Research Students Supervised: Past and Present

Brian Alden

An Expert System Approach to Retrograde Analysis

This thesis presents a system called RETRO, which has the capability of solving a number of retrograde-analysis chess problems of varying degrees of difficulty. Retrograde analysis problems are exercises in deductive reasoning, entirely different from conventional 'mate-in-n-moves' chess problems, their nature being such that any type of question, together with perhaps an assortment of associated initial conditions, may be presented to the solver, whilst their solution lies in asking pertinent questions at each step of the deduction process.

The aim has been to investigate the extent to which the deduction of past events may direct attention to these questions, a key concept being that of a 'Significant Event' (SE), which is an event of significance that must have occurred at some time in the history of the game. An SE is akin to a frame, the usual slots being replaced by the questions that the SE prompts, to which the answers may either provide a solution to the original problem or point to another SE, together with its association questions. It is shown that a small number of SEs possess widespread application.

RETRO maintains an explicit knowledge base of rules, grouped according to function, which are used both by the SEs and the system. This knowledge base is designed for easy amendment.

RETRO also maintains an explanation facility, which provides an account of its deductions, including a listing of rules utilised.

A number of examples are presented which demonstrate the effectiveness of the SE approach. There is also a discussion as to how the concept may be expanded, and its possible limitations.

The Open University
Faculty of Mathematics
PhD Awarded

Next