps check early extrapolations from a successful pilot, to assuming impact at scale - to encourage deeper design of a whole AI programme, not just a platform.
Cases like this illustrate an important baseline we should apply. When AI technology is being built and deployed for social good, it is imperative to have participation and social research skills within the development and deployment teams. Engaging with lived experience early in, and throughout development and deployment, can highlight design decisions that need to be revisited, identify risks of uncritical tech introduction, or better set expectations of impact.
But, we should be striving to go beyond this baseline. Delgado and colleagues put together this really useful matrix looking at the Participatory Turn in AI Development, asking about the goal of engagement, the scope and the form
Essentially they challenge work to not only look at the user-interface or features of an application, but also to move towards public engagement over whether and why a system should be built in the first place.
For me, one of the experiences that best exemplifies this point came when we ran the People’s Panel on AI, as a deliberative review of the UK AI Safety Summit.
Picture the moment.
-
A stage of AI luminaries, facing the stage lights in the British Museum Auditorium, talking about AI opportunity and AI safety: how we would have to adapt to manage risks of this emerging tech.
-
A question comes from the dark of the back row of the audience… “Aren’t we approaching this the wrong way around? Shouldn’t we start from talking about what is good for society - rather than starting from technology?”.
-
It was a question that was so at odds with the dominant narrative of technology in search of problems. And it’s a question that remains relevant in many of the proposals put to funders today. Are we prioritising the problems we use AI to address? Have we thought about other ways of addressing those problems? And are we clear on the hypothesis we’re testing by introducing data and AI as a solution?
I want to tell you a bit about how we got to that question being asked.
A few days before we had gathered in a community centre around the corner from the British Museum to welcome a group of 11 randomly selected citizens from across England, many with no prior experience talking about AI, who, over the following three days, had the chance to engage with AI hands on, to observe fringe and main events from the AI Summit, and to meet one-to-one and in panel discussions with AI experts. Supported by expert facilitators, the group identified questions they wanted to ask, and recommendations they wanted to make.
It was a powerful demonstration that when people are giving space for critical learning about AI, they can quickly develop nuanced views on how it should be deployed, and how governance should respond to it.
It also demonstrated to us that, when we bring together lived experience with a chance to learn and reflect on AI, many people grow in their confidence to express views on technology - and want to go on to use that confidence more.
That led us to think about the pathways of participation we could offer to members of the People’s Panel, and people who have been involved in other public engagement activities on AI - and over 2024/25 we helped to coordinate an ongoing Public Advisory Group for the Public Voices in AI Programme. This invited a panel of six members of the public to build on their past engagement and to meet regularly - mostly online - to provide challenge and advice for an £850k research programme. Their input shaped survey designs, grantmaking calls and decisions, and the presentation of findings - and they were part of sharing project findings with Parliamentarians at the end of the project - as well as speaking about public voice on AI on local radio as shown here.
One of the central messages from the group was that the awareness and empowerment they were gaining with respect to AI should be shared. In essence, we don’t want to see better designed AI still ‘done to’ other people: everyone should feel that AI is something that they have agency over, and that, as the 2025 UN Human Development Report puts it, enhances their agency.
One way that has come across is in the Let’s Talk AI campaign, launched last month through a co-design process with groups across the country - creating a series of scrollable webtoons that provide a balanced view of AI in everyday life - and offer a foundation for opening up local conversations.
Investing in local dialogue about AI matters: but so does linking this to organisational, national or even global decision making. I want to share two final case studies: one delivered, one in development - and show this potential.
In October last year, working with the Department for Education, we launched a toolkit for hosting distributed dialogues on Generative AI in Education. The simple lesson plan used a set of cards and worksheets to introduce potential educational applications of AI, and some of the issues to consider - and then asked classes to debate statements suggested both by the DfE, and by their peers, providing feedback through an online platform whether they agreed or disagreed. These results, accompanied by qualitative feedback from classes, fed into a video and report that directly influenced new national product safety expectations for AI in education: strengthening prohibitions on AI tools in schools exploiting emotional manipulation to motivate engagement, role-playing as humans, or straying into providing emotional and social advice.
But, through an independent evaluation, we’ve also heard how discussions at the classroom level have informed teachers’ decisions. Students spoke about the importance of the student-teacher relationship in ways that caused some professionals to rethink their use of AI lesson planning or marking. We’ve also heard how, being part of a national conversation really anchored classroom conversations - making them more likely to take place - and raising an interest in comparing attitudes towards AI across groups.
We’re now building on learning from this distributed dialogue model in planning for the Citizens Track on AI - an ambitious global initiative to embed participation in the new UN Global Dialogue on AI Governance, taking place for the first time in July this year. In short, we want to see communities across the globe starting from their lived experience, gaining critical literacy around AI, and identifying actions both for the local community level - and to feed into global policy making: building solidarity, confidence and agency to shape our shared AI future.
And at Connected by Data we’re working on that for a simple reason. The People’s Panel I spoke about earlier, the first of their 7 recommendations after their week in London was for inclusive global governance of AI - recognising that shaping AI for social good needs us to not just create different apps, but also to shape the markets, incentives and landscapes in which AI and its effects are produced.
So, I’ve probably run over time, but to very briefly recap, making the most of lived experience in shaping AI demands not just user-research or surveys, but deeper forms of power-sharing participation, that build public confidence to make choices about, and to shape, AI futures.
And what it makes possible is not just (or not even directly) better functioning tech, but digitally confident communities - with both individual and collective agency, driving better decisions about where we put resources to both create, and regulate, a socially desirable digital world.