访谈范·本特姆教授.doc

Transcription

1 ( ) (Johan van Benthem) (Bonsall) 90 the Institute for Logic, Language & Computation ILLC the European Foundation for Logic, Language and Information Folli the Dutch Academy of Arts and Sciences the Academia Europaea the Institut International de Philosophie

2 the Student Revolution : M. Martin Löb 1977 University of Groningen 1986, M Modal Correspondence Theory the Handbook of Philosophical Logic 1985 Modal Logic and Classical Logic invariance for bisimulation Gestalt Switch D. Dov Gabbay

3 The Logic of Time , Essays in Logical Semantics generalized quantifiers 1980 J. Jeroen Groenendijk M. Martin Stokhof A.Alice ter Meulen Montague Grammar F. Frans Zwarts, J. Jon Barwise R.Robin Cooper E.Ed Keenan 3

4 situations repertoire semantic automata natural logic 1991 Language in Action Categorical Grammar - operating system a hierarchy of categories B. Theory of Types Lesniewski Ajdukiewicz Categorial Grammar 50 J. Joachim Lambek - - glue - substructural proof systems Curry-Howard semantics Cut Elimination formal language natural language 4

5 Handbook of Logic and Language R. government-binding theory situation theory game-theoretic semantics ( )

6 men M. Maarten de Rijke Y. Yde Venema Dynamic Turn belief revision theory discourse representation theory H. Hajnal Andréka I. Istvan Németi the Guarded Fragment Tarski semantics the Dutch National Research Organization NWO Spinoza grants 2 6

7 Logic in Action P. Paul Dekker Y. Computational Logic Lab for Language and Inference Technology Didactics and Dissemination Jan van Eijck CWI Utrecht Logic in Games the Dynamic Turn.John Nash 7

8 Logic in Games 2004 strategic equilibrium ( compositionality valid consequence (conditional) (knowledge) (belief) 8

9 9 philosophy-internal (Liège) 2001 National Cognitive Science Committee 2003 (University Professor), 5 5 M. Michiel van Lambalgen

10 [1] Johan van Benthem1983, The Logic of Time, Reidel, Dordrecht, (Synthese Library, vol. 156). Revised and expanded edition published with Kluwer. [2] Johan van Benthem, 1985, Modal Logic and Classical Logic, The Humanities Press, Atlantic Heights. [3] Johan van Benthem, 1986, Essays in Logical Semantics, Kluwer, Dordrecht, (Studies in Linguistics and Philosophy, vol. 29). [4] Johan van Benthem, 1991, Language in Action: Categories, Lambdas and Dynamic Logic, North-Holland, Amsterdam, (Studies in Logic, vol. 130). MIT Press [5] Johan van Benthem, 1996, Exploring Logical Dynamics, CSLI Publications, Stanford. [6] Johan van Benthem, , Logic in Games, electronic lecture notes, ILLC Interview with Johan van Benthem A Life of Logic, Language, and Information Johan van Benthem is University Professor of logic at the University of Amsterdam, and Bonsall Professor of humanities at Stanford University. In the 1990s, he was the founding director of the Institute for Logic, Language & Computation (ILLC), a joint venture of mathematics, computer science, philosophy, and linguistics, for studying the structure and flow of information. Also in that period, he was the first president of the European Foundation for Logic, Language and Information (Folli). He published some 10 books and 300 articles, is an editor of many journals, and supervised some 50 master s theses, and 50 Ph.D. dissertations. He is a member of the Dutch Academy of Arts and Sciences, the Academia Europaea, and the Institut 10

11 International de Philosophie. In 1996, he received the prestigious Spinoza Award for his total research from the Dutch Government. Homepage: In a recent interview, Johan van Benthem spoke about his entry into the field of logic, the course of his career, and his general views on logic. Excerpts follow here. Fenrong: You are very famous in China because of your logic, but we know little about you personally. Could you tell us something about yourself As far as I know, you got a bachelor in physics. When did you begin to think about logic Prof. van Benthem: I was born in 1949, and became a student in the exciting 1960s and 1970s in The Netherlands. I liked every subject in high school, but had not fallen in love with any one of them in particular. When I entered the university, the Department of Physics shared a building with the Department of Philosophy, and I once accidentally walked into a logic course. For me, this was an amazing eye-opener: the fact that logic reveals an elegant mathematical structure behind the ordinary things that we do, like talking, reasoning, and argumentation. So I fell in love with the subject, switched departments, and did a Master's Thesis on temporal logic in Philosophy, and another on weak axioms of choice in Mathematics. Upon graduation, I quickly became an assistant professor in Amsterdam. These were the days of the Student Revolution which turned the old university system upside down. Being the youngest faculty member, I was immediately elected head of the Philosophy Department, and found myself chairing endless general assemblies of students, professors and secretaries, full of radical talk about a new life without power or authority, and a new type of human being, where students would grade themselves, and courses would serve the needs of the working class rather than an intellectual elite. Wonderful times, because we were all young, with a strong sense of liberation and a new world but also heavily politicized, with many problematic events on the darker side. Anyway, those 1970s are already archeological prehistory to our students today, as Dutch universities have long since returned to autocracy, hierarchy, and business models. 11

12 In late evenings, at home, I found an oasis of private time for doing logic research, as a counterbalance to preserve my sanity amidst all the revolutionary events. I did my Ph.D. on modal logic from 1973 to 1977 with Professor Martin Löb. In 1977, I became a professor at University of Groningen, with the task of setting up a logic environment between philosophy and mathematics. In 1986, I returned to the University of Amsterdam as Löb's successor on the Beth Chair. Since 1991, I also teach at Stanford University in California, from April to June every year. Fenrong: Can you say something about your dissertation, Modal Correspondence Theory You also wrote a chapter on correspondence theory in the Handbook of Philosophical Logic, why did you choose this topic Your book Modal Logic and Classical Logic was published in How is that connected Prof. van Benthem: The dissertation is about the core subject of philosophical logic: modality, but the methods are mathematical. My feeling was that philosophical logic should not be an isolated alternative to mainstream logic with its own 'industry', because one can understand it by classical methods. Based on translations of modal languages into classical ones, I used model theory of first-order and second-order logic to describe the expressive power of modal logics over all sorts of frame classes. In particular, this research led to the discovery that the modal formulas are precisely the first-order formulas that have the special property of invariance for bisimulation. Ever since that time, my methodology is to work 'in tandem'. I look at many things from two perspectives, modal and first-order, as a sort of Gestalt Switch. About the Handbook, what can I say The editor Dov Gabbay told me to write on this topic, and so I did! The monograph Modal Logic and Classical Logic was ready by 1982, though its publication history is complicated, running from Poland to Italy. It is the extensive version of my dissertation plus later work on a comprehensive account of correspondence between modal axioms and frame properties, modal definability of frame classes, and many other pleasant things besides. 12

13 Fenrong: The Logic of Time was published in What were the most important ideas that you wanted to present in this book Prof. van Benthem: This book continues an interest from my Master's thesis. But it is different in spirit from my dissertation. The Logic of Time is not primarily about logic techniques like modal correspondence, but its central interest is understanding Time, analyzing philosophical and linguistical issues about our common sense temporal representation and temporal reasoning with logical tools. What interested me was the fact that there are different ontological views of what time is, points or intervals, and different ways of reasoning about it. I tried to use logic to create different views of time, not just the canonical view from physics. By the way, the study of time has always been a common concern of many disciplines, and through the book I also acquired many contacts in artificial intelligence and even cognitive psychology. Because of this interdisciplinary audience, I guess, the book keeps being reprinted. Fenrong: Which view do you prefer, points, events or intervals Prof. van Benthem: Philosophically, I prefer events but mathematically, despite all my radical enthusiasm of that book: still points! Fenrong: In 1986, you published a new book, Essays in Logical Semantics, which is mainly about generalized quantifiers. It is very different from what you did before. Why this change Prof. van Benthem: What happened was that around 1980, I became interested in natural language. Why this shift As more often in life, it is not a matter of planning: you are steered by encounters with other people. In Amsterdam, I had some excellent students with an interest in semantics, such as Jeroen Groenendijk, Martin Stokhof, and Alice ter Meulen, well-known names today. People often say that a teacher influences the students, but it also works the other way round: students form their teachers! Through them, I became interested in Montague Grammar. The other 13

14 influence was an old high-school friend, Frans Zwarts, now an influential Dutch linguist and the Rector Magnificus of Groningen University, who took up the study of generalized quantifiers, as then emerging in the work of Jon Barwise, Robin Cooper, Ed Keenan and others. Fenrong: But what is so important about generalized quantifiers Prof. van Benthem: I had been raised in the paradigms of formal language philosophy, which says that natural language is not clear, but should be 'cured' by logic. But then through formal semantics, I realized that there is a lot of structure to natural language, which is a remarkably successful medium for human communication that shows no signs of abdicating in favour of logical formalisms. In particular, I became fascinated by the expressive power of language for describing situations by means of quantifiers. What can we say, and why is this repertoire natural and stable across the languages of the world I tried to give mathematical sense to this in Essays in Logical Semantics. The book shows how our human quantifier repertoire meets systematic requirements of expressive completeness in terms of properties like monotonic inference and other semantic constraints. It also shows how quantifiers strive for algorithmic simplicity, something which I modeled in terms of 'semantic automata'. A final theme, and still a continuing interest of mine, is 'natural logic': how much logical inference can be done directly on natural language surface forms, without deeper logical constructions Fenrong: In your 1991 book Language in Action, the focus is on logical and typetheoretic foundations of Categorical Grammar. In that work, the reader sees mainly lambda symbols and derivations, and you seem to have become a proof theorist. Prof. van Benthem: After Essays in Logical Semantics, I wanted to understand the more general mechanisms of natural language, the 'operating system', so to speak, in which quantification and inference function. The key idea here, that objects of thought form a hierarchy of categories, is an old one in philosophy and mathematics, which underlies both Bertrand Russell's Theory of Types and Lesniewski and Ajdukiewicz' 14

15 Categorial Grammar. The two traditions came together in the 1950s in the work of Joachim Lambek, and my book takes this much further. Natural language may be viewed as a sort of type theory or lambda calculus but importantly, it is one with only very limited resources, not the mega-rich structure needed for the foundations of mathematics. In my book I develop the idea that a small fragment of lambda calculus provides the 'glue' for composing meanings in natural language, and I show how this can be developed by using resource-sensitive versions of lambda calculus and substructural proof systems. In particular, I give a Curry-Howard semantics for Lambek-style grammatical derivations, which shows how proof and semantic meaning are really two sides of the same coin. In this connection, I also showed more generally how proof theoretic-methods like Cut Elimination can be used in mathematical linguistics and semantics. So it is no longer a matter of 'formal language' versus 'natural language'. The more sophisticated question is rather how much logical glue must be mixed with natural language to explain how it works. This still seems crucial in the contact between logic and linguistics. Fenrong: Well, the 1997 Handbook of Logic and Language is still in this general line that you mention. It has had great success in logical circles worldwide, and it has been very influential and widely referred to by Chinese logicians. As one of the editors, what do you think were its main features Prof. van Benthem: Interfaces between logic and linguistics go back to Antiquity, but they were never very systematic until the work of Richard Montague. The Handbook documents the main modern research lines, with their links to computer science and cognitive science. Our authors covered the main logico-linguistic paradigms, such as government-binding theory, situation theory, or game-theoretic semantics, in what we hope is a non-partisan manner. Moreover, the book contains the major descriptive topics and techniques in the field, such as generalized quantifiers, categories and types, temporal expressions, situations and partial logic, dynamic interpretation and inference, feature structures and unification logics. Our purpose with this big 15

16 enterprise was also to help create a community, by offering a standard reference facilitating back-and-forth flow of ideas between logic and linguistics. Fenrong: What is your opinion about the relation between logic and language Prof. van Benthem: My views on this have fluctuated a lot over time. The stronger thesis would be that language really has mathematical structure; the weaker claim is just that logic is convenient tool to model phenomena in natural language, and make them visible. In the Handbook, you will find both views represented. As for myself, I see natural language as much more than a logical formalism, for instance, because it also represents a process of biological evolution. Logical analysis in my style, no matter how close to my heart, only gives a partial view, of language as an information device, it does not say much about the other functions. Actually, for me the more interesting contrast nowadays would be different. Logic is primarily about artifacts: languages that are designed by people for some purpose. But natural languages are not designed; they are an ongoing cognitive phenomenon. Now, instead of grand questions about contrasts between the two, I am interested in their contacts. Nowadays, you have all these intermediate languages, designed but nevertheless, actually used. Examples are programming languages for communication between humans and computers, or the professional language used in a law court. I would like to understand how we manage to thrive in this mixture of natural and formal language. Some designed behaviors seem to 'fit' into our cognitive functioning, and are incorporated, others do not. Think of successful games that become part of our lives. To me, understanding such mixtures is a much more fruitful perspective on cognitive science than grand issues like 'Are Machines like Men' By the way, an aunt of mine used to say that men are certainly machines; the only real open question is if women are... Fenrong: So you spent 10 years of your life on temporal and modal logic, and then 10 years on natural language. In the 1990s, it seems that your interests were shifting again. Let me turn to your book, Exploring Logical Dynamics which was published in 16

17 1996. What was going on there Prof. van Benthem: First, I should say that the 1990s saw my return to modal logic, partly because of the pressure of having very good students, like Maarten de Rijke, Yde Venema, and others. They expected activities in modal logic from me, and who was I to refuse Nevertheless, my new interest came with a new twist. Modal logics were increasingly used for studying computation and process structure, especially, in the guise of dynamic logic. So I did work on that. Moreover, this went very well with a broader 'Dynamic Turn' in philosophy and linguistics, making cognitive actions and procedures of interpretation, or reasoning, into primary objects of logical theory. Examples of that trend are belief revision theory, discourse representation theory, or dynamic semantics in the Amsterdam style. I found this very exciting, because it made me aware of aspects of logic that had always hidden in the background of my mind, but never visible in a logical calculus. The book tries to convey the attractions of this new perspective, develops its logical structure, and demonstrates some surprising repercussions even for apparently standard parts of logic. Fenrong: What kind of technical results did you get in this book Prof. van Benthem: I give a treatment of modal languages which totally centers around bisimulation, and also characterize the expressive power of dynamic logic in terms of safety for bisimulation. Moreover, these languages come in hierarchies of expressive strength, all the way up to full first-order logic. In particular, this modal fine-structure view led to the discovery, together with Hajnal Andréka and Istvan Németi, of the Guarded Fragment, a large and hitherto unknown decidable part of first-order logic, which has been much studied since. This broad technical framework is used to study different process models, varieties of dynamic inference, as well as architectures for logics with both static and dynamic components. Another major interest in the book is the search for low-complexity systems, preferably decidable. One of the surprising repercussions of the dynamic perspective in the book is my analysis of first-order logic as a modal logic, provided you accept a generalization of 17

18 Tarski semantics. This reveals a decidable core to standard logic, while the usual undecidability turns out a side-effect of mathematical features of standard models. Fenrong: In October 1996, the Dutch National Research Organization (NWO) awarded you one of its first annual Spinoza grants, to an amount of 2 million guilders. As far as I know, this award is the most prestigious in Dutch science. Its rewards a very limited number of scientists with a large grant without any conditions attached. You used this grant to set up the project Logic in Action. What was that all about Prof. van Benthem: Actually, since there were no conditions, I had wanted to do something really spectacular, and buy a tropical island in the Florida keys as my gift to the profession. My idea was that this would serve logic for centuries to come. But alas, common sense prevailed. I then used the money to do three things which I has always liked to see happen at ILLC, but which had not. One was a research project on 'Logic and Games', headed by Paul Dekker and Yde Venema, bringing together modal logicians and semanticists in the study of communication and interaction. Amongst other things, this project has become a bridge between logic and game theory. The second project on 'Computational Logic', headed by Maarten de Rijke, was meant to give our theoretical research in modal and other logics a 'strong arm', i.e., a laboratory for computational experimentation. This has developed into our current Lab for Language and Inference Technology. After the Spinoza project ended, these project leaders all got permanent jobs at our university, so the story goes on. The third and final project was about 'Didactics and Dissemination', headed by Jan van Eijck, already a professor at CWI Amsterdam and Utrecht. It has produced innovative educational software for logic teaching, as well as a book initiative for high school texts on logic, starting with a volume called "Thinking Machines". Moreover, the project allowed us to have many foreign visitors over the years, subsidize dissertation prizes, and so on. It is really nice to be rich, if only for five years! Fenrong: In 1999, you started the electronic lecture notes Logic in Games. Can you explain why games make sense to logicians 18

19 Prof. van Benthem: Well, you can always predict my next step from an earlier book. Exploring Logical Dynamics just described processes, but it did not explore social multi-agent interaction, like communication, which is really the natural next step. Nowadays I see a social mini-process like asking a question and getting an answer as just as basic in logic as drawing a valid conclusion by a lonely thinker. This is the natural version of the Dynamic Turn, it seems to me now. And the best model for that are games. Games have been popular in lots of disciplines, and some people think all their mathematics is known, with the Nobel Prize for John Nash as a crowning event. But I think that the story is only beginning. Games are ubiquitous as natural activities. Moreover, they seem cross-culturally stable. I have taught logic games in every continent (including Asia) without ever encountering any barrier. There must be general cognitive reasons for this appeal. Moreover, games are a rich model for interactive cognitive processes about which we have vivid intuitions. Thus, they are attractive to logicians interested in information flow beyond the usual computational models. Games also offer a possibility of experimentation; they are a free cognitive laboratory. We can design games for logical purposes, and see how humans respond. In fact, of course, argumentation and many other logical activities have a gametheoretic flavour. Many of us have had the bitter experience of 'losing' an argument in a group, which is almost as bad as losing a football match. Fenrong: A monograph Logic in Games has been announced many times, but should now be published in How should we understand this title Prof. van Benthem: First, notice the 'should'. I keep publishing new papers in this new area, sometimes even in economics journals which seems to have raised my prestige somewhat with my more practically oriented sons. But the book is still in flux. Anyway, its purpose is to demonstrate and develop two directions of contact. First, 'Logic in Games' refers to the fact that many logical activities can be cast as games, witness the many existing 'logic games' for semantic evaluation, consistency maintenance, proof, or model comparison. Often, basic logical notions then have to do 19

20 with winning strategies for one of the players. The second sense of the title refers to logical structures in general games, such as economic competition, social cooperation, chess, soccer, war, or dialogue. All these activities involve rule-governed action by intelligent reasoning players. In this broader realm, logic helps clarify possible game structures, as well as the mechanics of deliberation by players as the game proceeds. The book has lots of results on logical systems which serve this purpose. Its total proposal is a joint enterprise where logic provides the fine-structure of the more global theory of strategic equilibrium in game theory. Fenrong: We seem to have covered most of your books. If I look at the topics that we discussed in this interview, I see a logician, a linguist, computer scientist, perhaps even a game theorist. But where is the connection with philosophy Prof. van Benthem: I am glad you raise this point, because it is a fair one. To me, there is no longer a clear dividing line between the label 'Philosophical Logic' and issues in surrounding disciplines. If you look at the 20th century history of philosophical issues of meaning, knowledge, or action, ideas travel between fields all the time. In my Stanford 2003 spring course 'Logic in Philosophy' I show this for a number of basic topics, such as compositionality, valid consequence, conditionals, knowledge, and belief. I see philosophy as a 'catalyst' or 'spice' which makes this interdisciplinary mixture work. By contrast, I have become skeptical about more 'philosophy-internal' issues in philosophical logic. These seem like intellectual black holes: they absorb a lot of clever energy, but nothing ever seems to come out. Fenrong: In the 1990s, you were the founding director of ILLC, and also the first president of Folli. You were awarded an honorary doctorate at Liège 1998, and elected a member of the Institut International de Philosophie in 2001.You are the Chairman of NWO s national stimulation program for the Cognitive Sciences. And your CV is even much longer. In 2003, the Universiteit van Amsterdam has appointed 20

21 you as University Professor, which is a very prestigious position. What do you think about these awards, and your general reputation How do you deal with all these interests, and all this busy work, while staying creative in your research Prof. van Benthem: I try not to think too much about the questions that you raise here. If I did, I might get seriously alarmed. I am always surprised by biographical descriptions like this. Sometimes I think there are two Johan van Benthems. One is just an individual person, who never grew up much since his student days, who likes to read in a corner, who talks occasionally to students and colleagues sometimes sense, sometimes nonsense who pursues some interests one after the other; and who likes to travel around the world to explain to people what he has just understood for himself. But then there is also a second, public Johan van Benthem, who wants to influence the field, who creates public institutions, and who tries to interfere with other people's lives, making his students do this, and his colleagues do that. Sometimes I ask myself how these two very different people get on together... Certainly the busy public life takes always time from research, but on the other hand it is also helpful, because you get greater exposure to new ideas, especially working with students, who can be sources of the most unexpected ideas. My public ego gives me many interactions that enrich my private ego, and stimulate research. Fenrong: So, you began your research in logic, then language, then information. It looks as if you often spend five years on a book, and then use it to close a chapter of your life. What is the next chapter, and your plan the next five years Prof. van Benthem: You have to realize that this five year planning is just a pretense. I look back in my history, then write down the resultant of five years of past work. But I never sit down to plan my next five years, in the style of a plan economy; if only, because I am not very good at predicting my own future. But if you insist, I would say that I want to learn a lot more about connections between logic and cognitive science, i.e., what people really do when they use language or reasoning, all the way to what 21

22 goes on in their brains. My being the chairman of this National Cognitive Science Committee reflects a genuine interest. Even so, unlike my faster creative colleague Michiel van Lambalgen, I have not yet published anything in this direction. A good example is what was said earlier about games, and designing new practices that can be inserted smoothly into human behavior. I would love to do joint research on that with empirical scientists, and find out which new practices work and which ones do not, and why. In fact, we are exploring such contacts now with groups in artificial intelligence and experimental game theory. And more generally, I am becoming interested in doing things with logic that really change people's lives. So you could say that I am getting old, returning to those revolutionary student days that we talked about, when it seemed as if academics could really help to make this world a better place. References: [1] J. van Benthem, 1983, The Logic of Time, Reidel, Dordrecht, (Synthese Library, vol. 156). Revised and expanded edition published with Kluwer, [2] J. van Benthem, 1985, Modal Logic and Classical Logic, The Humanities Press, Atlantic Heights. [3] J. van Benthem, 1986, Essays in Logical Semantics, Kluwer, Dordrecht, (Studies in Linguistics and Philosophy, vol. 29). [4] J. van Benthem, 1991, Language in Action: Categories, Lambdas and Dynamic Logic, North-Holland, Amsterdam, (Studies in Logic, vol. 130). MIT Press [5] J. van Benthem, 1996, Exploring Logical Dynamics, CSLI Publications, Stanford. [6] J. van Benthem, , Logic in Games, electronic lecture notes, ILLC [7] Amsterdam & philosophy Stanford, and book in preparation. 22