Inclusion is more than who is in the room – it’s about who has the power

by Professor Claire Paterson-Young and Professor Marion Oswald (both PROBabLE Futures team)

If you spend any time on LinkedIn during conference season, you might conclude that the sectors, disciplines and communities involved in AI have solved their diversity problem. The photographs are persuasive: packed auditoriums, enthusiastic claims about “bringing together the brightest and innovative minds in AI”, and the increasingly gender-balanced panels. AI-related communities have become very good at photographing diversity. The harder question is whether these communities have become any better at sharing power. Representation has improved, and that matters. But visibility is not authority, and a balanced conference programme can leave the distribution of power almost entirely untouched. Being visible in the room is not the same as having the authority to shape what happens inside it. Indeed, our growing ability to demonstrate diversity may make this distinction harder to see. Once an event looks inclusive, asking whether it actually distributes influence can seem unnecessary. It is not.

Counting women on conference programmes tells us who is present. It tells us far less about who is listened to, whose expertise is treated as authoritative, and whose ideas are allowed to set the direction of AI. This distinction may seem subtle, but it goes to the heart of how knowledge, authority and influence are produced within the world of AI. Conferences are places where research is presented and ideas are exchanged. However, they are also places where careers accelerate, collaborations begin, funding relationships develop and disciplinary and policy priorities are quietly established. They determine who becomes recognised as a leader, whose research is repeatedly cited, whose expert advice is heeded and whose perspectives shape the next generation of innovation, policy and change. Conferences do not simply reflect the AI community; they help construct it. They help decide whose knowledge counts as AI expertise, who becomes recognised as a leader, and who acquires the authority to define what AI should become. That authority travels well beyond the conference venue. Today’s keynote speaker can become tomorrow’s media expert, policy adviser, grant collaborator or member of an influential committee. The people repeatedly placed at the centre of AI conversations accumulate the credibility that brings them back to the centre again. Conference visibility is therefore not merely symbolic: it can become a mechanism through which authority is accumulated and reproduced.

Counting bodies tells us who entered the room. Positionality asks where they were placed once they go there. Are women positioned as experts whose ideas define the field and as leaders who set agendas, or as visible evidence that an event has met its diversity ambitions? Those two things are not the same and the concept of tokenism helps explain why. Nearly fifty years ago, Rosabeth Moss Kanter argued in her 1977 book Men and Women of the Corporation that tokenism was not simply a matter of numbers. Tokens can be highly visible while possessing relatively little power. We should have learned this lesson by now: visibility without influence is not inclusion. Yet nearly fifty years later, we are still remarkably willing to use visibility as evidence that inclusion has been achieved.

A conference can therefore look impressively diverse while quietly preserving a hierarchy of expertise. Women may appear across the programme yet remain less visible in prestigious keynotes and the sessions presented as defining the future direction of AI. Our own expertise lies in responsible AI, ethics and law enforcement. The problem is not that we are invited to speak about these subjects. It is that these areas of expertise can themselves be positioned as peripheral to the supposedly more exciting business of AI innovation. This creates a false hierarchy of AI knowledge: technical development occupies the centre, while law, ethics and responsibility are too easily positioned as conversations about the consequences. Yet these fields do not simply respond to AI after it has been built. They shape what can be developed, how systems should be designed and deployed, whose interests are protected and what institutions and societies are prepared to accept. Responsible AI is not peripheral to AI innovation. It is part of determining what responsible innovation can be.

This is where positionality becomes particularly important, because the reception of knowledge cannot be separated from perceptions of the person producing it. Knowledge is not evaluated independently of the individual; expertise is interpreted through assumptions about gender, race, age, institutional affiliation and professional status. A conference, therefore, is never simply a collection of presentations. It is a social environment in which credibility is constantly negotiated. Many women working in the sector recognise this negotiation and the fact that it can appear in surprisingly ordinary ways. Being assumed to be the communications lead rather than the principal investigator; being invited to speak about “women in technology” rather than the substance of our research; or finding that our expertise in responsible AI, ethics and law is welcomed while simultaneously being positioned as commentary on technological development rather than expertise capable of shaping its direction. These moments rarely dominate a conference, which is precisely why they are so easy to dismiss. Their significance lies in their accumulation. Experience enough of them over several days, across enough conferences, and they begin to shape not simply how an event feels but who is recognised as authoritative within it. We have both experienced such tokenism on multiple occasions. Representation addresses who is present. Positionality asks who is listened to, who is remembered and who leaves having influenced the direction of the field.

Sara Ahmed’s work on institutional diversity shows how easily diversity can become evidence that an institution uses to demonstrate its own success. The gender-balanced AI conference photograph risks doing exactly that. It is photographed, posted and celebrated; the image becomes proof of inclusion, while the awkward questions about who set the agenda, controlled the discussion and left with greater authority remain unanswered. Representation is visible and countable. Power is much harder to photograph. Conference organisers are asking the wrong questions. Instead of asking whether women appeared on the programme, ask who delivered the keynote. Who chaired the sessions with the greatest prestige? Who decided which research themes deserved the main stage? Who entered the informal conversations where collaborations and funding relationships were formed? Who left with more influence than they arrived with? Who was allowed the final word? And who appeared in the diversity photograph?

These questions are more difficult to answer than counting speakers, but they reveal the true extent of the conference impact. This matters because conferences distribute more than speaker slots. When those forms of power remain concentrated, a diverse programme can disguise an unchanged intellectual hierarchy. So by all means count the women on the programme. Then count the keynotes. Count the chairs. Count who is positioned as defining the future of AI and whose expertise is treated as commentary upon it; who gets quoted, funded and invited back; and whose ideas survive after everyone has gone home. If those numbers tell a different story from the conference photograph, stop calling it progress. Representation gets women into the room. It does not, by itself, redistribute authority. Inclusion begins when women share the power once they arrive.

For more information on RAi UK’s Keystone project PROBabLE Futures:

• https://research.northumbria.ac.uk/probablefutures/