•  
  •  
 

Abstract

This paper exemplifies how better knowledge about human judgement strategies known as heuristics can be used to improve software processes, especially estimation and prediction processes. Human judgement heuristics work well when they exploit a fit between their structure and the structure of the environment in which they are used. This use of environmental fit may lead to amazingly good judgements based on little information and simple computations compared with more formal approaches. Sometimes, however, the heuristics may lead to poor judgements. Knowing more about the strengths and weaknesses of human judgement heuristics we may be able to (1) know when to use formal process improvement approaches and when to use less expensive expert judgements, (2) support the experts in situations where the experts’ judgements strategies are known to perform poorly, (3) improve the formal processes with elements from the experts’ strategies, and (4) train the experts in the use of more optimal judgement strategies. A small-scale experiment was carried out to evaluate the use of the representativeness heuristic in a software development effort estimation context. The results indicate that the actual use of the representativeness heuristic differed very much among the estimators and was not always based on an awareness of fit between the structure of the heuristic and the structure of the environment. Estimation strategies only appropriate in low uncertainty development environments were used in high uncertainty environments. A possible consequence of this finding is that expert estimators should be trained in assessing how well previous software projects predict new software projects, i.e., the uncertainty of the environment, and how this uncertainty should impact the estimation strategy.

Share

COinS