Judgment and Decision Making: Neo-Brunswikian and Process-Tracing Approaches

environment to the laboratory. The combined error model combines the ecological framework in the form of PMM theory with the idea of stochastic components of the judgment process. The sensory sampling model accounts for confidence in sensory discrimination.

Together these theories represent a significant theoretical advance; we now have computational models that can reproduce most of the basic phenomena in the calibration literature. Nevertheless, these models are points of departure rather than a finished research agenda. Still no theories beside the PMM theory ( Gigerenzer, 1993; Gigerenzer & Goldstein, 1996; Gigerenzer et al., 1991) and the sensory sampling model ( Juslin & Olsson, 1997) provide anything like detailed accounts of the cognitive processes, and even these accounts remain largely untested in their processing details.

A problem with the combined error model is that--even if conjoined with the processing assumptions of PMM theory--the model remains underspecified in its account of the cognitive processes that form degrees of belief (internal probabilities). A second problem is that both the ecological model and its derivative, the combined error model, concentrate only at the processing of frequency information. Although it seems evident that subjective probability is responsive to frequency information, these formulations neglect data suggesting that probability assessment is also affected by similarity, or representativeness ( Kahneman et al., 1982). Further developments need to bridge the iron curtain between models that stress frequency and similarity, and connect research on subjective probability and calibration to basic research on memory and categorization.

A working hypothesis in our current research is that the exemplar-based models from categorization (e.g., Medin & Schaffer, 1978; Nosofsky, 1984) can inform research on probability judgment. The exemplar-based models are supported by a large body of data and provide algorithms that respond to both similarity and frequency in a sensible manner. In any event, we hope that the research reviewed in this chapter can serve to stimulate development of more refined and detailed computational models of the cognitive processes that support and underlie the assessment of subjective probabilities.

The research reported in this chapter was supported by the Swedish Council for Research in Humanities and the Social Sciences.

-92-