An investigation of the effects of computer spreadsheets on cognitive decision behavior
Paul Nevell · The University of Queensland · 1991
nnn The primary objective of this thesis is to contribute to the development of a theory of how computerized decision aids affect decision behavior. nnn This objective is attained by developing a cost-benefit-based model of how decision makers select from available cognitive information processing operators. The cost-benefit model assumes that decision makers establish qcognitive repertoiresq comprised of all the cognitive operators they are capable of executing. When making a decision, decision makers choose from this repertoire based on maximizing net benefit (execution benefits less execution costs). Decision aids are one of the task environment factors that influence expected execution costs. nnnn It is important to develop a theory of decision-aid effect so that decision aids can be matched more purposefully to the characteristics of the decision maker, the task, and the environment. This objective contrasts to the current situation, which is characterized by ad hoc implementation of decision aids on the assumption that aided decision making will improve the quality of decisions. More generally, expositions of decision process theories that help explain contingent decision behavior in different task environments are needed to improve decision performance. nnn Since the decision making model developed in this research was based on cognitive decision processes, a process tracing methodology (collection of verbal protocols) was used to gather data to test the model. This methodology relies on participants verbalizing their thought processes as they make a decision. Coded verbal protocol transcripts form the basis of the experimental data used for statistical testing purposes. nnn This research used one combination of experimental task (a financial analysis task), participant (intermediate-level professional accountants), and decision aid (a computer spreadsheet program). Participants were asked to assess the relative risk of seven competing loan applicants. Nine participants were provided with only paper-based task materials. Eight participants were also provided with access to a computer spreadsheet containing loan applicant financial information. nnn The central research proposition that decision-aid effect would be limited to increased execution frequencies of cognitive operators specifically supported by the decision aid was ratified. Apart from this expected effect, few differences were found in the decision behaviors of unaided and aided decision makers. However, post hoc tests of variance in decision behavior showed aided decision makers operate more consistently. This result cannot be explained easily within the cost-benefit theory used in the research. nnn By applying cost-benefit theories of decision behavior to individual cognitive operators, some tentative decision-aid design guidelines have emerged. When decision making is regarded as a sequence of cognitive operator executions, improved decision making performance results from improved cognitive operator execution performance. To achieve improved cognitive operator execution performance by using a decision aid implies that the affected operators positively contribute toward problem solution. Unless decision-aid designers analyze task requirements in terms of required (or desired) cognitive operators, the results achieved by providing decision aids will be unpredictable. Accordingly, maximum decision-aid leverage will be obtained only by knowing which cognitive operators result in information processing bottlenecks and by providing decision aids specifically designed to overcome these bottlenecks. nnn Because verbal-protocol-based research is expensive and time consuming, not all research propositions could be tested. In particular, research propositions relating to expected changes in decision behavior arising from the amount of task information embedded in the decision aid were not tested. Testing these propositions would add another factor to the empirical work, thus doubling required samples sizes and resultant data analyses. Appropriate experimental designs for future research are identified. nnn In summary, the research has contributed to the development of a cost-benefit theory of decision behavior that is useful in understanding the cognitive consequences of changes in the task environment. Because decision aids reduce expected costs of cognitive operator execution, they increase expected operator net benefit. Where operator net benefit is increased, increased execution frequencies follow. By analyzing other changes in the task environment in a similar fashion (that is, by determining the effects on expected operator execution costs and expected operator execution benefits), more accurate predictions of contingent behavior can be made.