Two of the most common reasons for not implementing a risk management program are cost and benefit. This paper focuses on whether the benefits of intervention can be shown to justify the costs. A confounding factor is that the acts of intervention during a risk management program may alter the outcome in ways we cannot separate and therefore cannot cost out. A second confounding factor is response bias – the tendency of individuals consistently to underestimate or overestimate risk, resulting in interventions that may be ineffective or excessively wasteful. The authors demonstrate that signal detection theory (SDT) can be used to analyze data collected during a risk management program to disambiguate the confounding effects of intervention and response bias. SDT can produce an unbiased estimate of percent correct for a risk management program. Furthermore, this unbiased estimator allows comparison of results from one program to another.

No abstract

The decision style of managers in applied settings were evaluated by raters. Two decision styles were identified, the socio-political and the data analytic. The consequences of these two decision styles were shown by evolving them in a cellular automata. The socio-political decision style produced unstable decisions. The data analytic decision style produced stable decisions.

This paper discusses a dynamic model of the user's interaction with the user computer interface. It addresses how the user's decision processes interact with different levels of complexity in the user interface. Cellular automata are used to model the interaction of user decision schema and user computer interfaces. The user's decision schema determine the navigation through the interface. Users with idiosyncratic decision schema can effectively navigate only simple interfaces while users with data analytic decision schema can navigate both simple and complex interfaces.

The software design community considers human factors engineering to be a source of bottlenecks in the computer interface design cycle. This paper suggests a way to shorten and improve the typically linear design process, without skimping on human factors engineering. It proposes a parallel design process based on the genetic algorithm of natural biological evolution. Although the genetic algorithm design process can take place within a facilitated group, it differs significantly from facilitation techniques in several ways: It is a parallel process, each design generation is evaluated against a usability fitness function, successive generations must include components from the previous generation, and evolution continues until convergence (consensus regarding the best design) is reached. The process is demonstrated here by a case study of the interface design for a corporate software program. Participants considered 40 designs and reached consensus on a final design in eight hours, thereby shortening the design cycle from two weeks to one day.