Computational Autonomy and Information Systems Open Reference

From Vision to Paradigm and Architecture

Erik Sandewall


The term "CAISOR" stands for Cognitive, Computational Autonomy and Information Systems. Besides being the codename for this website and for the sequence of research groups that have provided its contents, we also use it as the name for an alternative software paradigm. Let us briefly discuss how it relates to some current research directions.

The term 'computational autonomy' is a standard term, in particular in the agent-oriented programming community, for attempts to design software systems in such ways that they make active choices about how to achieve given goals, instead of just executing a given algorithm. In our particular interpretation of that term we emphasize life-cycle autonomy rather than autonomy within the short-term execution of a particular run of the computer or the software, which is the usual focus. In other terms, we take particular interest in the kind of autonomy where a software system manages its own version and variant control, its own documentation, its own distribution and update, and so on.

The term "information system" in our umbrella phrase does not indicate a particular application. Instead, it represents our view that a software system is an information system, in the sense of a self-managing structure consisting of programs and data that exists over a long period of time, and that has well-defined relationships to two (possibly overlapping) sets of human agents, namely its developers and its users.

The term "cognitive" in the umbrella phrase, finally, represents our view that self-sustaining, autonomous systems that are software systems and information systems at the same time, will be natural users of a number of techniques from artificial intelligence. We think in particular of techniques for reasoning about actions and change, planning, and other kinds of knowledge representation, but also of computer learning in the A I sense.

If these notions are applied consistently then the resulting system is going to be quite different from systems using contemporary software technology. In that sense this is a vision for a new and alternative way of designing software. We propose to make it tangible in two steps, by first formulating it as a software paradigm and then, more concretely, as an architecture. The acronym CAISOR will be used to designate the paradigm, and the acronym CAISA (Computational Autonomy and Information Systems Architecture) for its concretization as an architecture.


Posted on 2005-02-20 as part of the CAISOR website. [Version history].