Graham White

I am a lecturer in the School of Electronic Engineering and Computer Science, Queen Mary, University of London. Here is my curriculum vitae.

PhD opportunities

I have a PhD studentship available for work on our large corpus of chatroom logs: details here

Research

Here are some of the things I have been working on recently: they are all quite closely related to each other, but, if you want to bring them under a single head, you should take seriously the remarks I make about phenomenology in the last section.

Action

I have a long-standing interest in reasoning about action: recently I have written on the system which Reiter describes in his book Knowledge in Action. My paper “Davidson and Reiter on Actions” (Fundamenta Informatica 2007) shows that much of Davidson's system (which he formulated in untyped second-order logic) can be reformulated in a much weaker, typed system, and that, in so doing, one can resolve some of the open problems in his book. Further work on this theme is described in “Contexts for Human Action” (in Visions of Computer Science, ed. Gelenbe and Sassone), where I describe a similar system in which there is a notion of success and failure of actions. These systems are typed, and the types represent contexts for action (propositions of those types represent knowledge explicitly available at those contexts).

Causality

Action and causality are closely related: I've written a good deal on causality in the past (using substructural systems to explain causal reasoning, or arguing that physical causality is actually difficult to express in the logical systems used in AI). In my paper “Causality, Modality and Explanation” ( Notre Dame Journal of Formal Logic) I write on a system, invented by McCain and Turner, and known as ‘causal reasoning’ in the AI community. I argue that it is basically a modal logic of explanation (and its main use in the AI community is to reason about causal explanation), but it is far more general than that: it can be regarded as a generalisation of Clarke's completion to the non-closed-world case, and it has also been discovered numerous times (by, among others, Girard and Schroeder-Heister). I am also responsible for the Stanford Encylopaedia of Philosophy article on medieval views on causality, which I have just updated. The medievals are worth looking at if you are interested in causal reasoning, because their thought was in many ways as techically sophisticated as ours is, but they are also culturally very dissimilar from us: so if you want to seriously escape cultural bias, have a look at them.

Contexts

Much of this work is done in typed logics, where the types are to be viewed as contexts either for action, for reasoning, or for the interpretation of language. I view all of these activities -- action, reasoning, and interpreting language -- as strongly contextual: contextuality gives an explanation of the way in which there seems to be a distinct lack of overt content which could reliably ground any of this stuff. In a related area (the design of programming languages) contextuality is pervasive, and for good reasons (see “The Philosophy of Programming Languages” in The Blackwell Guide to the Philosophy of Computing and Information, ed. Floridi).

The Human World

I take a robustly phenomenological view of formalising human reasoning in this work: the aspects of reasoning which are phenomenologically salient (its contextuality, its pervasive use of modal contexts, its reliance on a world of structured objects) are those which should be candidates for formalisation (and also those which can, in fact, be formalised by the category-theoretic tools of typed logic). Conversely, it should be difficult to find a seamless fit between the human world and formalisations with less overt structure than the category-theoretic, strongly typed ones that I use: I argue this in “On Scholarship” (in Bartscherer ed., Switching Codes).
Graham White
PGP public key Last modified: Tue Jan 4 20:33:15 GMT 2010