My academic background is mainly in philosophy, with some expertise in logic and formal semantics of natural language. I am currently working on the following topics, often in collaboration with other researchers:
In collaboration with the ATR Human Communication Science Group, I study what we call ``meta-communications'' in human conversation. In our view, there are two different kinds of communications going on in a human conversation: (a) base-communications, namely, exchanges of information about the topic situation of a dialogue, and (b) meta-communications, namely, exchanges of information about the progress of the conversation itself. Meta-communications can be done by verbal means (texts, rhythms, pitches, powers, and speeds of speech) and non-verbal means (gestures, gesticulations, gaze directions, and inhalations), and they are often unintentional. Nevertheless, meta-communications play crucial roles in our joint-managements of conversation, for, without them, we cannot share the information about the progress of base-communications, problems occurring in them, turn exchanges, topic changes, and background knowledge of each participant.
Our publications in this area includes Shimojima, Katagiri, and Koiso (1997) on the possibility of a mathematical theory of meta-communications, Shimojima, Koiso, Swerts, and Katagiri (1998) on informational values of the prosodic features of echoic responses in dialogues, and Koiso, Shimojima, and Katagiri (forthcoming) on informational values of speech rate changes.
My Ph.D. thesis, entitled ``On the Efficacy of Representation,'' is concerned with the efficacy of different modes of information representations as an aid for human reasoning. I used the mathematical frameworks situation theory and its descendant information theory (Dretske, Barwise, Perry, Seligman, etc.) to characterize various interesting properties of representation systems that account for their inferential potentials. My study has also led to a plausible analysis on the conceptual boundary between so-called ``graphical'' representation systems and so-called ``linguistic'' representation systems.
The work is a part of the larger project, envisioned in the paper ``Surrogate Reasoning'' (1995) by Jon Barwise and myself, that aims at the general theory of information representation. My other publications in this area includes Shimojima (1996a) on the free ride phenomenon in diagrammatic reasoning, Shimojima (1996b) on the interventions of operational constraints in the manipulations of diagrammatic representations, Shimojima (forthcoming) on the conceptual distinction of "graphical" and "linguistic" representations.
On a much more philosophical side, I am interested in the development of what may be called the "qualitative theory of information." In contrast to the traditional theory of information a la Shannon, the qualitative theory of information is concerned with the content of information transmitted via a channel, rather than its average amount. It is too common a practice for us to explain the behaviors of cognitive agents in terms of the acquisition and/or transmissions of information with particular contents, and the qualitative theory aims at putting these ordinary "information talks" on the rigid mathematical basis.
As far as I know, this new line of information theory was first envisioned by Fred Dretske (1981), and was partly inherited by the information-oriented theory of natural language semantics in situation semantics (Barwise and Perry 1983). The development seems to have reached great mathematical maturity recently, when Jon Barwise and Jerry Seligman developed "channel theory" in their book, Information Flow: The Logic of Distributed Systems (Cambridge University Press, 1997). On the other hand, the potentials of the mathematical theory developed in the book are yet to be much explored, and I for one started exploring them in a recent paper (1998, written in Japanese). In fact, there is a sense in which our theories of human communications and representational efficacy described above are instances of big "information talks," and they could be developed into full-fledged mathematical models through the marriage with channel theory.