The dangerous seduction of AI in public policy

Some important and dire warnings about the way the current federal government is going about its business making policy on artificial intelligence. Or perhaps, more accurately, how it is letting AI make policy on artificial intelligence and other things.

One of the dangers of using AI for public policy can be seen in this article, published in Slugger O Toole, a blog on Northern Ireland issues. The author uses a context-free ChatGPT definition of reconciliation and then asks the question “how are we to practice the vague, abstract notion of reconciliation?” It sent alarm bells ringing in my head. Here is the response I wrote:

You’ve lost me at using a ChatGPT definition of reconciliation. Defining reconciliation is as much a part of reconciliation as enacting practices and structural reforms to sustain it. The context of reconciliation matters tremendously. In South Africa it was a crucial decision to take as a majority population finally assumed power once the state became democratic. The potential for terrible violence was present and the way the majority took power mattered, hence the Truth and Reconciliation Commission. Here in Canada reconciliation is a socio-historical imperative, to transcend centuries of colonial policy that have been called genocidal by the government’s own inquiries. But it is also a legal concept in which Aboriginal rights and title, which are recognized in Canadian law, need to be reconciled with the Crown’s rights and title and interest in lands. If you will excuse my directness, using a ChatGPT definition of reconciliation in the context of real, meaningful and very specific needs is not only lazy writing, but potentially dangerous and destabilizing public policy. Reconciliation is not a goal, it is a direction of travel and, perhaps, an evaluation criteria in which future generations can say: “they took actions which reconciled us.” It requires deep sensitivity to the dynamics of the present, and a commitment to that ongoing direction in the context of the unique affordances of time and opportunity which people might pursue together.

Seems to me that the essence of democratic deliberation is to work out what we can do together given the current state of play. I’m sure AI has it’s usefulness in deliberative practice in democracies, but our federal government’s blind pursuit of it as an engine for economic growth and Hughie Beag’s uncritical use of it to set a public policy agenda that ignores the reality of the society in which he is embedded show to of the dangers from succumbing to it’s seduction.