A Cognitive Perspective on Analogy-based Project Estimation

Analogical or case-based reasoning (CBR) is a knowledge management technology based on the premise that history repeats itself, but not exactly. This leads to problem solving can be based upon episodic memory and retrieval by similarity. CBR has been used for many software engineering problems including support for software project management in prediction and lessons learned. Unfortunately results from using CBR are varied often for no obvious reason. Consequently we cannot predict a priori when CBR will aid problem solving.The origins of CBR lie in psychological studies from the 1980s. Since then, much work (including our own) has espoused a 'black-box' approach to understanding problem-solving in software engineering. Recent interest in CBR, as a knowledge management tool, has resulted in a large literature. However, this has emphasised algorithmic approaches typically used for well-defined problem solving. To solve non-trivial or ill-defined problems, such as those encountered by software project managers, experts use other cognitive strategies (e.g. creative thinking), yet we have sparse evidence about cognitive processes in software professionals using CBR tools. Personality has been found to impact cognitive processes. Therefore it will affect problem solving ability and strategy. The proposed project aims to address the limitations of previous research by exploiting cognitive psychology and computer science to investigate the cognitive aspects of software engineering when using analogy-based tools to improve problem solving effectiveness. This project is intended as a pilot study for this novel approach. We will investigate the cognitive processes and personality of software professionals using CBR to support problem-solving (specifically cost estimation) in situ.. This will yield insights into the effective use and appropriate development of CBR tools.It is a joint proposal between Brunel University (who will provide software engineering expertise and the CBR shell), Southampton Solent University (who will provide cognitive psychology expertise and manage the RA) and EDS (who will provide expert practitioners, real world problems and research feedback). We plan to submit one journal and four conference papers, maintain a project website, a wiki, and publish post-prints of papers in our university open research archives. We will run a workshop for collaborators, academic and industrial researchers to disseminate findings and generate ideas to lead to a collaboratively authored publication. Feedback from the workshop will help develop an agenda for follow up research.