Eliminate the Interface

The design of an interface between two systems depends upon the way in which the systems are understood and characterized by the designer. Digital computation and the concomitant understanding of cognition as a computational process allowed the interface to be developed as a surface of contact between two symbol-processing systems. This paper suggests that as embodied, situated and to an increasing degree autonomous artifacts are developed and as cognition becomes understood as a dynamical process that new forms of interface will be required. It is suggested that direct actions in the world are the most robust means of means of communication. This eliminates the interface as an object or artifact of design and concentrates efforts on crafting the interaction that takes place between autonomous agents. Our theories of computation have evolved and interacted with and informed our understanding of cognition. There is a strong tendency to see each in terms of the other. This reciprocal relationship has had a profound impact on the kinds of interfaces that have been developed to mediate between them. Transformations in this relationship will have corresponding implications for the kinds of interfaces that need to be developed and the way in which they are implemented. The shift from computational to dynamical models of cognition has tended to emphasize both the embodiment and situatedness of the agent and to bring the environment into an active role in the cognitive process. Simultaneously, the widespread distribution of embedded processors has allowed a measure of responsiveness and interactivity to be generated from within the environment itself. This condition requires that notions of interface be reconsidered. The development of digital computation had symbolic logic as its foundation. Logic is an attempt to codify the proper structure of thinking independent of its content or context. This rule-based manipulation of symbols has Copyright ©2000, American Association for Artificial Intelligence (www.aaai.org). Allrights reserved. enormous power and flexibility. Instantiations of this process have been shown to be equivalent whether implemented by mechanical, electrical, hydraulic, or other means. It was widely supposed that biological, and in particular, human cognition is based on computational principles as well. The independence of process from substrate suggested that software could be considered apart from hardware and so also the mind from the body. Hardware independence lent plausibility to the project of extracting and transferring human cognitive capabilities to machines. Wearn (1998) notes it is common to consider cognition only within the context of the mental processes of an individual using as a model the human as an information-processing machine. Interfaces within this perspective are surfaces of contact between two kinds of information processors -surfaces upon which the symbolic tokens reside and are manipulated. Text, graphic and speech interfaces attempt to capitalize on the richness of the modes of communication that have developed in the human social sphere. They suggest a humanness of the intelligent machine that is consistent with the traditional goal of A.I. the duplication of human intelligence. They also accept an abstraction of humanness that is consistent with the cognitivist approach. But, the computational model of human cognition, especially in that it relies on notions of representation and logical operations, has been effectively challenged. We have come to understand ourselves as much more than this. But if the computational model is no longer convincing, then alternative interface strategies will need to be developed. There has been a consistent trend toward the diffusion of computing resources into the environment. Initially all computing machinery was concentrated in a small number of machines of legendary size at work on problems of scientific or political import. The development of personal computing has put a large number of (for their time) relatively moderate processors to work on issues that are related to the individual or corporate user. While few of the problems addressed by these machines individually have the social significance of those that engaged the first machines, together they have produced significant social and cultural transformations. At the same time, the number of information processing machines of all sizes is dwarfed 4 MAICS-2000 From:MAICS-2000 Proceedings. Copyright © 2000, AAAI (www.aaai.org). All rights reserved.