Article #20
Newsletter list 


Previous Article 

Next Article 


Research at University College London

Clare Harries
London, United Kingdom

1. Judgmental combination of Forecasts

Nigel Harvey and I are looking at the factors that affect our ability to combine other people's forecasts. We are doing this within the framework of a lens model, treating the forecasters as the cues, and modeling both the (artificial) environment and the judgmental combination of these forecasts. This work is part of a larger on-going project investigating judgmental combination of forecasts.

This year we began by investigating the effects of forecasts that exhibit typical patterns of behavior (such as a "trend bias"), on both explicit knowledge of the accuracy of those forecasts, and on tacit judgmental combinations. People seem to have a good explicit understanding of the validity of forecasters but are relatively insensitive to their intercorrelations. We found that forecasters were tacitly relied upon most ideally when they exhibited behavior that was atypically biased. We extended the research in two directions. In an investigation of the role of tacit and explicit forecasts we have looked at judgmental combination when one of the cues is in fact your own forecast, and which may or may not be labeled as such. In a further investigation of understanding of interforecaster correlations we have compared situations where forecasters tend to agree, and situations where they are equally accurate but do not tend to agree.

2. Models within the Lens Model Framework

At the SPUDM 1997 conference Ulrich Hoffrage demonstrated that a multiple linear regression did not distinguish between simulated judgements based on linear regression and those based on "take the best". In a paper delivered in December 1997 to the London J/DM group, Mandeep Dhami and I discussed the relative merits of using a "Take the best" model rather than a regression model to describe the judge and the environment within a lens model framework. We compared the "Take the best" model with the classic findings using regression models in terms of fit, agreement, cue use, consistency and self-insight. Since then we have tested our theoretical conclusions through reanalysis of physicians' decisions. We hope to present the results at this year's Brunswik meeting. If this does not happen, e-mail us to find out more.

3. Occupational Therapists' decision making

Earlier this year Cilla Harries (an Occupational Therapist) and I presented a paper to the 12th International Congress of the World Federation of Occupational Therapists. Cilla has looked at OT reasoning using traditional qualitative methods, eliciting the OT's concurrent verbal reports carrying out in-depth interviews based on a handful of cases. Together we argued for increased use of quantitative methods in OT, as a compliment to the currently used qualitative methods. Quantitative methods in OT are likely to be associated with reductionist experimental methods. In our paper we introduced judgement analysis and described its advantages in relation to other types of modeling. We described what would have been gained by extending a study of OT referral decisions to include judgement analysis. We argued that by using interviews, concurrent verbalization and judgement analysis we could obtain a fuller description of the cues that people attend to, the ways the cues are interpreted, and an objective measure of the ways in which cues influence decisions.

Contact Clare Harries

Previous Article                Next Article

Newsletter list