From Artificial to Ambient Intelligence

In this project, a reconstruction of the continuities and changes between the research programmes of Artificial Intelligence (AI) and Ambient Intelligence (AmI) shall be undertaken. The inquiry will focus on the implicit and explicit models of the environment in which the systems in question operate. While the importance of the role of the environment to a system`s capabilities and behaviours has been recognised in the AI debates since the early 1990s, AmI, here understood in an inclusive sense, offers a markedly different take on this topic. In seeking to unobtrusively modify the environment for perception and interaction, AmI systems provide an instructively indirect route to an understanding of how the human mind is placed in, and interacts with, the world. In a first step, a brief historical-systematic overview will be provided of the ways in which the environment did and does assume importance in various AI research fields and in the emergence of AmI. In a second step, an empirical inquiry into concepts and practices of modelling the environment in AmI research will be undertaken. Again, a comparative perspective between different approaches will be adopted, with a view on implications of the models in question on human-environment relations. In a third step, the empirical findings will be discussed on the background of critical reconstruction of three partly related theoretical accounts: the ecological approach to visual perception; the concept of "niche construction" in evolutionary biology; the "extended mind" hypothesis in the philosophy of mind. All of these theories conceive of environments as specific to an organism`s goals, dispositions and behaviours rather than merely its physical surroundings. Matching these theoretical accounts against the empirical findings, the questions for analysis will be: As one central aim of AmI is to make features of human environments adaptive to human goals, dispositions and behaviours, what kind of counterpart of an interaction will such a modified environment be? How is an AmI system supposed to keep track of human behaviours? How does it allow for humans to keep track of its purposes in turn, given that it lacks many of the cues of human-like intelligence that had been the mark of classical AI? These questions, although arguably relevant to AmI research, are not technical questions in the first place. In this project, they will be addressed as questions of model-building in the philosophy of science and as empirically rooted questions in the philosophy of mind. Ideally however, the results of this inquiry will loop back into the burgeoning field of AmI research.

Ambient Intelligence or Intelligent Environments are labels for computer-based technologies that, despite common origins, provide a striking contrast to the research programme of Artificial Intelligence (AI). Where classical AI sought to directly simulate human thought and action, aiming at some degree of human-likeness, Ambient Intelligence (AmI) systems are designed to simulate some key properties of human environments in real-time, adaptively and partly autonomously. They shall hence facilitate perception of those properties and offer novel opportunities for interaction. Where human thought was mostly conceived of as abstract, symbolic and disembodied in classical AI, AmI provides some evidence for recent views in cognitive science and philosophy of cognition as being an embodied activity that is embedded in perceivable environments and extends into them by means of artefacts. Thus conceived, AmI technologies are in the business of tracking natural information for perception in human environments. In this study, two basic modes of selecting and presenting that information are distinguished: Firstly, there are artefacts that provide information concerning a set of conditions in a concrete environment that the user could, factually or hypothetically, also perceive when using natural cues. Such is the case for the majority of Augmented Reality technologies. An example are driver assistance systems that direct a motorists awareness to relevant conditions, including his or her own focus of attention, so as to initiate an appropriate reaction to a traffic situation (which is a solution beneath the level of the self-driving car). Secondly, there are artefacts that provide information on a virtual environment that is integrated with what the user perceives in his or her real environment. Such is the case for Mixed Reality technologies. An example are gaming applications that place interactive virtual objects in real environments, where these objects are part of the gaming environment while sometimes also involving properties of the real environment (as in Pokémon Go).Either way, it is concluded, environments are modified for the respective users in a certain way: if cognition is an embodied activity that is coupled with a perceivable environment, and if that environment is enriched with embedded artefacts that adaptively and partly autonomously provide information for human activities, such an environment presents itself as a seamless, smooth and cooperative surface for those activities. Without necessarily being perceived as having been artificially modified in a certain way, it will affect the choice and pursuit of human activities. The nature of such an environment may come to be perceived as a matter of course until, that is, some interaction fails. To the extent that this peculiarity of intelligent environments gives rise to ethical issues, these will provide an instructive contrast to the question of the relation between human nature and machine simulation in classical AI.