Thread by @sburall

Next up at #AIEthics is a panel on Public Engagement. Feat @stephenjcave @sarahcastell, Kate Coughlan, BBC, @DrDihal, Jessica Montgomery, Royal Society and @HughMilward 1/

'how would you define #ai to a friend', lots of accurate answers, but many emotive 'scary robots' 'computers doing things to you, I had it' 2/ #AIEthics

That from @stephenjcave quoting survey by Kate (I think) 3/ #AIEthics

He notes that the public need a good foundation of knowledge about #AI, but that isn't enough. And big qu is how to engage public in debate when they are so diverse and numerous 4/ #AIEthics

@stephenjcave focusing on the question of what meaningful public engagement is. I'm going to try to tweet sparingly.... 5/ #AIEthics

@DrDihal looking at the extremes of the narrative about #AI and asking how this affects the public debate. Noting that AI is just a tool, but a tool with human attributes thereby distinguishing it from other tools. Provides promise that AI will become the master tool 6/ #AIEthics

In this way AI could solve all our problems which is therefore the apotheosis of the tech utopian debate. The flip side is that AI will develop own needs and goals, and that these will go against those of humans 7/ #AIEthics

@LeverhulmeCFI has identified 4 narratives, both utopian & dystopian. To be published in @nature in NY. These extreme narratives can help inspire + get people to focus on risks. However, also can skew debates so need other stories reflecting real tech capabilities 8/ #AIEthics

Now up is Kate Coughlan from the BBC. Why is BBC interested in AI? - see picture 9/ #AIEthics

Over 80% of people feel they understand AI, but only 10% believe that they can influence it. Highlights fear that it well be developed to benefit specific groups, but public will bear risk 11/ #AIEthics

Next steps for BBC after this survey, so they are asking lots of people what they think, looking for a representative understanding of the UK public and beyond 12/ #AIEthics

Now it's @sarahcastell going to reflect on her extensive dialogue work on AI and wider science and tech 13/ #AIEthics

Four reflections: Some groups resist learning more about some aspects of what AI can do, for example where it might be able to show empathy; people don't believe that their behaviour is predicted by data science, people like to feel they are an individual; 14/ #AIEthics

One of biggest challenges we face in 21 century she says, 'How do we understand ourselves as a data science human' ie as a data science subject and what AI can predict using that. ie people don't like being boiled down to data points, misses our stories/ context 15/ #AIEthics

dialogue always looks at costs and benefits, and how public trade them off. Works well she says, but is pretty rational. People start at point of ends justify means, so if AI can improve things. However, nature of benefit, in terms fairness, equity etc, more nuanced 16/ #AIEthics

People accept benefits, but are worried about missing the human connection, even if less effective. Ie might prefer to meet with a consultant who isn't as effective than be diagnosed by AI 17/ #AIEthics

Now it's @HughMilward. 'Trust in the technology is existential for us as a company [Microsoft]. <risks trust becoming an instrumental thing though. Focus needs to be on creating trustworthy institutions governing, developing, researching, regulating and deploying 18/ #AIEthics

Microsoft going to build and embed a set of ethical principles into their work. And this must enfranchise the public< we'll know if this has been done if M/soft stops doing something it wants to if public say they don't want it 19/ #AIEthics

Microsoft also wants regulation limiting use of AI. Finally it has a focus on training and education. Both tech, but also uniquely human skills such as empathy etc. 20/ #AIEthics

Finally up, Jess Montgomery. She's going to look at lessons from other technologies. History of Royal Society shows we've seen lots of waves of tech change and hopes for its impact on society 'We've always had hopes for what technology might do for us' 21/ #AIEthics

Dialogue on climate change, technical issues bleed into wider questions about scientific risk, and challenge of words (risk) meaning different things to different people. This makes public debate v difficult 22/ #AIEthics

She's looking at nuclear and GM too. Notes that unrealised hype causes problems in public debate, as does questions of/ concerns about who benefits and loses from the tech in question. 23/ #AIEthics

She notes, in context of gene editing, that public see significant benefits, but have hopes &fears. Key is context of the decisions made about tech. This mirrors findings of @sciencewise research, public often conditionally accept tech but... ukri.org/files/legacy/s… 24/ #AIEthics

my q to panel is whether data and AI is the right focus for a wide, diverse, sustained public debate involving multiple publics, or starting from the public's visions for the future, the problems they face and role that tech might play in that... 25/ #AIEthics

... panel answers agree right way round. @sarahcastell notes that questions of power are too often missed out of questions asked to public about AI [and more generally]< hearty agreement with this from me 26/ #AIEthics

@sarahcastell noting that any effective public engagement must have a real impact on a decision and not just develop some 'airy fairy' principles. 27/ #AIEthics