Connected by data

Rewiring the State for public voice

Tue, 11 Aug 2026 00:00:00 +0000

The Cabinet Statement on Rewiring the State was released at the end of July, with a strong emphasis on devolution. It’s striking how it takes the phrase “Rewiring the State” in a very different direction to the technology-focused themes of the similarly named Science, Innovation and Technology Committee’s report on delivering digital government, the Global Government Forum’s report on government transformation, or Darren Jones’ speech on delivering a new digital state.

(One could even say the Statement neglects the importance of technology, with the only slight mention being “a strengthened focus on timely and accurate place-based data from the Office for National Statistics and other bodies that will support outcomes-based decision making in places”. For a thoughtful piece on how data and digital services can be deployed in devolved and distributed systems, I recommend James Plunkett and Richard Pope’s work on (NHS) Operating Patterns.)

The part that I was particularly interested in was this:

We agreed to replace the current model of consultation by default with new and modern forms of public participation. This includes considering where performative statutory duties to consult can be revoked and how we can more actively bring the public into the policy-making process instead. This recognises that local leaders are better equipped to link public opinion and civil society to policy making.

I am really pleased to see this commitment to actively engage the public in participatory policy-making processes. After all, ensuring communities have a powerful voice in the decisions (about data and AI) that affect them is our mission at Connected by Data. It’s a particularly welcome shift because of the increasing use of AI by both those responding to consultations and those processing those responses. It is easy to see current consultation processes devolving into thousands of AI-generated responses being summarised by AI tools in ways that may miss nuance, overlook expertise, stereotype submissions, and flatten disagreement. More importantly, I know from working with disenfranchised communities that consultations where responses are “read” by AI feel uncaring and diminish trust. We can and should do better.

However, I also know that public participation can be just as performative as statutory consultations; I am concerned about deliverability across public bodies including local government; and I’m worried about how public voice will be combined with specialist expertise from civil society and academia. Let’s get into it.

Why engage?

It’s vital to take a step back and think about what we want to achieve through engagement, consultation, and participatory processes. To me, there are three reasons why we should want to see public bodies engaging with affected communities around the decisions they are making, whether these are about policy in DfE, planning in local government, or the operations of GP surgeries:

To inform planning and decision making by highlighting evidence, perspectives, and values that should be considered by decision makers.
To improve democratic legitimacy and trust by ensuring that public bodies are and are perceived to be responsive and accountable to the people they serve.
To build connection and power within communities affected by proposed changes, to reduce divisiveness and so that they can effectively advocate for their interests.

I’ll note here that some people see consultation as an unnecessary burden – something that blocks progress more than aids it. They would argue that trust is built through delivery. They might go so far as to say that the third of what I consider goals – building connection and power within communities – is an anti-goal because it enables and encourages small groups of energised people (NIMBYs) to effectively block things (most importantly, building homes, data centres and other infrastructure) that we need more of.

I would argue that trust based solely on delivery is vulnerable, because it is in the nature of our complex world that delivery often does go wrong, and that trust built through working together, including working through differences and finding compromises, is a stronger foundation to build on. In fact, I believe that tighter relationships between communities and public bodies enables faster correction when things do go wrong, acting as a safety net that enables organisations to be less risk-averse. In addition, I would say, in a democracy, people power needs to be at least as strong as corporate power, and the only way that’s achievable is through solidarity and collective action. I agree that the public sector should be faster at reaching decisions and acting on them, but removing the need to consult is not the only way to reach that goal.

So one thing to keep an eye on is whether the “performative statutory duties to consult” that are revoked are replaced by statutory duties for more meaningful involvement and collaboration. Having no statutory duties to consult, involve or collaborate with stakeholders outside the public sector will lead to a situation where the only people heard by governments are well-funded lobbyists, usually for big businesses.

Who gets to have a say?

When people engage in consultations or participatory processes, they draw on three things:

Their own lived experience, as well as that of their friends, relations and social circle.
Expertise gained through research and study.
Their attitudes and values – what they believe is right, fair and just.

All three of these are important. Different approaches to consultation and participation tend to be good at accessing some but not others.

To give an example, I was on the Oversight Group for the People’s Panel on Digital ID. Having observed it, I know there were people there who had professional expertise (for example in data protection or supporting people experiencing domestic violence) as well as relevant lived experience (for example who were young, old, disabled, trans or digitally excluded). One of the great things about the panel was seeing these different perspectives come into conversation with each other. As with other forms of civic participation (such as juries), people obviously also really valued connecting with people who were very different from them around a shared endeavour. It was a form of engagement that, as James Frith, who was Minister for Digital ID at the time, described as “problem solving, not point scoring”.

But there were undoubtedly areas of both expertise and lived experience that weren’t strongly represented within the digital ID people’s panel itself, and this was a big panel of 120 people – a typical small public participation exercise (involving less than 30 people) is very unlikely to include this range of perspectives directly.

I don’t think the digital ID consultation process was perfect, but it rightly also had two other strands alongside the people’s panel, which brought in different perspectives:

A traditional formal consultation to access expertise and perspectives from a wide range of stakeholders, including business interests.
A set of targeted engagements with smaller communities that might be particularly affected (in the case of digital ID, I believe this included homeless people and the digitally excluded, though nothing has been made public about who was engaged with in this way).

Public dialogues and deliberative approaches are not enough to access the full range of views that public bodies need to consider. Other forms of consultation, involvement and collaboration are needed to access both professional expertise and the voices of smaller and under-represented groups. It’s also important to bring these voices and perspectives into public participation exercises, to make sure the participants are being shown the full story.

What happens as a result?

It is always disheartening when it feels like consultation and public participation has very little effect on what public bodies actually choose to do. When people and groups actually get involved – whether writing responses to consultations, attending engagement meetings, or taking part in a people’s panel – it so often feels like a waste of time, which decreases trust in and the perceived legitimacy of decisions made by the public body, and means that people either disengage or adopt other techniques to get heard. These can range from adversarial campaigning to backroom conversations, which are systemically biased towards those that have the spare time, resources and connections to make them work.

The Cabinet Statement commits to revoking performative consultations, but getting public participation to not be performative requires addressing the performativeness, not just swapping one engagement approach for another. The Gunning Principles provide a benchmark for meaningful consultation – a strong benchmark that can form the basis of judicial reviews:

consultation must be at a time when proposals are still at a formative stage
the proposer must give sufficient reasons for any proposal to permit of intelligent consideration and response
adequate time is given for consideration and response
the product of consultation is conscientiously taken into account when finalising the decision

In any move towards public participation, we either need an equivalent appropriate benchmark, or these principles need to be adapted towards the style of engagement that public participation entails (where, for example, “adequate time for consideration and response” might look quite different from a 12 week consultation window). The key points about proposals needing to be at a formative stage, and the results of consultation taken into account, have to remain.

What I’d ideally want, especially as public bodies get up to speed on how to do public participation well, is for any public participation exercise to be accompanied with a (proportionately scoped) independent evaluation, which is then published. For a quick afternoon workshop, this could be a brief questionnaire to participants; for a longer-term series of engagements, it could include interviews and more sophisticated analyses.

Evaluations would help point out improvements for the public body themselves, for next time; it would ensure that the approaches taken by public bodies are shared with other organisations to learn from; and it would provide evidence that could be used in any campaigns or challenges around the exercise, supporting accountability in the process. In addition, when your work is being evaluated you tend to take more care over it, so this should be a way to drive up quality, even if it will doubtless feel uncomfortable and be politically challenging.

Where’s the infrastructure?

The final observation I want to make is about infrastructure, from two perspectives: the infrastructure within the public sector to support engagement, involvement and collaboration; and the infrastructure outside the public sector to participate in these activities.

Traditional consultations are relatively simple – they require public servants to write a consultation document, publish it, receive responses, and analyse those responses. Good public participation, on the other hand, requires public servants to have ways of identifying a representative set of individuals to take part; shaping appropriate exercises for them to undertake; pulling in expertise and advice to ensure the process is fair and balanced; facilitating workshop sessions; capturing qualitative and quantitative data; analysing the results; and then feeding back to the people involved. This not only requires a wider and deeper set of skills, but also more confidence in engaging with the public, which generally comes from seniority: consultation documents can be signed off, but in-the-moment responses in a workshop can’t, so public servants need to be on top of the messaging, and artful enough not to appear to be stonewalling.

I suspect that this is a place where local governments will have something to teach national bodies. Councils tend to have people who specialise in local engagement with their communities, while central government tends to think in terms of one-way public communications. But either way, we should not underestimate the skills and resources required to involve and collaborate with the public. There’s a range of infrastructure that could help – digital platforms that support deliberations and data analysis; tested workshop plans and facilitation materials that can reduce reinvention; opt-in participant databases to help identify representatives to take part and so on – many of which are described in Demos’s Citizens’ White Paper from two years ago. Now is the time for central government to put more in place to support this.

On the other side, civil society organisations play an incredibly important role in public participation, and more generally democratic accountability and control. They are frequently the route through which potential participants – particularly those in particular targeted communities – can be reached. They can aggregate the lived experiences of their communities to identify patterns and lessons. They are intermediaries for communication that invites the public to get involved, and can confer trust on the results of deliberations. They provide ongoing social infrastructure that brings people together and builds cohesive and powerful communities.

Central government should undoubtedly be less of a micro-manager of local government, but for devolution to work there needs to be effective local accountability. Journalists have rightly been highlighting the need for better funding for local media but frequently it’s civil society organisations that flag stories to journalists and inform councillors and parliamentarians about their communities’ needs. They need support too.

In conclusion

In my particular world of data and AI regulation, digital public services, and public sector AI adoption, the challenges of the pace of change, unpredictable impacts, and need for buy-in does mean that consultation is rarely fit for purpose, so it’s really great to see an emphasis on public participation in the Cabinet Statement.

But I don’t think we should be adopting an attitude of “traditional consultation bad; public participation good”. Rather, we should set our sights on democratisation of the form described by JP Spencer in his paper on Democratic Public Services – i.e. democratic accountability and control – and use whatever combination of engagement, involvement and collaboration techniques work best for achieving that.

Tim Weeknotes

Fri, 31 Jul 2026 00:00:00 +0000

Time to drop some irregular weeknotes with an update on recent events, and some thinking ahead.

Citizens’ Track on AI Governance

Launched in Geneva

A few weeks ago we shared two outputs from a rather intensive few months of work since mid-April:

A new whitepaper: A Citizens’ Track on AI Governance - Alignment, Agency and Accountability - launched with events at the Geneva Graduate Institute, and UN’s Geneva HQ
The PAVE Case Book - a collection of over 80 case studies of participatory practice around AI, which featured as the centre piece of an Expo stand at AI for Good at Palexpo.

I’m really proud of both: which have brought together lots of thinking from the last few years at Connected by Data - and were only really possible because of two factors:

(1) Prior thinking, partnerships and trust built up over the last few years of work.

The design and technical architecture of the case book builds on the case study research I started in my first months at Connected by Data, and the methodology was shaped by Susan Oman’s work on Public Voices on AI evidence review. The research work was enabled through collaboration with Susan Oman, Emily Rempell, Erin Mccloskey, Canning Malkin and Octavia Field-Reid. The whitepaper built on in-depth thinking Rich Wilson, Johnny Stormonth-Darling, Inaki Goni, Claire Mellier and others have been doing in recent years, as well as design work from Jesi Carson that helped us clarify the expression of a number of ideas.

All this was underpinned by Emily Macaulay’s operational efficiency: wrangling all the practical components to have the right people in the right place at the right time with the right kit when we were in Geneva.

There’s many people I’ve failed to mention here - as to produce an exhaustive list would mean I never finish these weeknotes. But when I was writing this week about next steps for the Citizens’ Track I was also struck by how many elements of our work it builds on:

Policy Design Lab - A global policy agenda on collective data governance (June 2023)
People’s Panel on AI and Resources for Effective & Inclusive Public Deliberation on Data & AI Governance (Design Lab) (November 2023)
Options for a Global Citizens Assembly on AI (September 2024)
Public Voices in AI (2025)
PAIRS (2025 and 2026)

And I’m pretty sure each partner invovled could cite a similar list of prior work they have been weaving from.

I could perhaps argue this is by design… though in practice I think it’s more from a case of being open to possibilities.

(2) AI Coding Agents

I’m slightly less comfortable with this one: but I should recognise that our ability to produce an effective Citizens’ Track website, and the PAVE Case Book, was only really possible on our timeline and budget with the aid of a coding agent: AntigravityIDE in my case.

Whilst I’ve wired up Airtable -> Jekyll websites in the past, or created maps and browsing interfaces - the ability to shortcut from idea to prototype in minutes rather than days and hours enabled us to have a dynamic interface to the PAVE Case Book, and made the last-minute set-up of a set of Raspberry Pi-based controllers to drive a touchscreen version that replaced external links with pop-up QR codes, or that could cope gracefully with wifi failures, a lot more straight forward.

Where next

In the weeks since we were in Geneva I’ve been focussed on working out next steps.

Operations and Governance - I’ve been sketching out the creation of a Citizens’ Track Delivery Unit that can take a start-up approach to building out the networks and infrastructure that could support public engagement in AI governance at scale. We’ve been exploring governance and fiscal hosting options for this, and getting closer to an agreed design.
Fundraising - Right now, the next steps of the Citizens’ Track doesn’t have a dedicated budget - so it’s a bit priority to secure funding. We’re working on a number of partnership proposals, and I’m working to try and keep different opportunities aligned.
Case Book 2.0 - Thinking about how the technical architecture of the case book might evolve - particularly towards a mulit-lingual data standard for describing participation processes and their outputs. I’ve been looking back over Aviv Ovedya’s past work on this, the MetaGov Interoperatiblity project, and work on dataset model cards, and reflecting on some of the specific challenges we’ll face.
Resources for deliberation - Exploring how to make sure we can reach the goal of hundreds of community assemblies feeding into the next UN Global Dialogue by providing resources: both to build general faciltiation capacity, and offer specific content that can be used in AI-focussed deliberations.

We’ve also been putting together workshop proposals for the Internet Governance Forum meeting in December, and exploring a possible fringe event of the UN General Assembly in September.

It’s feeling like a big challenge ahead - but a unique moment of opportunity between now and next June, and if we can continue to build on, and expand the partnerships that enabled action in the last few months, it should be both possible, and fun, to work on!

AI in Education

Yesterday I was down in Tunbridge Wells to faciltiate a workshop for the Good Futures Foundation summer school on AI governance. I’ve written that up here

Reading

I recently got a copy of AI and Assembly and whilst I’ve only read the introduction so far, it looks set to be a really critical read for current work.

At our local bookshop I recently picked up a copy of Kim Stanley-Robinson’s Ministry of the Future.

Signing off

My train is pulling back into Stroud, so that will have to do for now!

What we’ve learned about choosing venues for activities

Mon, 20 Jul 2026 00:00:00 +0000

A venue isn’t just the space where you hold an event or an activity – it’s part of the planning about how the session will run. A good venue will enhance the experience of participants and that likely benefits what you get out of it too. A poorly selected venue can be a barrier to engagement – if people even come at all. At CONNECTED by DATA we’ve tried to be intentional in our venue selection for every in person event or activity we’ve hosted (or in fact even if it is in person or online and where online…more on that later).

As a starter of course a venue has to meet your requirements: can it hold enough people, in a seating format that you would like, can it provide (or allow) food and can it provide any tech that you need? Beyond that though the decision making can feel more subtle but, our experiences would argue, is just as important.

Who is in the room?

We aim to design our events to be as inclusive and accessible as possible to all kinds of participants. This can include providing materials in accessible formats, providing ex-gratia payments for those who are otherwise unfunded, negotiating event times to meet most participants’ needs, among other adjustments. (You can read a couple of Weeknotes from Emily where she’s captured some of our approach to this – in participatory events and our unconference).

In terms of venue selection there’s the need to have it physically accessible for all mobility levels, but also consider how sound works in the space. Large rooms with high ceilings – particularly if you then have table-based discussions – can be challenging for those with hearing impairment. We always try to provide a ‘quiet’ room – a space for prayer / decompression (particularly for those with religious or neurodiverse needs).

If you’re holding an event that a good number of people are attending and you haven’t been to the venue before – try to. It’ll help you visualise how the space will work and we always do signage for our events directing people to toilets, food, the ‘quiet’ room, other spaces that we’re using etc.

Who is getting your money?

We started by ensuring any venue we consider is values aligned with us as an organisation. Generally this meant it was a charity or social enterprise, having social impact in a local or national setting. We used universities, research institutions or hotels on occasion, usually when holding events outside the UK.

We are a not-for-profit organisation and will always seek at least three options for venues. We won’t always choose the cheapest if we think a slightly more expensive one will provide the best experience, but we won’t unnecessarily overspend.

Sometimes we asked for spaces from partners – either as a form of ‘sponsorship’ or because they were involved in the work we were doing. Sometimes we were offered spaces by organisations wanting to support the work. It is always worth chatting with others before accepting these offers though. There’s something about ensuring it still meets your needs (not just making it work because the space is free of charge) but it is also worth considering how it may appear using the space. For example we had a workshop to run about how people felt about a particular implementation of AI. We were offered space by a tech company who we were working with in the project. After discussion we felt using the space may suggest a bias, or imply a preference, that we didn’t want to frame the discussions. In that example we used a local community-owned space instead.

What’s the vibe of the space?

There’s lots of different types of venue of course and the ‘vibe’ changes with each of them. We’ve never used high-end/hosted spaces but they do have a place. Community-owned spaces are great for knowing your money is going back into local needs, but they’re often slightly less glossy. They likely won’t have up-to-date tech options; they may not have matching chairs. Depending on your event this won’t matter and in fact a more ‘relaxed’ space can enhance the feeling of the session and encourage attendees to be slightly more participatory in how they engage (rather than just sitting back and being hosted).

We held an unconference in March 2026 in a community hall in a community-owned building. In addition to the relaxed vibe, the building was constantly in use by the local community. If – like us – you’re an organisation that believes in community involvement then this feels even more aligned (and in this specific case we invited the community to attend the event too).

One thing to consider when deciding what end of the ‘hosted’ spectrum to select is how much support you need when running the event. The more ‘hosted’, the more support and the more someone else is holding the support then the more you can focus on delivering a great session. (Of course if you have team members you can take along to do the ‘ops’ then you get that support without needing the venue to be hosting).

London? Or beyond…

The reality is that the UK is designed to get people into and out of London. Often this will mean that London is the best place to hold a meeting that includes people from across the country. However this can result in an actual or perceived London-centrism and so holding sessions outside of London can be appealing.

Most of our events (unless they needed a specific geography) have been held in London. In March 2026 we held a conference in Manchester. The travel related costs were a lot higher than normal (lots of London/SE based people travelling to the NW) but the feedback suggested that London-based attendees enjoyed being pulled out of their city and Northern-based attendees enjoyed not having to travel as far.

How can you reduce the carbon cost?

We have an organisational policy to seek to minimise the carbon costs of our events so we want to enable people to travel by train to our events. We won’t book a venue that’s not walking distance from a train station (we did this once for a residential event as our requirements had to be driven by the residential provision…but we organised taxis and car shares to transfer people from the station to the venue). Proximity to public-transport infrastructure also improves the general accessibility of an event.

We also offer vegan meals by default at our events (as well as catering for other dietary requirements on a person-by-person basis). This can be a varied experience. We’ve had great vegan food – usually from venues with an aligned aim to reduce their carbon cost so they’ve invested energy in developing a good vegan offer. In others we’ve had poorly developed vegan menus. The feedback from our participants is also interesting on this default position (which we are always clear about up-front). Some love the approach and the prompt to try vegan food (if it isn’t their own default). Others have criticised it (some quite strongly) and suggested it negatively impacted on their enjoyment of and engagement with the day.

In person or online?

Most people feel that in-person connection at events is a preference, that it enhances an event. However when balanced against travel time, convenience, and cost the balance can tip more towards holding an event online. We’ve found that hybrid events don’t tend to serve either need and can cause people to join online when, if online wasn’t an option, they’d be in the room.

Online facilitation of sessions is a skill. Although it is a version of the general facilitation skill, it is a specific skill and you need it if you’re going to run a session online. There’s a plethora of online platforms (or venues!) to hold events and who your audience is will need to impact the option you choose. Some organisations don’t allow Zoom access; for many that aren’t Microsoft/Teams based, the Teams experience can be less user friendly; and Google Meet tends not to have the range of ‘host’ tools.

Generally our principle has been to have short (<2 hours) or international audience sessions online, and try to have as much else as possible in person.

We have also experimented with more immersive online experiences including GatherTown and SpatialChat. Whilst both add value around informal and ‘corridor conversations’, attendees are less likely to be familiar with it and this may be a barrier to easy attendance and engagement. In testing people enjoyed the experience but it was felt it required too much “how to use this space” support to create a welcoming space.

And finally…

You won’t get it right for every attendee. Every event we’ve done has had positive and less positive feedback about the venue. We try to listen and learn each time and reflect on feedback when we’re booking another venue, but sometimes different people want opposing things. As the saying goes, you can’t please everybody.

Shoutouts

We’ve used some great venues over the years of CONNECTED BY DATA and here are some of our favourites that we happily recommend:

Toynbee Hall, London
Coram, London
Churchill Hall, London
Mechanics Institute, Manchester
Hawkwood, The Centre for Future Thinking, Stroud (the venue we used for our residential)

Data and AI Public Engagement Community of Practice – a summary

Mon, 29 Jun 2026 00:00:00 +0000

Since July 2024, we’ve been running a monthly community of practice for people who are trying to engage the public around data and AI. That was part of our project, ‘Giving communities a powerful say in public sector data and AI projects’, funded by the Mohn Westlake Foundation, which has now come to an end. (You can find lots more resources from that, such as case studies and event write-ups, here.)

Over seventeen events, we’ve heard from some great speakers talking about their projects engaging the public on data and AI – how they did them, what they’ve achieved, what the challenges were, and what they’ve learned. Notes of all our meetings are available to people signed up to the community, along with some videos of presentations. Some speakers have been able to let us share their presentations more widely – these are all embedded below. There’s also a list of all our speakers at the bottom of this post, with links to their work (where possible).

Some of the themes or questions that kept coming up across the events included:

How to recruit diverse voices, and value their contributions (including through payment)
How to help people engage with complex tech questions, especially given the speed of development, and given some find it boring, including connecting the issues to political narratives – the need for storytelling
How to talk about ‘AI’ when it’s a term that covers many different things
The importance of co-creation from the start
How to convince leaders of the value of public engagement
How to ensure there is a legacy, that decision makers honour the outputs of public deliberation, and to keep people involved in a project or process
How technology can help with deliberative and participatory exercises
The need to share resources and learning

Case studies of public involvement

We’ve been really lucky to have heard about how people went about running various brilliant public involvement initiatives, and what they learned as a result. Hopefully you’ll be able to draw inspiration and insight from them, too.

In October 2025, Giles Lane from the Royal Academy of Engineering presented on their People’s AI Stewardship Summits. These five public dialogues run across the UK – Belfast, Glasgow, Liverpool, Swansea and London – brought members of the public, innovators, entrepreneurs and experts together to explore how the public feels about the implications of AI, and ensure their voices are heard by those working on it. You can read the notes and Q&A, too.

The following month, November 2025, we heard about two quite different projects. First, Jonny Harper, director at Thinks (an insight and strategy consultancy), talked us through their public dialogue for the Home Office on the use of AI in policing. Members of the public considered three different use cases (answering non-urgent 101 calls, finding patterns in datasets, predictive policing) and came up with four positive principles and four things to watch out for.

Then, Sandeepa Arora, a postdoc at The Mohn Centre for Children’s Health and Wellbeing at the School of Public Health, Imperial College London, discussed ‘Data Detectives: 10-year-olds’ views on their health data and how it should be used’. The project’s starting point was that transparency around the use of our data is an essential right that children should be able to access too, and wondered what would happen if children were asked about data. They helped develop a game, ‘Data Detectives’, which allowed them to explore questions around data privacy. You can read the notes from Jonny and Sandeepa’s presentations and Q&A, too.

In December 2025, Janet McLaren from the Global Partnership for Sustainable Development Data talked about their Data Values Project. This included consulting with people and organisations to develop a Data Values Manifesto – the five calls to action include ‘Support people to shape how they are represented in data’ and ‘Invest in public participation for accountability’ – and supporting Data Values Advocates to bring the manifesto to life in their communities.

In February 2026, we heard from Rachel Astall from Beam, the company behind the Magic Notes (now Beam Notes) casework transcription tool. Beam worked with Nesta on a public deliberation process to understand public attitudes towards the use of the tool – the AI Social Readiness Advisory Label. Focus groups and deliberative polling found that 74% of people agreed that the benefits outweighed the risks – and identified possible inaccuracies and overreliance being the most concerning risks.

And in March 2026, Connected by Data’s own Tim Davies ran through some of our work, namely:

Participatory data governance in practice, which involved creating a library of case studies
People’s Panel on AI, a deliberative exercise hosted by Connected by Data around the AI Safety Summit UK in November 2023, which offered a proof of concept about how to do an agile deliberative process
Community campaigns on data, where we explored how we can support communities to develop plans of action and advocacy around data, creating a toolkit
The Good Governance Game, a card game to help people explore different methods of participation
Generative AI in Education - Have Your Say, where we developed a “workshop in a box” to enable teachers to run sessions contributing to a distributed dialogue to understand students views on generative AI in education
Participatory AI Research & Practice Symposium (PAIRS), an international community of people interested in participatory practice around AI which ran online and in-person in 2025, and online and in-person in 2026.

How to do public engagement well

We also had several presentations around how to do public engagement and solving particular challenges.

In February 2025, Maddie Gough and Elizabeth Cunningham from Involve talked about how they approach public engagement. They covered some key themes of good engagement – Clear purpose, Properly resourced, Accessible and inclusive, Gift of thanks and expenses, Balanced information, Accountable decision makers – before discussing the IAP2 spectrum of engagement and different methods. They also shared a couple of case studies: working with Camden Council on their Data Charter, and with Sciencewise on a project called ‘Future Flight 2’ to understand public views on technologies like drones.

In January 2025, we had two speakers focus on how to engage different communities, including historically excluded ones. Henrietta Hopkins, director of Hopkins van Mil (a social research agency specialising in deliberation), shared three case studies: Trust in Digital Identity Services (DSIT/Sciencewise), the Citizens’ Biometric Council (Ada Lovelace Institute) and Grown up? Journeys to adulthood (Nuffield Foundation/Ada Lovelace Institute). She said that some of the key themes were the importance of co-creating the process, creating safe and trusted spaces (in a physical sense, and in terms of facilitation), recognising those taking part were likely to be impacted by the topic, targeted outreach and working with gatekeepers to reach the people that matter, and ethics and safeguarding.

Emily Rempel, then Public Participation Manager at Liverpool City Region Civic Data Cooperative, also used three case studies: Round ‘ere (on what we should be collecting to begin with), Community data conversations (how to engage with minoritised groups), and the Residents’ assembly on data and AI innovation (intended to be as broad and welcoming as possible). Her tips for avoiding only engaging a WHIMS population (white, healthy, in retirement, middle class, skilled) were negotiating time and relationships, humility (something she’d also encountered working with First Nations people on a health data project in Canada), fair payment, working with and not for, and throwing things at the wall to see what worked. (Emily also spoke at Data Bites about the residents’ assembly, and at our conference.)

In October 2025, Connected by Data associate Anna Beckett presented on how to evaluate a deliberative exercise – specifically, the Liverpool Residents’ Assembly that Emily had run. She said that evaluations can be formative (shaping the thing they are evaluating) or summative (summarising what happened) and can focus on process (what happened) or impact (what happened as a result). A formative approach – engaging the evaluator from the start – could be most valuable, as it allows real-time feedback to shape the exercise (though there is a risk of ‘marking your own homework). A theme of recent evaluations was a lack of AI sceptics in the room – either they don’t engage at all, or start the process and then drop out.

And in June 2025, we focused on the use of technology to support better public engagement. Prateek Buch talked about PolicyLab’s use of Pol.is – although his presentation is not public, you can watch him talk about it at Data Bites in October 2023. Hannah Perry from Demos and Georgina Denis from Psi then talked about the Waves project, which aims to use tech to scale public participation and deliberation. The project is working with the London Borough of Camden and South Staffordshire council to pilot approaches, and has a learning network for other interested councils. You can sign up for more recent updates via Demos.

All previous speakers

1: July 2024

Robin Carpenter, Ethics and Governance Lead for the AI Centre for Value Based Healthcare and King’s College London, on Collaborating well with patients when developing healthcare AI
Elizabeth Nelson, Public Engagement, Communications and Impact Manager at Administrative Data Research Centre Northern Ireland, on ‘Co-producing data-driven research with historically excluded groups’

2: October 2024

Dr. Michaela Omelkova, Data Ethics Advisor, Scottish Government, on their public dialogues on data and CivTech challenge on the use of technology in public participation on the use of data and AI

We then had something of an education special:

Edward O’Garro-Priddie, Delivery Lead, and Jake Luscombe, Lead Data Scientist, Faculty, on the AI in Education Hackathon
Dea Begaj, Public Attitudes Researcher, Responsible Technology Adoption Unit, on research on parent and pupil attitudes towards the use of AI in education
Fay Skevington, Head of AI Alignment for Education, Data Directorate in the Department for Education, on how collaborative engagement on data has shaped policymaking.

Connected by Data has subsequently done lots more work with the DfE, through a distributed dialogue, which you can find at https://connectedbydata.org/ai-in-education/.

3. December 2024

This was a workshop on the question, what do public engagement practitioners need to succeed?

4. January 2025

This focused on how to engage different communities, including historically excluded ones, with:

Henrietta Hopkins, Director, Hopkins Van Mil
Emily Rempel, Public Participation Manager, Liverpool City Region Civic Data Cooperative.

5. February 2025

This focused on the different approaches that public engagement practitioners can take - i.e. the menu of different options for public involvement. The speakers were:

Maddie Gough, Engagement Lead and Elizabeth Cunningham, Engagement Lead – Science, Technology and Data from Involve
Sam Nutt, Researcher & Data Ethicist, London Office of Technology and Innovation.

6. March 2025

This celebrated the first year of the project, and featured reflections and discussion from:

Sylvie Hobden and Dea Begaj from DSIT
Annie Radcliffe from the LGA
Shayda Kashef from ADR UK.

7. April 2025

This heard from the Ada Lovelace Institute’s Eleanor O’Keeffe, Roshni Modhvadia and Helena Hollis, who spoke about their work as part of the ESRC Digital Good Network’s recent Public Voices in AI project.

8. May 2025

Julian Tait from Open Data Manchester, on developing principles for putting residents at the heart of how automated decision making is used
Louise Petre from the Government Digital Service, on the role of user research in understanding and improving the relationships we build around and with data.

9. June 2025

This focused on the use of technology in public engagement, with:

Hannah Perry from Demos and Georgina Denis from PSi talking about the Waves project
Prateek Buch from PolicyLab.

10. July 2025

Frederica Longfoot **and **Eli Holderness about the public engagement work of OpenSAFELY.

11. September 2025

Tracey Brown talking about Sense About Science’s Responsible Handover Framework.

12. October 2025

Giles Lane from the Royal Academy of Engineering on their People’s AI Stewardship Summits
Anna Beckett on evaluating Liverpool’s Residents’ Assembly.

13. November 2025

Jonny Harper, Director at Thinks, about their recent AI in policing project
Sandeepa Arora, Postdoc at The Mohn Centre for Children’s Health and Wellbeing, School of Public Health, Imperial College London, on ‘Data Detectives: 10-year-olds’ views on their health data and how it should be used’.

14. December 2025

Helen Miller-Bakewell and David Ross from the Office for Statistics Regulation talking about their new public involvement and engagement in official statistics toolkit for statistics producers
Janet McLaren from the Global Partnership for Sustainable Development Data, on ‘Centering People in Data Governance: Lessons from the Data Values Campaign’.

15. January 2026

Anna Steere from Understanding Patient Data presenting on ‘why tracking sentiment matters for health data policy’.

16. February 2026

Louis Horsley, Research and Insight Manager at National Voices, on the ‘Underrepresentation of ethnic minority data in genomic medicine’
Rachel Astall, Chief Commercial Officer at Beam, on the AI Social Readiness Advisory Label, a first-of-its-kind public deliberation process led by Nesta to evaluate the risks and benefits of AI tools in public services

17. March 2026

Tim Davies, Research and Practice Director at Connected by Data, on our work on participation around data and AI.

The future of the Data and AI Public Engagement Community of Practice

Mon, 29 Jun 2026 00:00:00 +0000

Between July 2024 and March 2026, we ran a community of practice for people who are or are wanting to engage and involve the public around data and AI. You can read a summary of the seventeen monthly(ish) meetings we held in a separate blogpost.

This was part of our project, ‘Giving communities a powerful say in public sector data and AI projects’, funded by Mohn Westlake, which has now come to an end. (You can find lots more resources from the wider project, such as case studies and event write-ups, here.) In this post, we reflect on the experience of running the community of practice, and what comes next.

What attendees thought

We ran a survey (well, spun up Google and Microsoft Forms) to ask what people they thought. Of 565 people on our mailing list, ten people filled out the surveys. Of those, nine had attended between one and four meetings; one had not attended any, but sometimes accessed the meeting notes. Other respondents also sometimes accessed the meeting notes (one regularly), more so than watched the videos.

Other key findings:

We asked people how useful they found the Community of Practice for their work (out of 10), which gave an average score of 6.7
Comments on what people found useful felt more positive: people thought the presentations were interesting and relevant (and often went beyond their own fields), valued hearing about and learning from what other people were doing, and found the links to other case studies useful. People also valued the networking element
When asked what was less useful, several respondents thought the Q&A could be ‘a bit dead’ with fewer people on the call than expected (one answer noted Wednesday afternoon is a particularly popular meeting time and another slot might work better). Some respondents lamented that they did not have the time they would have liked to engage with the resources and the community in general. One respondent found the sessions interesting ‘but I often ponder what should I be taking away from this as a civil servant in terms of how I could help the public be more involved in issues that affect them related to data and AI’
Eight respondents said they would come to future meetings if the community continued (the other two said ‘maybe’)
When asked what people would keep and what they would change – several respondents said they would prefer in-person (‘People who go have really committed and are more up for discussion’ and it’s better for networking), although one said they found online events much easier to attend. One comment said they liked that the focus was ’not academic and addresses broad societal issues around AI’.

We also asked people to suggest possible new host organisations, or people they’d like to hear speak. We include them here to signpost other interesting organisations and relevant work:

PEDRI (Public Engagement in Data Research Initiative)
Understanding Patient Data
Office for National Statistics
DARE UK (Data and Analytics Research Environments UK)
Open Data Institute
Ada Lovelace Institute
One HealthTech
Civic Health Innovation Labs at University of Liverpool
National Co-ordinating Centre for Public Engagement.

One respondent wondered if a collaborative effort – a shared secretariat across different organisations and sectors – might provide a way forward for the community of practice.

We also asked which networks we should share details with, especially inside the civil service. One answer said it would be helpful to loop in someone from the Senior Civil Service to encourage others to come along – ‘more junior civil servants can be understandably really reticent to go to public events for fear of “putting their foot in it”’, and ‘the visibility is not good about this within the CS for this sort of thing. Which is a shame.’

Our reflections

Staying on that theme, one of the things that has struck me most has been our struggle to get civil servants along to community of practice events. We have tried to plug into some relevant networks, but it’s difficult – especially from outside the civil service – to identify what these are, let alone how to work with them. As someone who helps run several other public service-facing event series (including Data Bites and TransformGov Talks), getting civil servants along to events feels like a wider challenge.

Anecdotally, there are understandable reasons for that – time pressures (especially with staff cuts), changes to working patterns post-Covid, other networks which may fulfil their needs (e.g. LGA networks). But there are also more troubling ones: the difficulties in sharing useful events with civil servants, civil servants having a sense of powerlessness to change anything having become used to firefighting in turbulent times, and damaging signals from ministers, such as the civil service speaking ban and clampdown on staff networks, which disincentivise civil servants from seeking out opportunities.

This may have affected the liveliness of the discussion after the presentations, as some survey respondents said – our first ever event was our best-attended and the discussion, on camera and in the Zoom chat, was the most flowing we had. That said, perhaps we could have structured the sessions differently to encourage others to share thoughts and insights, and taken more of a workshop approach on some occasions (as we did in December 2024).

One of my reflections a year into the project (back in March 2025) was that we were still in the relatively early stages of public servants involving the public in discussions and decisions about data and AI – perhaps we’ve not yet reached critical mass of practitioners who see themselves as such. (Relatedly: we were sometimes still grasping for the right language – public engagement? Involvement? Participation? – and the right way to talk to other people about it.)

That was reflected in some of the discussions we had in the community of practice, with lots of practical questions about how best to do things. This ranged from finding a diverse range of participants, to what methods to use, to managing people’s expectations about what the work could lead to, to how to evaluate the success of a deliberation. It felt like attendees really valued these insights, and the chance to learn from others.

My main reflection is this: there’s some really great work going on to involve the public that leads to better policies and outcomes. It’s tempting for organisations to see public engagement as an expense (and it does take time and money to do well) and a box to tick to add a veneer of trust. But the work we heard about was of real value in designing better products and policies, and supporting people to adopt new technology. Those involved were generous in giving up their time and insights to support others to run similar public exercises. There are already some brilliant case studies on how to do public involvement work well (see our other blogpost for some more detail), and a strong desire from people to learn from them and conduct their own deliberative and participatory exercises. The field may be relatively new, but it’s got off to a good start.

What next

As we said at the beginning of this post, the funding for the community of practice (and the wider project of which it was part) has come to an end. Where we had permission from speakers, we’ve published videos of presentations on individual meeting pages, on our YouTube channel and collected them in this blogpost. We’ll also be keeping the notes live for people signed up to the community of practice.

We don’t currently have any plans to run more meetings of the community of practice, and we are not handing it over to another organisation. But we hope others might pick up the mantle. You can also take a look at some of the other organisations in the field further up this post, some of whom run regular events (such as PEDRI).

All that’s left for me to say is a huge thank you – to everyone who came along and asked great questions, to the team at Connected by Data for their support, and to all of our excellent speakers. Thank you!

Ten lessons on public sector AI from our work with education unions

Mon, 22 Jun 2026 00:00:00 +0000

Between July 2025 and April 2026, as part of our project to build worker power to shape public sector AI, Connected by Data and the TUC worked with nine education unions on an action learning programme to test how to catalyse the union movement’s capacity to understand, negotiate and advocate on AI.

This work threw up several lessons that apply more generally to the adoption of AI across the public sector – the risks and benefits; the challenges and opportunities; and the roles that workers can and should play in AI’s development, deployment and evaluation – which I’ve thrown together into this listicle.

1. Public services are not in control of how the public uses consumer-facing AI, but have to respond to it.

Educators are already having to respond to the ways in which students use consumer-facing AI such as ChatGPT to help them with homework and how parents use it to write complaint letters. Their duty of care means they need to understand and respond to the impacts of the use of AI on children’s mental health and resilience (just as they have had to respond to children’s changing media diets, and use of social media). They also need to be thinking about how to equip children with the right skills for the future.

These examples illustrate how public services don’t just need to focus on how they adopt AI, but also on how they respond to the societal adoption of AI, which may change what they need to do and the volume of work they need to handle. In the medical profession, mental health practitioners need to understand how to treat AI psychosis; frontline doctors need to learn how to diagnose patients who report their symptoms using language shaped by AI or have been taken in by AI-generated misinformation. Reduced friction in interacting with public services is leading to additional demands on Employment Tribunals and an increase in quantity and decrease in quality of Freedom of Information requests.

All this means that the government’s approach to regulating consumer-facing AI matters to public service provision, and that at the same time as public bodies are considering how to adopt AI, they also have to work out how to adapt to demands that have changed because of AI.

2. Public services are in a poor state and AI cannot solve some of the deep challenges they face.

The education union’s joint statement produced by this project said “We recognise that AI and education technology (EdTech) can complement human-centred education. But we know that AI cannot solve the education system’s deeper problems: underfunding, teacher shortages, overwhelming workloads, or the social and economic pressures educators face every day.”

There are similar deep challenges across the public sector, brought about by under-investment and changing demographics and requirements.

AI should not be used to work around and temporarily alleviate challenges that require more radical rethinking. Having a clear agreed vision and values, and taking a problem-first, rather than technology-first, approach, helps to align everyone’s incentives towards prioritising changes that will make a difference.

3. The impacts of the adoption of AI by public sector workers go beyond their immediate interactions with AI systems.

The most transformative uses of AI (and other technologies) require end-to-end reinvention of services and processes: the introduction of technology is a trigger for rethinking existing ways of working.

Non-transformative introduction of AI simply automates pieces of processes or layers systems over existing (broken) services. For example, the IT infrastructure in many schools is outdated, so simply introducing an expectation to use AI introduces more work – getting laptops working in a classroom; helping pupils to reset the passwords they’ve forgotten and so on.

The effective introduction of new technologies like AI are organisational change processes. These require the involvement and buy-in of the workforce, which has intimate knowledge of what works and what doesn’t on the ground.

4. All public sector workers are feeling the impacts of AI.

There’s a tendency to focus on AI systems specifically supporting teachers, doctors, or social workers. However, AI is being built into systems across the board, from CoPilot in Microsoft Office, to HR systems and sector-specific administrative software, and therefore affects all public sector workers.

Specialised software, targeted at particular public sector workers, should be co-designed with them. (Beam’s Magic Notes was designed with social workers, for example, and the government has committed to AI tutoring systems being co-designed with teachers.)

On the other hand, when non-specialised software, such as office suites, start to incorporate AI, there needs to be workplace-specific negotiation to co-design policies around its use. This is particularly the case around HR systems that may be introduced to support decision making around recruitment, and job progression, but also applies to developing workplace norms for recording and transcribing meetings, answering emails or writing reports.

5. The introduction of AI in the public sector affects the public as well as public sector workers, and the relationships between the two, which in turn affects public attitudes to the state.

One narrative about AI, evident in the Blueprint for modern digital government, is that introducing AI will allow “public sector workers to focus on delivering the crucial front-line work that will always require human relationships, judgement and empathy”.

The reality of the impact on relationships is more nuanced; other work on pupil attitudes to EdTech reveal that some pupils find it disrespectful for teachers to mark their work using AI, for example – something many technologists (though not many educationalists) might otherwise view as an admin task. Doing a relational activity more efficiently changes the quality of that activity for both parties; optimising for respect may be a way to ensure that AI does not end up diminishing the public’s experience of the public sector.

6. There is the same variability in how public sector workers (and union members) feel about AI as there is in the general population.

Some public sector workers are early and enthusiastic adopters of AI, trying it out even in places where it hasn’t been proven. Some are willing to adopt it but need help and support to feel confident to do so. Some are deeply skeptical, believing that AI is immoral, actively damaging and should be resisted.

This diversity in worker attitudes inevitably leads to variability in use and impacts across settings and brings into play questions around the balance between professional autonomy (both to adopt fast and to not adopt at all); the need to innovate and learn-by-doing; rules and safeguards; and training and support.

7. AI products and services implicitly encode values and political choices, particularly as they reduce relationships between citizens and the state to numbers.

In education, student-facing EdTech can only measure certain things, such as time taken on tasks, percentage accuracy on questions, or when homework (or preparation) is started and completed. This data is readily available, and it therefore becomes something that appears on teacher- and management-facing dashboards.

Goodhart’s Law states that when a measure becomes a target, it ceases to be a good measure. Similarly, when something starts to be measured, it becomes what’s important, especially when evaluating and managing performance. But many things in education are not easily measurable, particularly within applications or through screens. It’s hard for an algorithm to measure conceptual comprehension or identify the missing piece that’s blocking understanding in a student; hard to measure their critical thinking or creative skills; hard to measure how happy they are, how frustrated, how fulfilled, or how much they trust the adults in their lives.

The voices of public sector workers and unions, as well as academics and other civil society voices, are important to work out what’s useful to measure, as well as to challenge the dominance of measurable things in the way we think about what’s important from public services.

8. This all means that the impact of AI interacts in complex ways with the public sector context in which it is used.

The hope is that AI can help under-resourced public bodies catch up; the reality is messier than that.

Some struggling public bodies may be able to use the introduction of AI to strengthen failing IT infrastructure and transform practices and processes that weren’t working, or find that AI fills a genuine gap in their provision, whereas public bodies that are already doing well may find that AI doesn’t add a meaningful benefit on already-great services.

On the other hand, other under-resourced public bodies may be the ones that struggle most with AI adoption, precisely because they are under-resourced. For example, the leadership of under-resourced schools are likely to be too busy fire-fighting to have the time, capacity, or imagination to incorporate AI thoughtfully. This can either mean they don’t adopt AI at all, or more likely adopt it in ways that are more likely to harm, and less likely to help, their students or staff. Meanwhile, more well-resourced schools may be able to have the time and resources to target their adoption of AI more effectively, building on top of strong existing capacity, thereby increasing the gap between provision between the best and worst schools.

9. The evaluation of AI tooling is particularly challenging in these circumstances.

First, real-world evaluations of AI tools are lacking, and public bodies such as schools are adopting AI without evidence, as the result of persistent sales tactics, political pressure, and a fear of missing out.

Second, evaluations necessarily trail implementation, and traditional evaluation timelines and techniques do not meet the demands of rapid innovation and adoption. Individual applications also change underneath the feet of public bodies as new versions are developed to take advantage of new technologies: something with proven benefits that they adopt today, can change tomorrow as new features are added or algorithms tweaked.

Finally, evaluations tend to be narrow and seldom account for the varied levels of impacts discussed above. All this means that evaluation has to be rethought, to get meaningful and timely evidence into the hands of decision makers faster. Worker and union voice is key to reflecting the felt impacts of AI adoption, and mechanisms are needed for them to be heard within the real-world evaluation of these tools.

10. Unions have a particular role to play in looking beyond the near-term and individualised impacts of AI adoption, towards the systemic effects on the workforce they represent.

For example, unions are attuned to situations in which AI service providers are capturing data that may then be used to create AI systems that replace human workers. Unions think about both deskilling of the existing workforce, and building pipelines of new workers to support future sector needs.

As such, unions are well placed to assist the government to think through the transformational changes brought about by AI, and the steps that need to take place to ensure these are beneficial in the long term.

Emily Weeknotes

Thu, 11 Jun 2026 00:00:00 +0000

What I’ve been doing

Adam left our team last month. Obviously this is sad for us as he was great at what he did and a nice energy to have around, but he has made his part time role with TUC a full time one and frankly that’s great for everyone the TUC helps. There’s some administrative processes around someone leaving - as they prepare to leave (including a leaver’s discussion to capture their reflections on organisational culture) and once they’re gone (such as shutting down accounts and deleting data). So I’ve been doing that. It now means there’s only three of us in the team, and that’s a 2.0 FTE at the moment. As a result I’ve jiggled our meetings a bit and rolled all our checkins into Confab - that’s now taking place weekly again on a Monday morning. It’s about being adaptive to the needs of the organisation.

I promised (in my last weeknotes) a selfie with a bus stop if I found one of the “Let’s Talk AI” ones…I found one, so here it is.

I’ve been trying to finalise the annual report (my yak shaving experience was unsuccessful which I’ve found hard to accept and had to send a “confession” email to Jeni explaining that I couldn’t do the annual report upload); and we’ve had the draft end of year accounts come through to review. There’s been the usual end / start of month activities such as monthly reconciliation.

Since my last writing there’s been a decision that we - with others - will be running a Expo Booth at the AI For Good Summit in Geneva in early July. I’m trying to wrangle and drive the logistics on that. Around the Booth itself we’re trying to support people booking travel and accommodation, and securing funding for that. We’ve never done an Expo before, and doing one as a collective means a number of different voices/opinions. It is very busy! I am incredibly lucky to have been asked to attend and staff the stand for the week though, so … I’m off to Switzerland! That’s exciting and hopefully I’ll have a couple of hours each evening to stroll and explore the city too.

What I need to take care of

There’s still lots to do for the Expo so that’s my priority.

Getting the end of year accounts finalised next week will be important to then get everything published and re-forecast some elements of the budget for this year without Adam’s time included.

I really really need to make some progress on the regeneration options for Connected by Data that we need to think through, review and iterate - all ahead of a Board meeting in September where a decision will be made. September is rapidly approaching. I’m struggling a little to find the focus / headspace chunks of time to do that work.

I’ve got some associated actions around that too. Writing an ‘ending well’ plan…although as I note that here I realise that will best be done after the Board have made their decision in September as they will have an impact even though whatever they decide it won’t be Connected by Data as it is now. I also have a couple of reflective blogs to share too. Memo to me: Create a ‘project’ called ‘Ending Well’ on the website to tag related blogs etc. [Edit: I’ve now done this]

We’re also in the early stages of planning a public sector focused event in mid/late October; a Green Party Conference associated event (2-4 October); and an input to the TUC Annual Congress (13-16 September). It certainly won’t be a quiet summer!

What I’ve been inspired or challenged or moved by

In some weeknotes in February I mentioned how I’ve been following the weekly Premier League football match predictions by the usual pundit (Chris Sutton), a changing celebrity guest, and - for the first time this season - Co-pilot (LLM). As we’ve now finished the season here is the final result.

Going into the final day of games Co-Pilot and Chris Sutton were level on the number of weeks where they had an “outright win” (more points from their correct predictions than anyone else). But Chris Sutton was in second place as he had less weeks (2, to Co-Pilot’s 4) where he had tied with someone else - so he needed to score more points in the final round of predictions than Co-Pilot in order to win.

He didn’t.

The human predictor only managed two correct results with no exact scores, giving him 20 points. Co-Pilot managed four correct scores and no exact results, to end up on 40 points.

380 Premier League games were played this season and the result is the over the course of that Co-Pilot, using the predictive LLM power that it is, was more accurate than an expert human. I wonder what best explains this. There could be something about a human being more likely to factor in context elements (a new manager, something happening off the field that’s upset the players etc) and perhaps this doesn’t impact on results as much as we think. Or perhaps the LLM simply plays the law of averages. Co-Pilot of course didn’t know it was involved in a competition - it got the same single prompt every week - whereas the human may have been factoring in games where they could take a risk on a score - and if they got it right would’ve scored more points. If I start thinking about it too much I wonder if even “when” they did their predictions will have had an influence and if they didn’t predict at the same time then the context may have been different (e.g. a player picking up an injury in a midweek game, a manager announcing in a press conference that they’re not likely to play someone). I find it fascinating. I’ll be interesting to see what happens next season and what is the most successful prediction method (and whether Chris Sutton changes anything in his process).

What I’ve been reading

I finished the book I was reading in my last weeknotes and immediately put the sequel on my ‘to be read’ list. I very excitedly then received a notification from my local library that Philip Pullman’s ‘The book of dust volume three: The rose field’ - the final book in the Book of Dust sequence (which is set both before and after the famous ‘Northern Lights trilogy’) was available for me to collect. As I picked up the reservation (it is very much in demand so I’d been waiting around six months) I discovered it was 621 pages (I’m normally a ~300 page book reader) and knew immediately I wouldn’t get that finished in the three week loan period. So I have relinquished that, to come back to it when it is no longer in demand and I can have an extended loan.

Having had a recent weekend in a hired camper van (and living all my best Enid Blyton inspired dreams) I have started my annual re-read of ‘The Secret Island’ by Enid Blyton which I want to get through quickly as in the library yesterday I also picked up my next read which I already know I’m going to enjoy (non-fiction)…what a tease…

Adam's leaving blog

Thu, 28 May 2026 00:00:00 +0000

I joined Connected by Data in January 2023, just months after OpenAI unloaded ChatGPT on the world. Suddenly it seemed that the politics of technology was on everyone’s lips and mine - though often in very different ways.

We sorted through the breathlessness of the hype to figure out whether the foundations of social and political life were really being shaken. I spent a lot of time with podcasts and reading around data and AI - and listening to my colleagues. Three years later, as I step down from Connected by Data to move full time (from part-time) as the TUC’s policy lead on AI, a lot has happened but the uncertainty has barely moved.

Fortunately, Connected by Data had a firm grounding. Founded by Jeni with the premise that what matters most about data and AI are the collective dimensions, stretching beyond individualised rights and technocratic framings. This of course was true before ChatGPT but as a form of automation technology, generative AI does throw up pretty stark and familiar questions; of risk and reward, and who has the power to shape technology. Nowhere is that more apparent than at work. The power relations of working lives have long necessitated the collective endeavour of workers, in workplaces and as a structural countervailing force to other actors, such as employers, big tech and policymakers.

That was the case I made when I joined. And in a manner characteristic of Jeni and of the organisation that I am deeply grateful for, I was given the autonomy and support to build out a deep programme collaborating with unions and the wider labour movement. This included visiting the working men’s clubs at Port Talbot steel works and picket lines at Amazon’s Coventry warehouse, to figuring out how to embed worker voice in Welsh public sector procurement, or support long-term capacity building projects with the CWU in the telecoms sector and with unions representing hundreds of thousands of education workers in partnership with the TUC. These are the experiences that connect the big ideas to the lived realities, and are a powerful check on the temptation to get overly abstract when discussing AI.

This work of building countervailing sources of insight, policy ideas and political power was paired by the development of the Data and AI CSO Network. The Network set out to bring together fragmented, siloed and sometimes competitive civil society and social movement actors into a common space. Not as a formal coalition - which would have scoped out too many participants - but as a forum for exchange of tips and gossip, of expertise and of friendly support; from submitting joint legislative amendments, to amplifying research or co-hosting events. The Network had some pretty high profile interventions, notably the massive open letter ahead of the UK AI Safety Summit that challenged Rishi Sunak’s corporate and ‘existential risk’ dominated event. The letter set out the demands of leading computer scientists, unions representing millions across the world, big brand NGOs and smaller community outfits and businesses to be part of framing and shaping the AI debate. With media from Vanessa Feltz to regional radio and international broadcasters it helped to pull AI’s societal harms into the frame.

More impactfully, though less obviously, the Network has helped to catalyse the relationships and infrastructure that are the prerequisites of any effective movement. This continues, though it’s fair to say that the work of convening disparate groups is a real challenge, and we have learnt as much as we have succeeded!

As I join the TUC full time, the power and centrality of workers is increasingly recognised as vital to the practicalities and politics of AI - in the workplace or at the ballot box. I feel genuinely privileged to have played a role in that. The hopeful shift also underscores a hypothesis Connected by Data was built on: that you can’t understand a technology as an individual or technical challenge only, but one that has powerful collective, political and governance implications.

I still spend a lot of time with podcasts and reading around AI - and listening to my excellent colleagues. Regardless of how the issues unfold, it’s clear in my mind that I will carry as much from the culture of Connected by Data as from the substance of the issue itself: Real autonomy with accountability among colleagues, collaborative and generous with partners and culturally open yet decisively led, with both competence and care - combinations that are all too rare. I aim to emulate this in my next steps - thank you all for the opportunity!

Do reach out at my TUC email or on LinkedIn.

Tim Weeknotes

Mon, 11 May 2026 00:00:00 +0000

As always, it’s been a while since my last weeknotes - and a bit of a whirlwind since I returned from a fortnight off over Easter inter-railing with family around Europe. I’d only just got back, when I was off on the rails again to Geneva to run a roundtable exploring next steps for our Citizens Track on AI Governance project. That got a really positive reception, and so I’ve been somewhat sprinting since to both follow-up on momentum, and make sure we hit some key milestones back in Geneva for the first UN Global Dialogue on AI Governance and the AI for Good Summit in early July.

We’re at an interesting point of having lots of energy coming into the project, with many partners contributing and more wanting to contribute, but activity perhaps getting ahead of both governance and sustainable funding - so trying to keep my eyes both on delivering an evidence-based case for centring public voice in AI in July and the kind of long-term inclusive field-building that we started with PAIRS.

Reading and reflecting

I’ve got quite a lot of writing to do in the coming month, working on the Citizens Track white paper. As I’ve been spending the last few months on more organisational tasks and not as much analytical writing - and my note-taking practice has rather collapsed, I need to get back into training, so I thought I would get back into the right frame by both making more time for reading, and spending the rest of these ‘weeknotes’ reflecting on what I’ve read.

Mysticism, black boxes and data

Olivia Guest, Nancy Abigail Nuñez Hernández and Mark Blokpoel’s new pre-print on Understanding Artificial Neural Networks: Mysterianism about Known Mechanism is Mysticism challenges the commonly heard claims that ‘no-one really understanding how AI works, not even the people building it’. Guest et. al. convincingly deconstruct these claims, showing that we can (and do) have mechanistic understanding of many systems even when their behaviours are unpredictable (they use the analogy of the double pendulum), pointing to the idea that an inability to predict all outcomes, or audit all internal states of a system without running it does not equate to a lack of understanding, nor justify an abdication of responsibility to offer appropriate explanations.

The paper also offers a useful treatment of ‘black boxing’ (analysing a system based on inputs and outputs (I/O)), which provides useful additional grounding for the approach we often take in participatory processes of avoiding discussion of the internals of AI systems, and focussing primarily on I/O. This links to one of the critical closing conclusions of the piece: the fundamental element of AI that is practically unknowable is not the mechanisms of computation, but “the dataset and the world” (p 15). In full, the authors state:

“…we can and should admit that slapping a statistical test, whether a single logistic regression or a billion onto a dataset will never bring understanding of the dataset or the world. Only science can do that. Laborious, slow and steady, deep and thoughtful scientific theorising…” (Guest, Olivia, Nancy Abigail Nuñez Hernández, and Mark Blokpoel. ‘Understanding Artificial Neural Networks: Mysterianism about Known Mechanism Is Mysticism’. Preprint, Zenodo, 7 May 2026. https://doi.org/10.5281/zenodo.20071869.)

For me, this feeds into current reflections about the importance of bringing data back into view in discussions of AI, and makes me wonder about the kinds of participatory exercises we might design that help people get to a specific understanding of the fact that one critical explanation of the behaviour of an AI system in relation to a particular domain or task comes from looking at the training data it might have had access to, and exploring how representative or comprehensive such data is, and what unknowns scientific research or data collection efforts are still seeking to address.

AI Agents and Democracy

In A blueprint for using AI to strengthen democracy Andrew Sorota and Josh Hendler present a useful diagnosis of some of the threats to a public sphere that might be brought by an era of AI agents, exploring how delegation of political debate to individual AI agents could lead to a “collection of private worlds, each internally coherent but collectively inhospitable to the kind of shared deliberation that democracy requires.” While this points to real risks, I want to pick up on two particularly problematic assumptions in the article.

(1) Firstly, it assumes universal adoption of AI (“A public sphere in which everyone has a personalized agent”) and argues forward from this assumption. Prior experience tells us technology adoption will be unequal, beset by digital divides. The situation we must be addressing is not one of perfect technology adoption, nor a prior perfect democracy only disrupted by technology - but one of an imperfect and unequal existing democracy, confronted by technologies that, even if better aligned with democratic values, will not reach all hands equally.

(2) Secondly, I question the assumption that ‘more and better AI’ is necessarily the solution. Rather than additional layers of technological mediation, what of dis-intermediating political dialogue? Local and face-to-face democratic conversations don’t face the kind of identity verification challenges that the authors highlight digitally-mediated discourse will have to contend with - and they have a much higher chance of supporting the social, solidarity-building and connection-building functions of democracy - not just the informational.

While I agree with the authors that “What is needed is a new generation of democratic infrastructure”, the omission of social from the specification of this as a ‘technological and institutional’ project is telling.

Unpacking the Civic AI Imaginary

Earlier this year M. R. Sauter posted a pre-print exploring ideas on AI as an Excuse in Participatory Systems: Identifying and Describing the Civic AI Imaginary and its Impact on Representative and Participatory Governance. I’ve been meaning to read it for a while, but Sorota and Hendler’s piece discussed above brought it to the top of my papers pile.

Sauter surveys some of the assumptions and values baked into a ‘Civic AI Imaginary’ such as:

5.1. The value of digital equivalency: a convincing performance of intelligence via natural language is the same as an individual intelligent human for most purposes…

5.2. The imperial epistemology of data: those things that are important for governance can be rendered into computationally legible data…

5.3. Generative and predictive systems are truth-telling systems… [that] … can reveal the core hidden truths of individuals preferences, desires, and behaviors…

5.4. Democratic governance is a product or procedure that can occur absent the governed

Each of these are deeply problematic, and yet, as Sauter notes, they lie behind many efforts to digitise democratic discourse. I particularly enjoyed the critique of the idea of ‘data doubles’: the creation of AI agents to stand in for human subjects:

Though often called data doubles, they are more precisely described as data homunculi, as they are fundamentally misshapen depictions, warped toward the perceptive limitations of the technology itself. [Sauter 2017]

The article goes on to outline how democratic governance is much richer than simple the aggregation of preferences, or production of decisions at scale - but instead involves relational processes of working out how to live in community with others. The piece powerfully concludes that:

The civic AI imaginary permits, encourages, and excuses the use of computational systems as replacements for people in participatory and representative governance structures. It is fundamentally corrosive to the ethos of address-ability and response-ability that is fundamental to any functioning democratic project. However, the success of the civic AI imaginary in the space of participatory and representational systems is reflective of how the US democratic project has already eroded. The pressures of efficiency, technocratic quantification, and optimization already present in actually operating governance structures mesh with the values of efficiency, industry deference, and imperial data epistemologies that compose the civic AI imaginary. Actually existing AI systems and the civic AI imaginary are not specially suited to deployment in participatory and representative systems. Rather, they function alongside existing structures of power to provide intellectual and ideological cover for the continued erasure of the governed from structures of governance.

In work on the Citizens Track paper, we’re going to need to grapple with questions of if, and how far, we turn to, or allow, uses of digital platforms and AI systems in supporting particular aspects of distributed and transnational deliberation. Keeping in mind the problems of a ‘Civic AI Imaginary’, and making sure any use is justified in terms of a richer democratic imaginary, will be vital.

Citizens' Track on AI Update - May 8th 2026

Fri, 08 May 2026 00:00:00 +0000

Thank you to everyone who has been getting in touch to support the Citizens Track.

We were deeply disappointed not to be able to connect with many of you in person last week due to the troubling cancellation of RightsCon, but I am happy in this update to share news on how you can input to our Case Book to showcase existing public participation on AI, as well as an invite to an online update call next week (Tuesday 12th May, 1pm UTC) where we will be sharing more about plans for activities at the UN Global Dialogue / AI for Good in July.

Update Webinar: Tuesday 12th May - 1 UTC

Join us on Tuesday 12th May at 1pm UTC (9am Washington / 2pm London / 3pm Geneva / 6:30pm Delhi) for our next open call to discuss:

The White Paper on a Citizens Track
The Case Book of participatory AI
Citizens Track presence alongside UN Global Dialogue

Register for the zoom call here.

Help us showcase existing public participation on AI

Have you been involved in commissioning, running or researching participatory processes on AI over the last few years? If so - please nominate or provide an entry for the Public AI Voices and Engagement (PAVE) Case Book.

This will form the heart of a showcase of existing public voices on AI, and a collection of inspiring and good practice examples of engagement methods, that we will be showcasing alongside the UN Global Dialogue on AI Governance and AI for Good in July.

You can nominate a project to be included in the case book by answering five short questions, or, you can provide a full description of the What? Who? How? And What was said? from your project through this simple online form.

We’re looking for examples from around the globe of:

Inclusive public attitudes research;
Public dialogues on AI;
Deliberative processes;
Citizens assemblies;
And other projects that have invited communities to shape AI development, policy or practice.

We recognise that participation takes many forms, and so the full case study form asks for the unique contribution each case makes to good practice for the particular context (geographic, sectoral, etc.) that it comes from.

We’ll be building the Case Book on a rolling basis, so early submissions are welcome, and only submissions received before 5th June can be considered for the July 2026 Showcase.

The Public AI Voices & Engagement (PAVE) Case Book project is funded and co-hosted with by the University of Sheffield’s Centre for Machine Intelligence, and co-hosted by Connected by Data in partnership with the Citizens Track on AI, the Participatory AI Research & Practice Symposium, the Ada Lovelace Institute, Geneva Graduate Institute, ISWE Foundation & partners.

Citizens Track at the AI Equality Festival of Ideas

We’re taking part in the AI Equality Network’s Festival of Ideas on 22nd May, 20:00 - 21:30 UTC with a session titled “Towards a Citizens’ Track on AI: Putting People in the Lead”.

This event, for a public audience taking part in the Festival of Ideas, will be an opportunity to hear about some of the ingredients that inform a Citizens Track - from grassroots engagement to global deliberation - and to explore the vision of embedding citizen voice in AI governance.

Our Global Dialogue consultation submission

We made a written submission to the Global Dialogue co-chairs last week. Thank you to everyone who contributed to discussions, or shared comments on the draft.

We were happy to also collaborate with a number of other organisations and networks submitting responses, and to see the Citizens Track recognised across a number of submissions.

The Ada Lovelace Institute have written about their consultation response, including a focus on “Public deliberation as essential infrastructure”, noting that:

“Public deliberations must feed directly into governance decisions. The Global Dialogue has a concrete opportunity to institutionalise public deliberation from the outset. It is easier to do this now than to fix later.”

Other updates from the Citizens Track

We’ve got a lot of activities underway over the coming weeks. Right now, our priorities include:

Strengthening governance of the Citizens Track project, and drafting partnership agreements / structures. Look out for a request soon about how you want to formally affiliate to the project;
Finalising and launching a project website;
Planning for side-events and activities around the UN Global Dialogue on AI Governance in Geneva (6th - 10th July), including a presence in the AI for Good Expo;
Fundraising/mobilising resources to be explore supporting distributed dialogues linked to the UN Global Dialogue on AI Governance and 20267 Geneva AI Summit.