In this section, we present three theoretical perspectives that serve as the basis for theory that explains the role of the application domain in IS problem solving. The theoretical framework for establishing roles for both IS and application domain knowledge is provided by research in cognitive psychology that examines problem solvers engaging in the simultaneous solution of two tasks. Formulating problem solving in IS as a dual-task model, that is, with tasks in each of the IS and application domains, allows us to consider situations in which a cognitive task in one domain has different types of influences on the performance of a cognitive task in the other domain. We present the theory of cognitive fit (Vessey, 1991) as the theoretical basis for determining what happens in such circumstances. Finally, we present theory on the structured nature of the problems under investigation, which we propose as a contingency factor in establishing cognitive fit between the dual tasks.
Following an introduction to the cognitive psychology literature on the simultaneous solution of two tasks, we present a model of dual-task problem solving as a way of thinking about the interrelationship between the two tasks.
Cognitive psychologists have long investigated what the community calls ‘dual-task interference’, a phenomenon that occurs when problem solvers perform two tasks in rapid succession. It is manifested in performance degradation on one or both of the tasks (see, for example, Durso et al.,1998; Koch and Prinz 2002; Navon and Gopher, 1979; Pashler, 1994; Van Selst and Jolicoeur, 1997; Wickens, 2002). When dual-task interference occurs, it is difficult for the individual to allocate attention effectively between tasks, resulting in reduced performance.
Much of the research in the area has focused on the resources needed to conduct the two tasks simultaneously, and therefore the allocation of resources between them (see, for example, Durso and Gronlund, 1999; Kahneman, 1973; Wickens, 2002) and the likelihood of a processing bottleneck (Pashler, 1994; Pashler and O’Brien, 1993; Van Selst and Joliceour, 1997). Although there is still substantial debate regarding the mechanisms that underlie the phenomenon, the effects have been observed consistently (see, among others, Koch and Prinz, 2002; Navon, 1990; Navon and Miller, 1987; Pashler, 1994; Pashler and O’Brien, 1993; Van Selst and Joliceour, 1997; Whitaker, 1979).
In this research, we apply the basic premises of research on dual-task interference to our specific context of IS problem solving in which interaction occurs between the IS and application domains and therefore between tasks in each of those domains. Under these circumstances, we propose that dual-task problem solving does not always lead to dual-task interference, and we address the circumstances in which the simultaneous solution of the two tasks does lead to dual-task interference and when it does not. We therefore use the term dual-task problem solving in our current analyses, rather than dual-task interference.
As we have seen, IS problem solving consists of solving problems in a variety of application domains, and therefore knowledge in both the IS and application domains may play a role in problem solution. The basic premise of our theoretical model is, therefore, that tasks in each of those domains must be solved to reach a solution. We present a dual-task problem-solving model as the framework for examining the interrelationship between the two types of tasks.
Figure 1 presents the dual-task problem-solving model that describes the cognitive process involved in solving a problem in which two types of relevant knowledge interact. This model is based on three repetitions of the basic problem-solving model used to describe cognitive fit (Vessey, 1991), extended to include the notions of distributed cognition proposed in Zhang (1997) and Zhang and Norman (1994). One problem-solving model is used to describe each contributing cognitive task, shown in dashed boxes in Figure 1, with a further model for their interaction.
Problem solvers first form mental representations for each contributing task, that is, the cognitive tasks of understanding the application domain (developing a mental representation of the application domain) and the IS domain (developing a mental representation of the IS task). They must then integrate these two representations into a mental representation that facilitates task solution (the mental representation for task solution).[1] Each contributing task is supported by an internal problem representation (knowledge the problem solver has of the domain of interest: IS or application) and an external problem representation that presents explicit knowledge related to the solution of the task in the IS domain.
Formulating IS problem solving as a dual-task model opens the way for us to consider situations in which one task might either facilitate or inhibit the other. The theory of cognitive fit that is used as the foundation for the dual-task problem solving model therefore serves as the basis for a theoretical analysis of the circumstances in which application domain knowledge facilitates problem solving, and those in which interaction between the IS and application sub-tasks results in dual-task interference, thereby inhibiting performance.
Here we present the basic notions of the theory of cognitive fit (Vessey, 1991). Although more complex forms of cognitive fit have now been identified (see Vessey, 2006), the theory of cognitive fit is most simply explained in terms of its original formulation as the performance effects resulting from matching the external IS problem representation to the IS task to be solved (Vessey, 1991). A match or cognitive fit occurs when the information emphasised in a particular external IS problem representation matches that required to complete the type of IS task under investigation. Decision making is facilitated because the problem-solving processes used to act on the problem representation are similar to those needed to solve the problem.
Using decision making using graphs and tables as our example (see Vessey, 1991) ‘symbolic’ tasks such as determining train departure and arrival times, which involve discrete data values, require the use of analytical processes and are therefore best supported with external IS problem representations that also require the use of analytical processes. In this case, such tasks are better supported using tables (symbolic formats) than by graphs. On the other hand, ‘spatial’ tasks such as determining the relationships among the performances of a number of sales regions, which involve making associations or perceiving relationships in the data, require the use of perceptual processes and are therefore best supported with external IS problem representations that also require the use of perceptual processes. In this case, such tasks are better supported using graphs (spatial formats) than by tables. Note that when problem-solving processes match, the decision maker is effectively guided in reaching a task solution.
Alternatively, when the type of information emphasised in the external IS problem representation does not match that emphasised in the IS task, there is nothing to guide the decision maker in working toward task solution, and they must exert greater cognitive effort to transform the information into a form suitable for solving that particular type of problem (Vessey, 1994). This increased effort will result in decreased performance (that is, decreased decision accuracy, increased decision time, or both).
We propose that the types of interactions between the tasks in the IS and application domains differ depending on the nature of the problem under consideration. The aspect of the problem that is key in these circumstances is whether it is well- or ill-structured. We distinguish different ‘fit’ situations based on whether the problem to be solved is well-structured or ill-structured in nature (see Reitman, 1964).
Well-structured problems are those that have a well-defined initial state, a clearly-defined goal state, a well-defined, constrained set of transformation functions to guide the solution process, well-defined evaluation processes, and a single optimal solution path (Greeno, 1978; Sinnott, 1989; Voss and Post, 1988). Further, the information needed to solve the problem is contained in the problem statement.
On the other hand, ill-structured problems are those for which the initial and goal states are vaguely defined or unclear (Voss and Post, 1988), and for which there are multiple solutions and solution paths, or no solution at all (Kitchner, 1983). Further, with such a problem the problem statement does not contain all of the information needed for its solution; hence it is not clear what actions are required to solve it (Chi and Glaser, 1985).