Modelling the thing-in-itself – a philosophically motivated approach to cognitive robotics

In this paper, we will try to outline a few major questions that the design of artificial cognitive systems should tackle. The argumentation will be led from a philosophical point of view, reviewing some of the latest epistemological positions, namely from the 20 century. The notion of the thing-in-itself, i.e. the substantial core of an object, will be central in the discussion of abstraction in cognitive systems. A second important point is intentionality (in the philosophical sense): Doing something is always guided or biased by a task that is currently pursued. Last but not least the old question of symbol anchoring is to be posed. A preliminary system design for abstracting knowledge is shown that forms a possible implementation for tackling these aspects for the domain of robotic vision. 1. Why start from philosophy? It has become obvious in the past years that the study of cognitive systems, in which epigenetic robotics takes an important part, not only profits from but even requires an interdisciplinary approach. Philosophy has an outstanding role in the history of sciences, often termed as ”mother of all sciences“. Its unique approach can provide researchers with new perspectives and simultaneously retaining the big picture. The usual suspects of disciplines involved in the study of cognitive systems, namely biology/neuroscience, developmental sciences, psychology and artificial intelligence, have evolved from philosophy when sciences started to split into more detailed directions. Nevertheless, their roots are philosophical. Second, at least since the ground-breaking article ”Epistemology Naturalized“ by Willard van Orman Quine (van Orman Quine, 1969), natural sciences and philosophy started to move together. Quine rejected the notion of a ”prima philosophia“ (first philosophy) in the sense of a philosophy being in the elitist atrium of sciences, but rather stood up for philosophy being seen as a science like any other, simply arguing in a more abstract sense (Schneider, 1998). Last and maybe the most exciting point is, that philosophy has the tendency to pose the hard and uncomfortable questions. There are some issues in cognition that are spooking around but are tackled explicitly only in philosophy (e.g. the discussion of free will). As we will see later, exactly the problem of what makes a cognitive system cognitive is – sometimes with other terms – one of the oldest philosophical topics. An approach to cognitive systems from a philosophical epistemology seems to make an important yet arguable implicit assumption: cognitive systems are intended to be human-like or humaninspired. As we will show later on, this paper argues for a very restricted human-relation, however. Summarised, there are issues that philosophy might help to address. With this paper we try to start with the philosophical notion of the thing-initself in the context of abstraction in cognitive systems. Second, we tackle intentionality as the philosophical notion for the system’s bias due to the task currently pursued. Finally, semantics are to be looked at – a topic well-known to artificial intelligence. To put it in a nutshell, it is our firm belief that the look at epistemological positions has the ability to bring up new approaches and perspectives to an understanding of cognition in engineering as well. 2. On modelling cognitive systems The goal of this paper is to tackle some of the hard questions in the interesting research field of building artificial cognitive system. In (Ziemke, 2001), the main approach to cognitive systems is from radical constructivism, whose basic principles are: knowledge is not passively received but actively built up, the tendency of the system is guided towards viability and – probably most important – cognition serves the subject’s organization of the experiential world, not the discovery of an ontological reality. We are apBerthouze, L., Prince, C. G., Littman, M., Kozima, H., and Balkenius, C. (2007). Proceedings of the Seventh International Conference on Epigenetic Robotics: Modeling Cognitive Development in Robotic Systems. Lund University Cognitive Studies, 135.