Much thanks to Dave Snowden and the Cynefin team for inviting me to be one of the Eagles at the Cynefin Exploratory, Balancing Human & AI Reasoning in a Complex World, held at Trinity College Dublin on 1–2 July.
The two-day Exploratory was designed to bring together researchers, practitioners, policymakers, and organisational leaders to explore the evolving relationship between human reasoning and AI in situations characterised by complexity and uncertainty. Through a mixture of provocations, dialogue, and collaborative inquiry, the event encouraged participants to move beyond simple narratives about AI and engage with the challenges and opportunities these technologies present for organisations and society.
As one of the event’s Eagles, my role was not to provide answers but to offer a provocation—an idea intended to stimulate discussion and critical reflection. My contribution centred on a simple question that I believe is largely missing from current debates about AI. It was this:
Which AI, for whom, and under what conditions?
Here is my argument:
Most debates about AI become too broad too quickly. We ask whether AI is intelligent, creative, biased, dangerous, transformative or beneficial, but those claims only make sense once we specify three things: which AI, for whom, and under what conditions.
By ‘which AI’, I mean that LLMs, machine learning and predictive systems, recommendation systems, decision-support tools, autonomous agents, and robotics are not the same thing. They have different histories, capabilities, users, risks and consequences.
By ‘for whom’, I mean there is no generic AI user. A civil servant using a decision-support tool, a teenager shaped by TikTok recommendations, a scientist using machine learning, a student using ChatGPT, and a warehouse worker interacting with robotics are not encountering AI in the same way. The benefits and harms are unevenly distributed.
By ‘under what conditions’, I mean more than context. Conditions force us to specify the task, setting, stakes, expertise, training, access, free versus paid systems, oversight, accountability, laws, language, culture, institutional pressures, and consequences of failure.
So, the point is not to ask whether AI is good or bad in general. The point is to ask what kind of AI, used by whom, for what purpose, under what conditions, and with what consequences.
Here is a link to the PDF of my presentation
HERE IS A LINK to my artwork
