Construction of attribute spaces for personality disorders by querying experts

Gudrun Wesiak

University of Graz, Austria

Abstract

In the presented pilot study the field of personality disorders is integrated into the theory of knowledge spaces, developed by Doignon and Falmagne (since 1985). The Knowledge Space Theory constitutes a mathematical model, in which the representation and diagnosis of knowledge are directly related. The model not only allows an economical and efficient assessment of an individual's knowledge state, but also provides concepts for the organization of specified sets of problems within a certain field of information, i.e. for the construction of a knowledge structure. Individual elements are ordered by means of surmise relations (or rather, as a generalization of the concept, by surmise functions) which map each item to exactly one subset of prerequisites (or clauses) for the respective item. Any combination of problems that is based on the given surmise relations is called a knowledge state. Among others, the consultation of experts constitutes one method to construct a knowledge space. In the course of a query a given set of items is related by computing explicit implications (which the expert can either accept or reject) and their implicitly drawn inferences.

Assuming that a mathematical model permits a precise organization of specified sets of elements, the mathematical structure of the Knowledge Space Theory was applied to establish prerequisite relationships for personality disorders and their corresponding symptoms. The method of querying experts was employed to obtain a first sketch of the knowledge, or rather attribute structure in question. The procedure of the investigation covered a total of four queries, including two experts who each related one set of 28 items and one of 58. The problems contained in the itempool consisted of four distinct attribute groups, viz. the DSM IV personality disorders, two prototypal symptoms per disorder, the gradation or counterpart of each symptom, and a set of attributes taken from two dimensional models, one of which integrates personality traits (Horowitz, 1988), the other personality disorders (Millon, 1981; 1990; Millon & Everly, 1985). Each query resulted in a rather incomprehensible attribute structure, which was consequently reduced to a quasi-ordinal attribute structure, and thereafter to a symptom structure. The obtained data show that personality traits can actually be structured by means of surmise relations, even though the current investigation needs to be improved by some aspects.