Entre la transparence et l’intrusion des machines intelligentes

Abstract

Smart machines are devices endowed with artificial intelligence (AI) applications. Some smart machines operate in tight coupling with human users within human-technology-world loops. This paper explores the phenomenological landscape that accompanies smart machines of this sort. Drawing centrally on Heidegger’s analysis of everyday encounters with equipment, I map out two phenomena, namely transparency and intrusion, that together give shape to this landscape. Transparency occurs when a device is manipulated with skilled expertise in a hitch-free manner, and, as a consequence, disappears from the conscious experience of its user. The user sees through the device to the task at hand. Smart machines may sometimes be transparent in this sense. Intrusion, by contrast, occurs when a device is present as an entity in our experience. Roughly speaking, the more intrusive the device is, the more we have to think about it. Smart machines may sometimes be intrusive in this sense. Having characterized these phenomena, I confront a design-problem – one with significant social implications – that thereby comes into view. Transparency-in-use is often treated as a mark of good design. However, there is compelling evidence that smart machines will sometimes make classifications that diverge from user judgments. Under conditions of transparency, user activity may be affected negatively by these divergent classifications in ways that the user won’t notice. I develop the beginnings of a response to this problem, a response that requires us to find a way of encountering smart machines that is located between transparency and intrusion.