Action
and Time
Most formal models used in traditional AI think of events as functions
from state to state. In NL work and commercial models, however, the
basic concept is of a process occupying an interval of time. We are
developing a formal framework which reconciles these opposing views
by integrating point- and interval- based temporal logics, thinking
of actions as occupying a 4-dimensional spatiotemporal `volume', and
separating deduction from prediction. This work is currently supported
by NSF and NIMA, and is part of a larger project (see below)
- A Catalog of Temporal
Theories, Pat Hayes; Tech report UIUC-BI-AI-96-01, University
of Illinois 1995 (Available online in two .ps files containing
chapters 1-3 and chapters 4,5 and appendices . )
Semantics
of Diagrams
Diagrammatic and hybrid representations are of interest for education
and interface design generally, and there has been much controversy
in conitive science on the nature of `mental images'. However, much
of the discussion is confused by the lack of a clear semantic theory.
We are developing a unified formal semantics for hybrid representations
which generalises both Fregean model theory of logics and a `similarity'
approach to semantics of diagrams. The central idea is a strict adherence
to the principle of compositionality.
Naive
Geographic Reasoning
In the tradition of naive physics, we are starting (5/1/97)
a project to develop a systematic axiomatic description of the useful
content of many qualitative spatiotemporal concepts used in geographical
thinking. The eventual aim of this effort is to create a geographic
knowledge base which can be used to support a useful concept of geographic
consistency. This work is currently supported by NIMA.
Nature
of Expertise
The history, philosophy, and sociology of science inform us that expert
knowledge is comprised of context-dependent, personally constructed,
highly functional but fallible abstractions. Experts can be understood
as performing a societal role that they were chosen to play as a result
of a constituency selection. In this way, we propose, evolutionary
epistemology, or more specifically a natural selection analogy, provides
a compelling basis for believing that some expert knowledge is more
than merely disposable cultural myths or highly local personal fabrications.
This perspective avoids both simplistic realism and complete relativism.
Computational
and Philosophical Foundations of AI and Cognitive Science
Work underway at IHMC is aimed at identifying and buttressing the
computational and philosophical foundations of artificial intelligence
and cognitive science. In particular, there is as yet no fully satisfactory
account of exactly what makes something into a `computer', and we
are developing a new approach to this question. Part of this effort
is a series of detailed critiques of various mistaken, though popular,
`proofs' that AI is impossible.
- About Artificial
Criticism: A Reply to Harry Collins. W.G.Barnes, K. Ford and P.
Hayes, Phi Kappa Phi Journal, Winter, 1995
- The Missing Link;
a reply to Joseph Rychlak. J. Adams-Webber, K.Ford and P. Hayes,
International Journal of Personal Construct Psychology, 1993
- Turing Test Considered
Harmful, P. Hayes & K. Ford, Proceedings of IJCAII-95, Montreal
- The Prehistory of
Android Epistemology (with C. Glymour and K. Ford), in Android
Epistemology, ed. Ford, Glymour & Hayes, MIT Press, 1995
|