Autonomy, autonomous systems and robots

In the context of AI, the term "autonomous" is mostly used to describe a computer-based system that performs complex tasks and to do so independently of human influence or control (to varying degrees). The individual steps in solving tasks are no longer prescribed or monitored by humans, but are chosen solely by the machine. For example, when a so-called autonomous vehicle is given a destination, the machine determines by itself the route, the speed or the driving manoeuvres it must execute according to the situation on the road.

The autonomous characteristics of a system are often understood as the next step towards automation, yet automatic systems originally were essentially about thinking ahead to all contingencies in the programming of a system. They can perform complex tasks, but they still follow the rules that humans have established. In other words, automatic systems are pre-programmed. Autonomous systems, on the other hand, are able to abstract patterns from unknown data (“machine learning”). They are programmed in such a way that they can also adapt step by step to what has not been determined in advance. This gives them the ability to adapt and act on their own.

Autonomous systems are never fully self-driven (they are programmed by humans) or self-sustaining (they depend on energy sources and technical infrastructures). The term “autonomous system” is therefore somewhat misleading. The autonomous system has its own materiality (its hardware), which essentially determines what the system can do. Materiality also plays a major role in terms of the extent of human control: interfaces (screens, switches) determine how we interact with the system (speaking, gestures). These interfaces also offer us options for how to control the system – or can withhold such options from us. All these interactions determine to what extent humans are in the loop: Are humans part of the decision or not? When can humans intervene?

AI is not a purely technical or abstract idea, rather the very notion is embedded in complex social and cultural contexts. The concept of autonomy alone – a concept that has been constantly renegotiated for centuries – shows how much our understanding of AI is shaped by European intellectual history. In addition, societal expectations and fears around AI are strongly linked to fictions and imaginaries (e.g. the representation of robots in popular culture). These cultural images have a strong influence on how we experience AI. Cultural comparisons, the kind we draw in our research, therefore sharpen our view of AI.