This description is obviously not finished yet. It isn't at the top of my list of priorities either, but if you're interested please ask me if you want any more information.

My original reason for embarking on this line of investigation was deep disgust at the general irrationality of public debate on issues ranging from politics to religion ( as in ridiculous to sublime ); the specific issue which set me off was a "discussion" on the Treaty of Waitangi conducted in the correspondence column of the New Zealand Herald a long time ago.

It seemed to me that there must be better ways, and I wondered whether there was some way in which help might be drawn from computers. My first thought was that something in the nature of artificial intelligence was a promising way to go; the ( travesties of ) arguments put forward were clearly supposed, by their originators if by no one else, to be models of logic and clear thinking, and logic is one thing we can do reasonably well with computers. Eccles naturally came to mind.

Further study suggested that this idea, while it would be fun, would only rarely be useful. The problems turned up before the logic was extracted from the surrounding verbiage, and the extraction itself would be a non-trivial exercise in language understanding, and far from easy. Instead, the real problems were about interpretation. A and B could argue about the significance of statement X, with each obviously attaching quite different meanings to X, and apparently not noticing. Further, the inferences drawn from X, on analysis, clearly implied that A and B were making rather different assumptions which they were using in the argument without stating them explicitly.

It therefore seemed that a better line of investigation would be to produce some means of making clear the steps which people had taken in their arguments, and the assumptions which had been made. This is not so much a task for artificial intelligence as for clever databases, and that did seem promising for computers - they are, after all, the best machines we've ever invented for reliably doing routine and laborious tasks with large quantities of information, and some at least of the problem seemed to lie in that area.

The first task is to design and implement a system which will record an argument, initially in the original words, then provide means for the argument to be annotated. The annotations would be, first, the author's interpretation of each step in the argument in terms reasonably suited to logic rather than political debate or rabble rousing, then, second, the author's explanation of how each step is derived from the previous one. An objector to the argument can then take issue with the steps, now clearly laid out and stripped of their rhetorical decorations, and state alternative interpretations. This process can go on for as long as the parties will put up with it, which might at least keep them too busy for further public pronouncements.

Some students have put in some work on debate moderation from time to time : Roy Davies; Peter de Vocht.

More if you want it :

WORKING NOTES : AC70 ( Schemata and debates ); AC74 ( Treaty of Waitangi ).
TECHNICAL REPORT : 55 ( Ideas and a manual simulation ).

Go back to me ( Alan Creak, in case you've forgotten );
Go back to Computer Science.