The bootstrapped model--Lessons for the acceptance of intellectual technology

This paper is intended as a non-technical introduction to a growing aspect of what has been termed 'intellectual technology'. The particular area chosen is the use of simple linear additive models for judgement and decision making purposes. Such models are said to either outperform, or perform at least as well as, the human judges on which they are based, hence they are said to 'bootstrap' such human inputs. Although the paper will provide a fairly comprehensive list of recent applications of such models, from postgraduate selection to judgements of marital happiness, the work will concentrate on the topic of Credit Scoring as an exemplar - that is, the assignment of credit by means of a simple additive rule. The paper will also present a simple system, due to Dawes, of classifying such models according to the form and source of their weights. The paper further discusses the reasons for bootstrapping and that other major phenomenon of such models - that is, the one can rarely distinguish between the prescriptions of such models, however the weights have been arrived at. It is argued that this 'principle of the flat maximum' allows us to develop a technology of judgement. The paper continues with a brief historical survey of the reactions of human experts to such models and their superiority, and suggestions for a better mix of expert and model on human engineering lines. Finally, after a brief comparison between expert systems and linear additive models, the paper concludes with a brief survey of possible future developments. A short Appendix describes two applications of such models.