Article #5
Newsletter list 


Previous Article 

Next Article 


Integrating Social Judgment Theory and Simple Heuristics

Mandeep Dhami
London, United Kingdom

In January 2001, I will be taking up a post-doctoral position at the Department of Psychology, University of Maryland, to work with Tom Wallsten.

Over the last year, I have been working on a few ongoing Brunswik-related projects, and most are close to completion. One project that also fulfills a personal goal is the completion of my doctoral thesis entitled "Bailing and Jailing the Fast and Frugal Way: An Application of Social Judgement Theory and Simple Heuristics to English Magistrates' Remand Decisions." In this, I conducted three studies that address issues of interest to neo-Brunswikian researchers. These included decision maker-related questions concerning cue use, consistency, agreement, post-decisional confidence, and "insight." A main aim was to compare the relative validity of different models of the judgment process. The other questions concerned the effect of task characteristics such as information availability and time pressure on the decision maker. I also compared policies captured using what Brunswik (1956) would call a representative design, via an observational study, and policies captured using systematic design, namely hypothetical cases comprising an orthogonal cue set.

In England and Wales, magistrates (most of whom are lay judges) make decisions on around 98 percent of all criminal cases. Their decisions have huge ramifications for defendants, their families, the criminal justice system, and the general public. However, until now, they had escaped scrutiny from psychological research. Similar to past SJT research, I found that most magistrates demonstrated inconsistency in their decisions in a test-retest situation. The modal response was recorded for each of a set of cases, and each magistrate disagreed with this response on at least some cases. The cues that were found to influence magistrates' decisions according to their models differed from those magistrates directly reported using. Here, the use of extra-legal cues was under-reported. Despite these results, all magistrates reported feeling highly confident in their decisions.

Unlike most past SJT research, I compared alternative descriptions of the judgment process. Brunswik (1943, 1952) favoured the use of correlational statistics for describing the process of vicarious functioning. Following this, past SJT research has tended to use the multiple linear regression model to capture and represent individuals' judgment policies. These static, structural models depict peoples' judgments as being the product of a linear, compensatory integration of multiple cues that are differentially weighted. Also, the same cues are used in the same way on each case. As such, these models do not meet the criteria of psychological plausibility, flexibility and adaptability. Brunswik (1955, 1956) recognized that there may be alternatives to the correlational model, as did Hammond (1955, 1996).

Recently, simple, fast and frugal heuristics have been proposed as viable alternatives to the regression model (Gigerenzer & Goldstein,1996). These are precisely defined process models that comprise principles for information search, stop search, and decision making. Many are non-compensatory. I developed a model for binary categorization tasks, called the Matching Heuristic. In this model, few cues are searched in order of their validity, and search stops once a cue that points to a punitive decision is found. Only that cue is then used to make the decision. There is no integration of cues.

In each study, I compared the relative ability of three models (i.e., Matching Heuristic, a unit-weighted linear model and a differentially weighted linear model) to predict magistrates' decisions, both on the cases used to form the models and on a new set of cases. The results were consistent across the studies: the Matching Heuristic proved a better descriptor and predictor of magistrates' decisions, than the other models. This result held when magistrates were presented with an orthogonal cue set, and a representative cue set. It also held when magistrates worked as individuals, and when they worked as benches (small groups). Some of these results are already in press (Dhami & Ayton). In a study in the medical domain, the Matching Heuristic was found to do as equally well as a logistic regression model in describing English doctors' prescription decisions (Dhami & Harries, in press).

Although the simple heuristic proved descriptively and predictively valid, I recognised that it lacked prescriptive utility in the legal domain, where accuracy is not the main criteria by which to evaluate the quality of a decision. Packer's (1968) definition of the requirements of justice and due process, more closely resemble the workings of a regression model than a fast and frugal heuristic. I have been consulting with the British government on these findings and thinking of ways to improve performance. In their exposition of SJT, Hammond et al. (1975) stated that "social judgment theorists firmly believe that all students of human judgment should engage in research that will help provide better social policies and thereby increase our chances for a decent life on earth" (p. 306). I guess you could say I have been trying to do my bit.

Contact Mandeep Dhami

Previous Article                Next Article

Newsletter list