How does AI affect human Autonomy?

Dr. Carina Prunkl, Senior Research Scholar, Future of Humanity Institute, University of Oxford

 Autonomy (autos= self; nomos = law) in the context of human beings refers to the capacity of self-governance or self-determination. This also implies that an individual’s actions are neither the product of external manipulation, nor imposition of external forces. Autonomy in this sense plays an important role in Western culture and is often considered desirable for the individual. When we speak about ‘autonomous systems’ in the context of artificial intelligence, we similarly refer to some sort of ‘self-governance’, but in contrast to human case, this ‘autonomy’ has little to do with acting true to one’s own beliefs, desires or motivations. Instead, it refers to the capacity of the system to learn and perform certain tasks without human guidance or supervision. A well-known example of such ‘autonomous systems’ are self-driving cars that navigate themselves through traffic to bring their passengers from A to B. But of course this type of ‘autonomy’ is not limited to the mechanical realm and we may easily conceive of virtual ‘autonomous systems’, such as virtual assistants that organize our lives by making appointments, doing (online) grocery shopping, taking notes, etc. By outsourcing seemingly trivial tasks such as driving and grocery shopping – not to mention some highly non-trivial tasks, such as those now performed by soldiers but that might at some point become automated – we are handing over more and more responsibilities to ‘autonomous systems.’ How will such a development affect our own autonomy? It is difficult to imagine that at least those of us who are somewhat indifferent to the joy of driving, will feel or be less autonomous by having a car that takes us to where we want to go faster and safer. This is at least in part because it is we, after all, who decide where to go and when. But what about when such ‘autonomous systems’ not only navigate us through traffic, but also through life? When they learn from our behavioral patterns, our preferences, our relationships, to make predictions about, say, what groceries we would like to eat next week? Here the situation is much less clear. Do we gain autonomy by not having to be bothered with boring grocery planning and shopping, and instead having time for the things we would really like to do? Or do we instead forfeit autonomy by not being the ones who make the choices about our nutrition, returning almost to the childlike state of not having to take responsibility for certain aspects of our lives? These are questions we urgently need to ask ourselves.