Computerized financial planning: Discovering cognitive difficulties in model building.

Abstract

This research uncovers cognitive difficulties financial analysts encounter when building financial planning models using generic software packages. These difficulties are the result of the cognitive dem and s inherent in the task and the mental manipulations required by the interface representations implemented in software packages. The results of the study indicate where errors are most likely to be produced with the two most popular interface representations and which formulas require more mental effort. More importantly, the research reveals why these errors occur and why some formulas dem and more mental effort than others. The empirical observations and the methodology used in the research show how to approach the design of the interface for financial modeling systems and the design of computer tools for guiding and assisting the user. The dissertation was executed in three steps: (1) The task of building financial models was characterized by breaking it down into its essential operations independent of any given software package. (2) Cognitive models (GOMS models) were used to represent the mental processes that skilled financial analysts are likely to follow when executing these essential operations. These GOMS models served as the basis for making predictions of the relative frequency of errors and the amount of mental time needed to translate financial concepts into formulas. (3) An experiment was conducted with 31 experienced financial analysts whose performance supported in general the predictions made by the cognitive models. The first of the systems studied in the experiment is a spreadsheet package with an interface representation characterized by non-mnemonic naming and absolute referencing. The second system is a comm and -driven package with a mnemonic and relative referencing interface. Although this second interface (called keyword) requires more time overall, it produces seventy-five percent fewer simple errors and requires less mental effort. Given the overall serious cost of errors in financial models, the interface having the keyword representation is considered far superior.