WHAT WE LEARNED
Lessons and limitations
The process of participating in the roundtables yielded insights into the importance of engaging with civil society and building a thriving AI ecosystem across stakeholder groups. These insights will inform the next phase of our work and our hope is that they may be useful for others too as we continue to explore together how to make public engagement impactful.
Working with organisations with specific expertise in each topic enriched the discussions
Working with civil society and social impact organisation to convene each conversation meant that we had a more diverse mix of participants and a richer discussion than we could have achieved alone, or with any single organisation. Working with expert groups also helped us to grapple with the broad scope of the topics we wanted to discuss. Some topics, like data and safety, proved so fundamental and expansive that knowing where to focus required collaboration. Agendas were designed to ensure conversations would not be dominated by a single perspective or a narrow set of risks.
Pre-reading was influential in how conversations were framed
The process of compiling the right pre-reading for participants was instructive and also required close collaboration with the convening parties. Without wanting to overly influence the conversation’s direction, it was important to create shared context and establish a baseline understanding of the questions under discussion and the goals of the roundtables. Thoughtfully composed pre-reads were important to achieving this balance.
A range of organisations had advantages and trade-offs
While working with different organisations for each roundtable meant that we could delve deeper into the topics we wanted to explore with a broader range of people, a single organisation may have lent more consistency to the roundtables. We chose not to be overly prescriptive in how the conversations and their outputs were structured and two conversations diverged even from the roundtable format¹². That approach was effective, but a single organisation may have increased cohesion and allowed parallels and thematic insights to emerge more readily.
Imagining the future productively is hard
Future scenarios are challenging to engage with, especially when the trajectory and ideal destination are unclear and the technology is evolving at pace. In some cases, it was easier to reach alignment on existing issues than on how best to handle possible future scenarios. In others, discussion focused on potential future governance mechanisms without bridging the gap between these ideas and the current state of play. Established frameworks for thinking about the future, like the Three Horizons model,¹³ were drawn on in some of the roundtables to productive effect.
The Three Horizons (3H) model is an adapted framework that helps us conceptualise long-term social change – in the case of our roundtable series, related to education and learning.
12 One was a workshop we contributed to with Brookings Institute on global governance, and the other was an exhibition we sponsored with UAL: Central Saint Martins (CSM) to design an experiential exhibition imagining AI Futures.
13 Leaders Quest, ‘Three Horizons Model’: https://leadersquest.org/three-horizons-introduction/#:~:text=It%20charts%20Horizon%201%2C%20the,our%20 desired%20future%20a%20reality. and McKinsey & Co., ‘Three Horizons Model’. “https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/ourinsights/enduring-ideas-the-three-horizons-of-growth”
Recency bias means that generative AI can dominate conversations
Generative AI dominated many of the conversations because it is so prevalent in the public imagination at present. Yet AI capabilities extend far beyond it and many debates have much broader relevance. Communication and education by AI developers is a prerequisite for thoughtful civil society engagement because not all sector-specific experts will also be AI experts.
Connections were forged and ideas generated between groups
Participants who met through the roundtables identified areas of shared interest and forged connections, leading them to pursue ideas together that we hope will yield concrete and beneficial outputs.