Universität Bielefeld - Sonderforschungsbereich 360

Public Information and Mutual Error

Wolfgang Heydrich and Hannes Rieser

Introduction

The general aim of this paper ist to discuss problems of the interraltion between action, communication and belief. More specifically, we are trying to develop empirically adequate descriptions of cooperating agents - descriptions, which can be used to explain, how convictions and beliefs, as well as the dynamics of their constant revision both affect and direct verbal behavior and joint action.
In order to study this empirically, we use the following setting: Instructor (Inst) and constructor (Const) are separated by a screen. Instructor has a block's assembly situated before him which he built up just before.

Fig. 1 Instructors block's world

He was instructed to describe it to the constructor. Const in turn was told to build up the same assembly according to Inst's directives. Among the blocks are cubes, rectangular blocks (cuboids), triangular blocks, cylinders ans segments of discs of different colour and size. Inst and Const are allowed to freely converse during the task.

Basically, Inst and Const have to solve a problem of cooperation. This problem, however, involves matters of coordination as well. E.g., Inst and Const have to gain awareness and common understanding of kind, number, and position of the blocks they are dealing with. They are already in need of this kind of understanding (a shared frame of reference, a common situation) in order to accomplish such elementary task as making transparent to each other which blocks they are talking about and which properties and relations they are attributing to them. The blocks Inst and Const have at their disposal are located on their respective sides of the screen. Inst's blocks (i.e. those which are perceptually accessible to him) are not directly asscessible to Const, and Const's blocks are not directly accessible to Inst. In this sense, information about blocks is private in our scenario - at least initially, i.e., as long as only direct perception i concerned. In order to be able to refer to and describe the blocks in a transparten way, private information has to be made - in some way or the other - intersubjectively accessible or public. To achieve this, verbal communication suggests itself as the most appropriate means in the given situation.

In this paper, we are dealing with some of the more fundamental aspects of the process of transforming private (e.g. only perceptually given) information into public (i.e. linguistically mediated) information. We will concentrate on some central features of this process connected with the opposition of private and public information.
Additional features will be discussed under this perspective, namely

(a): The make-up of discourse patterns and the notion of expectability of moves in an interaction (and, more specifically, the notion of expectability of actions in a conversation).

"Discoure patterns" and "expectability" are important notions for an adequate understanding of the required complexity of agent's beliefs and convictions in the succession of moves in an interaction. Effectively, only finitely many and finitely complex convictions must be sufficient, since it is hard to see, how empirical agents could possibly sustain beliefs of an arbitrary (let alone infinite) complexity or number. We assume that, basically, it is due to the rôle of entrenched patterns and established expectations in discourse that a finite number of finitely complex beliefs is sufficient. In this paper we mainly concentrate on the internal structure of the discourse pattern of giving directives, which is of central importance in construction dialogues. Directives, it will be shown, are typically processed as interactive procedures in order to carry out plans. They exhibit a dynamic structure of cycles and stages.

(b): The notion of ratification of actions as a means of establishing consent among discourse participants.

Ratification in discourse consists in turns which are used to affirm or reaffirm information and signal mutual acceptance ore agreement among discourse participants. This is a central procedure in discourse constitution, and it is an important one for our topic. Ratification is a means to tag informatin as mutually accepted and henceforth intersubjectively available. Although we will not go into the micro-analysis of ratifications in this paper, we will emphasise its important role for the dynamics of cycles and stages in the interactive pattern of giving directives.
There is one important aspect of our topic we just point at, but are not going to discuss in detail, namely.

(c): The origin of obligations in verbal interaction.

When dealing with the construction dialogues of our sample, we are confronted with a variety of speech-acts, many of them indirect. Typically, Inst's task is to ask, request ore order. Typically, Const's task is to accomplish requests and to perfom whatever he is asked to do. (This, however, is an oversimplification. There are many passages in our transcripts, where Inst and Const "change rôles" such taht Const takes the initiative and Inst reacts.) Clearly, the general distribution of rôles is already implied by the definition of the scenario and the overall purpose of the conversation (the construction task). Given this background, Inst may establish obligations indirectly. E.g., instead of explicitely asking Const to arrange some blocks in a specific manner, Inst may reach the same effect by simply describing the configuration of blocks he intends to be realised. It is well-known that there are more or less general mechanisms involved in procedures like those of using descriptions in the place of requests. We do not have to say much about these mechanisms here, and we leave the topic to the theory of indirect speech-acts.

Anke Weinberger, 1995-10-26