Information search, use, reasoning and being strategic (but never all at once)
Clare Harries, Centre for Decision Research, Leeds University Business School E-mail: firstname.lastname@example.org
I’ve continued work on three or so lines of research (see Harvey, this newsletter). See the Brunswik website www.brunswik.org for a report on the one day international meeting on Clinical Judgment Analysis.
Vicarious functioning, vicarious reasoning and inequalities in health care
With Damien Forrest, Nigel Harvey and Ann Bowling.
In our three year project on the role of decision making in inequalities in the British healthcare system we have extended a classic clinical judgement analysis in several innovative ways and with interesting results. Each physician asked for the information and tests they wanted prior to making a multivariate treatment decision for each of 72 patients. This process focus allows us to include testing for heart disease as both a decision to be analysed, and as a piece of information influencing subsequent treatment decisions. It also allows us to measure the influence of information in relation to the proportion of cases on which it was sought. We elicited subjective policies not as ratings but as a series of graphs indicating the relationship between each level of cue, and the decision (we picked 5 or 6 decisions that we asked them to focus on). Thus we measured the subjective functional relationship between cue and decision and avoided ambiguity of the meaning of ratings. In addition to this, participants listed reasons why they would do each treatment or test and reasons why they would not do that treatment or test.
The preliminary results are striking. Analysis of the idiographic patterns of behaviour show that (only!) half of the physicians in each specialist group (Cardiology, Care of the Elderly, and General Practice) were less likely to treat the old and they did so in terms of different aspects of their patient management (from the amount of basic clinical information they collected, to ordering tests, ordering angiograms, and revascularization). Elderly people with chest pain and suspected angina in the UK are vulnerable to a cumulative effect of discrimination. Most importantly, different physicians gave different reasons for their decision making. For example, some physicians implicated age (in our rationing based society) as a simple contraindication to treatment (a “fair innings” argument), others implicated age as a cue to co-morbidity, which was itself seen as a contraindication to both testing and treatment, or as a cue to reduced potential benefit;
other physicians implied that old age was a cue to a lack of desire for treatment. The end pattern of age-related decision making is a product of both vicarious functioning and vicarious reasoning.
Strategy and knowledge in decision making I have finally moved from the theoretical analysis of a Brunswikian perspective on strategic decision making to a series of studies that tease out different parts of the process. These studies build on work on dynamic decision making, complex problem solving as well as lens model analysis. One focus is on the strategic decisions that change the ecological validity of predictive cues. Those decisions that change the levels of each cue can be seen as more tactical. Another focus is on the distinction between uncontrollable and controllable aspects of the environment. We’ll see where all this gets me next year.
Information search in simple heuristics Mandeep K. Dhami, and I continue to work on the collaboration that we started five years, four countries and five cities ago. We compare physicians’ information selection to the search specified in simple (matching heuristic) models of their decision making. These simple heuristics are good at predicting physicians’ decisions, but of course the physicians select many more pieces of information than specified in the model. For most physicians the simple one-reason for decision making identified by the model is the cue looked at first, sometimes it’s the cue looked at last, and for one or two physicians it’s the cue looked at after some information and before other information. This work will be presented at the Brunswik meeting this year.
Advice taking and trust in advisors
Nigel Harvey University College London
continuing work on the advice-taking paradigm. In this task, all cues and the criterion refer to the same variable. For example, people are given forecasts of sales of a product that have been produced by four different advisors. On the basis of this information, they make their own best estimate of what sales will be. Some advisors are
better than others but judges can this only by experiencing outcome informs them of true sales figures
find out about feedback (which for the product).
We have directly compared results with those from an equivalent
from this multiple
probability sales had
learning task. In the to be forecast from
latter case, product different variables,
Newsletter 2002 page 13 of 28