FAPR '96: The International Conference on Formal and Applied Practical Reasoning

Bart Verheij (1)

In Bonn, June 3 to 7, 1996, the International Conference on Formal and Applied Practical Reasoning (FAPR '96) was held. In my opinion, this conference was of unique import.

Not just because of the fine quality of the usual conference elements, such as tutorials, workshops, invited lectures, and paper presentations. The import of this conference was its stimulating goal: bringing together researchers who otherwise do not meet, but who should meet.

One of the problems of modern scientific research is that it almost inevitably requires a high level of specialization. The advantage is that the resulting ivory towers, in which most individual researchers spend their scientific lives, can more efficiently be built higher; the disadvantage is that these towers do not simply add up to an integrated scientific village.

For me, an argumentation researcher, mostly living in the ivory tower called 'Artificial Intelligence', this meant that I had to focus on the YAAICs (YAAIC for Yet Another Artificial Intelligence Conference). Although I do not deny the importance of those conferences, they often gave me the feeling of being trapped within too narrow limits. Allow me to explain this feeling.

Research on argumentation can have rather different aims, for instance:

Of course, if one does research on argumentation, one does normally not have only one of the aims above. For instance, even if one is mostly interested in designing intelligent computer programs, one can be inspired by actual human argumentation, and be led to the enhancement of one's initial model of argumentation. Nevertheless, research is often differently biased towards each of the mentioned aims of study.

As a result, the biases of individual researchers or research programs can be depicted in a triangular 'biases diagram'. The three corners of the triangle correspond to the three mentioned aims of study, and are suggestively labeled 'Minds and humans', 'Machines and programs', and 'Theories and models'.

Each researcher or research program has its own place in the diagram. The closer a dot is to one of the corners, the more the corresponding researcher or research program is biased towards the aim of study of that corner. (2)

The biases triangle can enhance awareness of the relative position of individual researchers or research programs. For instance, most research at YAAIC seems to be biased towards the top and right sides of the triangle. The left side bias is virtually absent.

FAPR '96 is an excellent example of a conference that attempts to overcome this limitation. There were researchers of all three biases. Let's for instance consider the four invited speakers. Jaako Hintikka, mostly working on mathematical logic(s), represented the upper corner of the triangle. Doug Walton, investigating the nature of fallacies, represented the lower left corner. Ray Reiter, now focused on applied research, aptly called 'cognitive robotics', represented the lower right corner. Lotfi Zadeh, the patron of fuzzy logic, positioned himself somewhere near the center of the triangle by explicitly claiming that fuzzy logic can narrow the gap between different kinds of models of argumentation.

It was interesting to see that the different biases of researchers (and their corresponding backgrounds) were sometimes even obvious for the casual observer: Several researchers representing the lower left corner gave lectures in the literal sense, by reading aloud, in the classical scholarly tradition; researchers of the lower right corner gave slick presentations, as if for a funding audience; researchers of the upper corner sometimes fell in the common trap of mainly focusing on technical details.

I will continue with a brief description of some of the talks that I attended at the conference. All of them contain an element of integration.

Doug Walton, in his invited lecture, started by pointing out the difficulties one encounters when teaching logic, not for its own sake, but as a tool for the analysis of reasoning in practice. He argued for a reorientation of logical researchers, who should acknowledge the fundamentally pragmatic and goal-driven aspects of reasoning. He emphasized the need of what he called 'applied logic', as an interdisciplinary subject. He suggested the development of software to teach critical thinking. Interestingly, and apparently independently, on one of the workshops of the conference, devoted to Computational Dialectics, (3) Thomas Gordon presented the ZENO project, in which Web software is developed for the mediation of argumentation. Ron Loui's Room 5 project has a similar goal, focused on legal argumentation.

Raymond Reiter discussed his current research on cognitive robotics (and did not even mention his default logic, that, as Dov Gabbay pointed out, comes up in any discussion of nonmonotonic logic). The goal of his research is nothing less than the implementation of agents that reason, act, and perceive in a changing, largely unknown world. Reiter explained that he did not want to choose between the seemingly opposite 'good old-fashioned' and 'cockroach' AI techniques (as they have been called by some), but was striving for an integrated framework. He presented a programming language called GOLOG, based on an enriched version of McCarthy's situation calculus, and specialized for the efficient generation and use of frame axioms.

Erik Krabbe, who was invited speaker at the workshop on Computational Dialectics, discussed commitment and retraction in argumentative dialogue. As an example, he gave the following dialogue: 'Do you ever answer a question?' 'No, I don't.' 'Is that your answer?' 'Yes.' 'So, you answered a question!' 'Well, sometimes.' He gave a taxonomy of types of propositions, with regards to commitment and retraction. For instance, an assertion by a person is a statement that must be defended if another person doubts it. A concession is a statement that one does not have to defend, but can also not doubt. Krabbe discussed how the more restricted Lorenzen-type dialogue systems, focused on logical rigor, and the more free Hamblin-type dialogue systems, focused on fallacies and errors of reasoning, could be integrated.

Frans van Eemeren, one of the members of the panel on the future of argumentation theory, discussed the perspectives of pragma-dialectics, the argumentation theory developed by Rob Grootendorst and himself. He argued that the normative and descriptive aspects of theories of argumentation should not be separated. He discussed several research projects, for instance, one on metaphors of argumentation. Because these metaphors can be as different as battle, journey, and solid construction, the different perspectives they provide influence actual practice of argumentation. He finished by mentioning the plan to start the formalization of the pragma-dialectical approach, which can lead to the integration of formal and informal research of argumentation, that developed separately for so long.

As a final example of mixedly biased research, I mention Carl Vogel's interesting paper on human reasoning with defaults of the form 'Xs are typically Ys'. Subjects were shown sets of such defaults in graphical or in sentential form, and then asked whether some conclusion could be drawn. The thus collected psychological data was compared with existing formal inheritance systems. It turned out that the data supported positive chaining, i.e., from 'Xs are typically Ys' and 'Ys are typically Zs' conclude 'Xs are typically Zs', but also negative chaining, i.e., from 'Xs are typically not Ys' and 'Ys are typically Zs' conclude 'Xs are typically not Zs'. The latter is usually considered incorrect. Vogel points out that it is not surprising that people make mistakes with negative default information, since this is also the case for negative strict information, as is well-known. He argues that AI researchers should take this kind of psychological information into account, since for the development of reasoning systems it is important to know what people actually do, even if that is considered to be wrong.

To conclude this report, the FAPR '96 conference was a succes, especially because of the inspiring broad perspective on argumentation it provided. Not only the good conference contents, but also the sunny weather and the nice terrace at the conference site eased the mingling of so differently biased researchers. Dov Gabbay showed to be an enthusiastic organizer, devoted to the goal of integration, by repeatedly suggesting outlines of combined papers for several of the invited speakers.

Since in my opinion the FAPR '96 conference fills a gap, it certainly deserves a follow-up. There is one planned for next year, combined with ECSQARU (focused on uncertainty). I hope my fear will not come true that the pleasingly broad FAPR then becomes yet another YAAIC.

Proceedings

Gabbay, Dov M., and Ohlbach, Hans Jürgen (eds.) (1996). Practical Reasoning. International Conference on Formal and Applied Practical Reasoning (FAPR '96; Lecture Notes in Artificial Intelligence, Vol. 1085). Springer, Berlin.

Notes

(1) bart.verheij@metajur.rulimburg.nl, http://www.cs.rulimburg.nl/~verheij/

(2) The triangle has barycentric coordinates. One can think of the triangle as the set of points (x, y, z) in the plane x + y + z = 1, such that 0 <= x <= 1, 0 <= y <= 1 and 0 <= z <= 1. For instance, the corners of the triangle are the points where one of the coordinates is equal to 1. The sides of the triangle are the points where one of the coordinates is equal to 0. The values of each of the three coordinates represent the bias level towards one of the corners.

(3) More information on this workshop can be found on the World-Wide Web at http://nathan.gmd.de/projects/zeno/fapr/programme.html.