tract infections and for sepsis. A combination of methods is used to study reasons for non- adherence. In general practice, think-aloud methods using vignettes showed that many treatment decisions are made without any contemplation of potentially relevant cues, suggesting that doctors often rely on simple decision routines. In some cases, this involves routinely prescribing second- choice drugs. This was recently published in Qual Saf Health Care 2002; 11: 137-43. In hospital, preliminary results from a lens model study using actual cases did not provide a good explanation for non-adherence. Factors included in the analysis were a range of potentially relevant and irrelevant case characteristics, such as age, gender, liver and
kidney function, comorbidity, hospital
fever, leucocytes-counts, acquired infection. It seems
that many of the empirical antimicrobial treatment decisions are routinely made choices for second- choice drugs.
Effects of Systematic and Representative Stimulus Design on Policy-Capturing
Mandeep K. Dhami University of Victoria, B.C., Canada E-mail: email@example.com
Judgment policy-capturing is perhaps one of the main avenues for neo-Brunswikian research. This literature suggests that among other things, judgments are the result of a linear, additive process, where few differentially weighted cues are used. The findings tend to be robust: applying to many different types of judges and judgment tasks. However, the fact that most studies involve hypothetical stimuli that comprise a factorial (orthogonal) combination of cues, draws attention to
the questionable validity of the findings.
stimulus design matters. The judgment policies that researchers try to capture are not adapted to the systematically designed stimuli that are presented to participants in the laboratory. Rather, policies are adapted to the stimuli that people naturally sample during their experience with the task. Brunswik believed this called for a new methodology: representative design. Representatively sampled stimuli may differ from systematically designed stimuli in terms of the number of cues, their values, inter-correlations, distributions, and ecological validities. Therefore, if Brunswik is correct, we should expect to find differences in the nature of the policies captured under systematic and representative conditions. Furthermore, we should expect that policies captured under systematic conditions would be poor at predicting judgments made on representative stimuli.
I conducted an experiment to test these two hypotheses. Senior college students were asked to judge the desirability of two sets of apartments for rent. One set included real apartments that were randomly sampled from an Apartment Directory given out to all students by the University housing office. These real apartments were described in terms of seven cues. The other set consisted of hypothetical apartments comprising an orthogonal combination of the cues. Overall, the policies captured for each individual differed: people 'used' fewer cues in the systematic condition, and the weights attached to the 'most important' cues differed under the two stimulus conditions. Furthermore, the policies captured using systematically designed stimuli were poor
predictors of individuals' representative set of cases.
These findings contribute to a small, but growing body of research demonstrating the effects of stimulus design on policy-capturing. Perhaps we need to sit back and think about what policy-capturing research over the past 50 years has 'really' told us about human judgment. The need for policy-capturing researchers to think about methodological issues is of great importance because of the impact their findings often have for professional judgment and social policies.
Statistical properties of the parameters of the lens model equation
Hassan Dibadj, SUNY at Albany E-mail: firstname.lastname@example.org
I am beginning work on a dissertation about exploring the statistical behavior of the parameters of the lens model equation and understanding how this behavior reflects the characteristics of the environment, the information system and the cognitive system. The goal of this study is to discover the statistical properties (distributions) of the parameters of the extended lens model equation under various assumptions about the judgment system and the environmental system being judged. The simulations will include factors and situations that affect judgmental performance as well as factors that affect the performance of judgment analysis techniques. The results of this study will facilitate the use and interpretation of the parameters of lens model equation.
Newsletter 2002 page 8 of 28