I’m going to try something new with writing week notes at the start of the week. Working part-time means finding Friday capacity for week-notes always feels tricky; but looking back at last week gives a better chance for reflection.
Power to Shape AI
Last week started with some outreach for the Power to Shape: CSO Voices on AI online engagement I’ve been supporting Clever Together Futureproof to run, along with the Data Tank. It’s a two-week online dialogue exploring how civil society organisations can have a stronger voice in governance of AI both globally and locally.
The dialogue kicked off fully on Friday, and is open for two weeks. There is still time to join. It’s running via an anonymous idea-sourcing platform. Leda and Lydia at Clever Together Futureproof explain the rationale for this in the intro note: seeking to reduce a focus on organisations and individuals, and emphasise ideas.
The bit I really like about this dialogue is that it’s emerged as a ripple from the governance panel at the Participatory AI Research and Practice Symposium, and the Tech Policy Press Op-Ed we worked on together after that.
PAIRS Planning
Last week also brought a couple of steps forward to planning the next Participatory AI Research and Practice Symposium: catching up with a volunteer for the next programme committee, working to schedule a meeting to get new programme committee established, and a fortunate chance to catch up with Astha Kapoor of Aapti Institute at some Digital Public Infrastructure Drinks in London and think about the way we can structure hosting for PAIRS in the future.
With it looking likely the next AI Action Summit in India will be in February next year, we’re going to be turning to concrete plans for a hybrid online and offline PAIRS, and seeking to fundraise much earlier this time around.
Where next for Public Voices in AI?
Last Monday also brought a catch-up with the Public Voices in AI team on ideas for future activity.
PVAI, for which we ran a public advisory panel, was a rapid year-long project, producing a wealth of insight into public attitudes on AI, and supporting a range of groups to build up their capacity to advocate for public perspectives. In thinking about what future phases could include, I explored a four ideas:
-
PAIRSx UK Symposium: Organise a UK edition of the Participatory AI Research & Practice Symposium to profile key findings from the PVAI program and support ongoing field-building.
-
Public Engagement Infrastructure - Peoples Panel Pool: Establish a standing public advisory panel to provide diverse public input on RAI UK projects and other AI programmes. This would aims to engage 20+ people as “Public Advisors” and to develop models to scale and sustain access to public input on AI.
-
Rapid Response Research Messaging & Active Outreach: Bringing evidence on public attitudes and voices about AI into mainstream media and policy narratives, and to the awareness of AI developers. E.g. capacity to place media coverage, having conference stands for public voice at events like AI UK, and facilitating engaging encounters between publics and the research community.
-
Public Education Campaign: Building on the recommendations of the Public Advisory Panel to develop a neutral, research-backed public education campaign on AI in collaboration with public participants.
Those have gone into the mix with others that the wider PVAI team are thinking about for a future proposal: but I’m also thinking about if and how some might be taken forward in other ways.
Community Tech and AI Commons
On Tuesday I joined a workshop organised by Rachel Coldicutt on Community Tech and AI, run as a series of discussions on different aspects of AI at the community level. And on Friday I hosted a small workshop in Stroud (mostly in an independent community role) to explore a potential local response to the New Commons Challenge on developing data commons to support local decision making (with generative AI).
Along with preparation work for the Perspectives on AI course I’m running at Hawkwood in a few weeks, I’ve been thinking a lot about how locally grounded approaches to experimentation with artificial intelligence (particularly using offline models, and community level data) offer pathways to more intentional governance: taking things ‘out of the cloud’ and into material space where they are more legible and amenable to discussion and decision-making.
Our discussions on Friday were super-productive: highlighting the potential for work on grassroots governance of health data - and linking really nicely with some of the discussions I’d be in the data before with National Voices at our “Aligning AI innovation in healthcare with patient need and the future NHS: roundtable”.
Demand-driven AI healthcare innovation
On Thursday, along with National Voices, we were at the Wellcome Trust offices for a roundtable discussion on AI in healthcare. We had a great group around the table, and some really in-depth discussions. We’ll be working with National Voices to write-up a short brief from the work soon, but for now, the below expansion of my outline notes (using qwen2.5-7b-instruct-1m in LMStudio with prompt to summarise, edited by hand) provides a flavour:
The NHS faces significant challenges. As expectations around healthcare evolve, there is a need to renegotiate the social contract surrounding health services. The pace of technological change, particularly in areas like AI, is rapid but relatively slow in its actual implementation within healthcare settings. This complexity necessitates frameworks that ensure responsible AI use, with accountability extending beyond ethical considerations to include listening to affected communities.
There are places where AI can bring real benefits from supporting diagnostics, to improving administration: but different uses require different forms of governance. In many cases, when clinicians go to the market with a specific problem, existing AI solutions are not available.
Building on existing work, such as Foundations of Fairness and other research on public attitudes to healthcare data sharing, offers a valuable starting point for integrating patient voices into the innovation ecosystem. Engaging patients throughout the process—from initial research through procurement and audit—can prevent waste and promote better outcomes. Recognising that industry has already started canceling AI projects due to poor returns, it is crucial to avoid wasting billions in NHS resources on unfeasible initiatives. To achieve this, transparency about data use and digital engagement is essential, alongside clear mandates for public input embedded within key frameworks.
The NHS needs an engagement ecosystem that supports access and support to engage patient groups in co-production. Leadership should focus on how public engagement can feed into shaping markets for, development of, and purchasing of AI. By co-designing challenges, products, and strategies with patients, the NHS can create trusted solutions that prevent waste and deliver better health outcomes.
Other activities
Last week also included:
- A chat with Sarah Spencer of the SAFE AI project about guidance resources under development on participatory approaches to AI for the humanitarian sector. We touched on Reema Patel’s recent and incredibly rich Framework and Self Assessment Workbook for Including Public Voices in AI and to Delgado et. al’s great framework in The Participatory Turn in AI Design that invites consideration of both depth of participation in decision making (consult; include; collaborate; own), and dimensions of involvement, (goals; scopes; forms).
- An interview for a new public engagement Thinks are running for the Department for Education on use of student data in training AI.
- Working with our Fellow’s Claire and Judith on a proposal for the Digital Good Network’s latest call, focussed on supporting school-level decision making about digital technology governance.
Outside of work, I also made it along to a Stroud District Green Party meeting where we were celebrating both recent success at the County Council, and remembering the life of John Marjoram, the first Green councillor in the country, and catalyst of both much local Green political action, and Stroud community action. A reminder of the power of persistence and building alternatives step-by-step, door-by-door, and conversation-by-conversation.
Reading
I’ve been reading Queer Data by Kevin Guyan which is a fantastic text on many levels. I picked it up to feel better informed in supporting action to respond to attacks on trans identity in proposed amendments to the Data bill, but the concept of queering data collection, as opposed to simply collecting data on queer identity, was also invaluable in a Create Gloucestershire board and team sprint last week as we were thinking about our Diversity, Equity, Inclusion and Belonging policies and practices: and reflecting on the need to avoid being constrained in our thinking by Arts Council or other demographic data collection categories.