JUDGEMAP: A Model of Judgment, Based on the Cognitive Architecture DUAL : A dissertation submitted in partial fulfillment of the requirements for the degree of Ph.D. in Cognitive Science

Abstract

Judgment of a target stimulus can be systematically shifted in either direction depending on the context. The experimental findings are robust, but often controversial – seemingly similar experiments can lead to opposite effects depending on the stimuli or the procedure. There is no single theory that can explain and predict all these controversial data.
In addition, very few computational models of the judgment process are implemented in a way that simulation experiments can be run on them. The hope is that building such models could contribute to the understanding of these controversial data in a coherent way. Our assumption is that several mechanisms interplay in the process of judgment, each of them leading either toward assimilation or toward contrast, and the behavioral data reflect only the outcome of the competition between these mechanisms. That is why seemingly similar experiments could lead to opposite results.
Since the model of judgment should be dynamic and context-sensitive, it seems natural to use the dynamic and context-sensitive cognitive architecture DUAL as a basis for modeling. In this way, the model of judgment will be integrated with other cognitive processes rather than being considered in isolation.
Moreover, an attempt has been made to use the same basic DUAL mechanisms that have been used for analogy-making to model judgment as well. This exerts a strong restriction on the range of possible models of judgment and does not allow us to construct any possible mechanism that will lead to the desired data. On the contrary, if successful, the results would mean that the same mechanisms could produce analogy and judgment as well as the assimilation and contrast effects. This is a very challenging task and of course, it could not be fully accomplished within a dissertation, however, this is the driving force behind the studies.
An underlying assumption behind the JUDGEMAP model is that judgment is based on a process of mapping between a set of stimuli and the set of possible scale values. In this way, judgment is considered to be a close relative to analogy-making and based on the same mapping mechanisms. This mapping is trying to keep the relational structure of the two domains in correspondence, i.e. better stimuli to be mapped on higher ratings. Thus ordering relations play important role. Another important assumption is that the set of stimuli used as background when evaluating a single target stimulus (called comparison set) is dynamically constructed in WM by spreading activation mechanisms and based on similarity with the target stimulus. Then the target stimulus is included in this comparison set and all elements of this set are mapped on the scale competing for the specific scale values.
The result is that very few modifications of the existing AMBR mechanisms have been made and some new mechanisms have been developed, holding all principles of the DUAL architecture. A wide range of psychological data has been replicated with the model. In addition, the model has made some predictions that have later on been tested and confirmed. Some of these predictions have seemed quite strange at first but have turned out to be true. Confirmation of non-trivial predictions is one of the model’s strengths. In addition, it has been demonstrated that the same mechanisms that are used for judgment can also produce choice and can even replicate some context effects on choice. At the same time, the model could not replicate all the existing experimental data and thus further development is needed, especially integrating it with perception and category learning.