Diaspora of a Mathematics of Argument


There has been a strong anti-intellectual bias in AI since "AI Winter"
and the ensuing "Halpernization" of the field.  

One of the most significant possible shifts of paradigm, however, is
the shift from logic to argument.  It is an intellectual result of AI
in the past decade and an idea which computer science exports to the
social sciences.  Work on argument has been quietly proceeding in
pockets where AI is still concerned more with concepts than tools.

I will trace the rise of process-based models of reasoning which
culminate in the current models of argument:  Keynes's influence on
Wittgenstein, then through Waismann, Austin's on Hart, then Hart's new
foundations for a half-century of social science through Rawls,
Toulmin, Barry, Raz, Ladd-Chisholm-Pollock-Sosa, Gauthier-Searle,
Wellman-Feinberg, then AI's flirtation with nonomonotonic reasoning,
PROLOG, defeasible reasoning, and negation-as-failure.

I will inventory some technical issues surrounding syntactic
vs. explicit defeat, meta-argument, and the shifting burden of proof.

I will discuss how argument affects the foundation of probability
(Reichenbach-Kyburg-Pollock), the foundation of decision
(Simon-Savage-Shafer-Tversky), models of legal reasoning
(Cardozo-Levi-Raz-Sunstein-Bayles), models of belief revision and
obligation (Hohfeld-von Wright-Alchourron), non-game-theoretic models
of negotiation (Fisher-Pruitt-Sycara), and the informal logical
traditions in rhetoric (Toulmin-Perelman-Walton).

Finally, I will discuss possibilities for an argument-based description
of fairness and an understanding of protocol-based gameplaying as
computation.


----------------------------------------------------------------

Ronald P. Loui is Associate Professor of Computer Science at Washington
University and an affiliate in the Legal Studies Program.  

This year he is co-editing a special issue of AI Journal on AI and Law,
was conference chair of ICAIL (the international conference on AI and
Law), and will be conference chair of SEP (the annual meeting of the
Society for Exact Philosophy).

He will be happy to discuss other research projects on:  gnu's
independent co-malloc and gawk's average hash chain length,
decision-theoretic approaches to virtual memory management, FPGA's,
purely probabilistic negotiating agents, real-time risk-defensible
object recognition, and introductory programming through cgi.