As a scientific concept, fit has its origins in the biological sciences where the emphasis is on understanding the process of fitting in order to understand the fit that emerges. Evolutionary selection is viewed as a process of survival of the fittest. In this evolutionary context, the explanatory power of fit is not so much in the outcome of the process of fitting (i.e. selection of some biological feature of a species) as in the process itself; that is, in how the biological feature came to be selected in the evolutionary process (see, for example, Dawkins, 1982). In a human behavioural context, as opposed to a purely biological context, ecological psychology has explored the dynamic and emergent nature of fit (e.g. Kochevar, 1994). The definition of fit presented earlier can be shown to be entirely consistent with the concept of dynamic fit in ecological psychology.

In ecological psychology behaviour is the product of the interaction between an individual and the environment. Gibson (1979) coined the term affordance to refer to the possible actions that may result from this interaction between an individual’s knowledge and the properties of the environment. A situation can afford a particular action for an individual with appropriate knowledge and abilities, and an individual can have the knowledge and abilities to carry out a particular action in an environment that affords such actions (Greeno et al., 1993). As Kochevar (1994) puts it: ‘Environments provide information structured to support specific behaviours, and adapted individuals are sensitive to such information patterns’. Thus, the concept of an affordance is concerned with the complementarities between an agent’s knowledge and abilities and the features of the environment (i.e. the information it provides and actions it supports). This notion of complementarities is essentially one of fit, but ‘fitness’ for what purpose? A given environment may afford many different actions for agents of even limited abilities. Affordances do not determine action; they merely define the set of possible actions available to a given agent in a particular environment. Newell’s Principle of Rationality, discussed earlier, provides the ‘filter’ for action selection: an agent will select the action (realise the affordance) that appears to best attain his or her task goals. Indeed, Heft (1989) states that the affordances an individual perceives in a given environment are determined by his or her intentions or goals. Notably, the other aspects of task, information cues and required acts, relate to the affordances themselves — what information the environment provides and what actions it supports.

Ecologically, what a situation affords an individual at a given point in time depends on the fit amongst the individual’s knowledge and abilities, the actions supported by the environment and the information it provides, in the context of the task goals that are present. Thus, the earlier definition of fit as the degree of the correspondence between an individual’s knowledge and the structure and features of the environment that specify supported actions and information, in the context of a specific task goal, is entirely consistent with ecological psychology principles. The complementarities between the environment (information cues provided and actions supported) and an individual’s knowledge define the set of all possible actions the individual may take in that environment — the affordances. The task at hand (goals, required acts and information cues) determines what affordances an agent perceives in that environment at a given point in time. More specifically, the task goals serve as a filter in the selection of the appropriate action to take. At least implicitly, this filtering reflects the degree of fit amongst the environment (information provided and acts supported), the agent’s knowledge (ability to use information and carry out acts), and the demands of the task (goals, required acts, and information cues — of which goals is the most determinative element since equifinality implies substitutions may be possible with respect to acts and cues).

This filtering is not a cognitively complex task. Rather, ecological psychology argues the somewhat extreme position that individuals ‘directly perceive’ the task relevant affordances in an environment. As Gibson (1979) notes:

Ecologically speaking individuals engage in a continuous perception-action cycle as they seek to maintain the fit between their knowledge and the environment in their attempts to satisfy specific task goals — fit is dynamic and task specific. Maintaining fit is a process of becoming sensitised to the affordances in a given environment and task context, and of fine-tuning this sensitivity. Problem solving behaviour can be characterised as ‘gap closing’ (Lave, 1988); attempting to improve the fit between the individual’s knowledge and the environment in the task context. Gap closing involves taking the processes used in the past to handle similar classes of problems or tasks and iteratively manipulating them until the present task or problem can be accomplished or resolved.

Neisser's (1976) Perceptual Cycle characterises this gap closing process (see Figure 3). The individual has some knowledge of the environment based on past experiences. This knowledge drives exploratory action in the actual environment; for example, behaviour driven by a crude notion such as ‘my experience suggests that this [action] usually fixes problems like this’ (Orr, 1990). Feedback about the success of the action relative to the goals results in modifications to the individual’s knowledge (their sensitivity to the affordances of the environment is fine-tuned) that then drives further action. This process iterates until the individual is satisfied that his or her knowledge of the environment and the actions that derive from it adequately address the problem (i.e. the task can be completed). Satisfactory fit is achieved. Importantly, this conception of fit allows for new and unique combinations of actions to emerge (Kochevar, 1994). In AST terms, it allows for novel, unique and unexpected or ‘unfaithful’ appropriations of a technology environment.

Neisser’s perceptual cycle is essentially a model of learning. Learning is thus an integral component of a dynamic, ecological view of fit. As an individual learns more about the environment, his or her fit with it changes and the affordances (possible actions) available change. Old ways of doing things are supplanted by new and better ways. However, in technology supported work environments, there are multiple types of fit, as there are multiple knowledge agents (user, technology-as-tool, technology-as-representation as per Table 1). While an individual may be able to ‘directly’ perceive affordances and consequently their fit with an environment for a given task, the assessment of fit for an agent other than themselves (e.g. as in representation-reality fit) is clearly a more cognitively complex task.

From a cognition perspective, two information processes influence an individual’s judgments of fit: inference from past performance, and prospective analysis of available affordances in light of one’s goals. Both of these processes can introduce errors into fit judgments. For example, performance feedback is often delayed in real world environments. Moreover, it is also difficult for an individual to causally separate performance variations due to actions taken from those due to natural variation in a stochastic world. Indeed, prior research has shown that human decision makers do not cope well with dynamic feedback systems, particularly when there are delays (Diehl and Sterman, 1995; Sterman, 1989a; 1989b).

Inferring fit from a goal-based consideration of available affordances as a form of prospective thinking — evaluating the desirability of realising an available affordance without actually taking the action — is also subject to error and bias as it is likely to be carried out heuristically (e.g. Tversky and Kahneman, 1974). The problem is exacerbated when a user is inferring fit when the technology is the knowledge agent. For example, assessing representation-context fit requires that the user have knowledge of what the environment affords the representation. Consequently, this implies that the accuracy of subjective judgments of representation-reality fit by a user is influenced by the degree of user-representation fit.^[3]

Judgments of fit, whether directly perceived or otherwise, drive individual learning and motivate efforts in task or systems refinement and development. Learning in this dynamic view of fit can be proactive, reactive, or passive. Proactive learning aims at improving fit by making available to the individual new affordances. It is driven a priori on an assessment of how well the current affordances correspond to the individual’s goals. Reactive learning occurs in response to dissatisfaction with the performance evaluation of actions taken. Learning can also occur passively as an individual acquires knowledge in a non-purposeful manner while interacting with the environment. In the information systems context, proactive learning could occur when a user works through an interactive tutorial before attempting a task with a new piece of software. Reactive learning could occur through resorting to the tutorial after having difficulties carrying out a task with a new piece of software. Passive learning could occur when a user discovers new ways of using a software application through casual observation or interaction with another user — it is merely an incidental outcome rather than a deliberate goal of the user.

Judgments that fit is unsatisfactory can also lead to efforts to change the environment to improve fit; through system refinement and development, for example. Such efforts could be as radical as a major hardware or software upgrade, or as simple as a change to the toolbars displayed in a word processor. Importantly, these changes may also occur exogenously if they are imposed on users rather than instigated by them.

Since users’ inferences about the different types of fit can become biased, their subjective judgments of fit may differ substantially from more objective measures of fit. In the IS context this is problematic both for the user who seeks to maximise job performance through IT and for the researcher or practitioner trying to evaluate a system. More specifically, misjudgements that fit is good, when it is not, reduce the motivation for learning (as per the perceptual cycle) or to refine a system to improve fit. In a similar vein misjudgements that fit is bad when it is actually good (such as may occur in highly stochastic environments) could lead to unwarranted efforts at system refinement or task learning.