According to research on naturalistic decision making, proficient decision makers are recognitionally skilled. They are able to recognize a large number of situations as familiar and to retrieve an appropriate response. In these kinds of situations, they are able to act quickly and effectively on "intuition."

Recognition falls short, however, in unfamiliar, novel, and uncertain situations, where no pre-existing pattern fits. The classical point of view is that decision makers must then switch from intuition to analysis: They break the proplem down into components, make judgments regarding each component, and then put the pieces back together again by a mathematical formula that leads to an "optimal" solution. This analytical process is time consuming and demanding. There is often no more reason to trust the inputs to such an analysis as there would be to trust a direct judgment about the best decision. And the results are typically not in a form that can directly support concrete anticipation and planning.

According to naturalistic decision making models, on the other hand, analysis is not the only tool available for real-world decision making in novel and complex environments. Another key component of proficiency or expertise is the ability to adapt pre-existing knowledge to fit new circumstances. The Recognition / Metacognition model, developed by CTI, provides a general, systematic, and precise account of processes that decision makers use to understand and plan in such situations.

According to the Recognition / Metacognition model, proficient decision makers are both recognitionally and metacognitively skilled. Proficient decisions makers adopt a two-tiered strategy: (1) recognitional activation of expectations and associated responses, accompanied by (2) an optional process of critiquing and improving the products of recognition when necessary. Together, these processes build, verify, and modify mental models (or "stories") to account for a relatively novel set of events and to construct effective plans. Decision makers reflect on and improve the results of recognition, but do not reject recognition altogether (as in analysis-based approaches). Also by contrast to analysis, a soltuion is always available if time runs out: immediate action can be taken on the best solution so far. No matter how they evole, the decision maker always has a situation picture and plan together with an appreciation of its strengths and weaknesses.

We have developed a computational model of these critical thinking processes and have created and tested training and decision support methods based on the model.

As shown in the upper right box, the Quick Test is a control function that decides whether or not to act immediately on the recognitional solution or to think more. If the stakes are high, time is available, and the situation is unfamiliar or uncertain, then the initial recognitional response may be inhibited, and a cycle of critical thinking about the decision maker's mental model begins. Additional cycles of critical thinking occur until the cost of delay becomes too high, the stakes change, or uncertainty is sufficiently resolved. Then action is taken based on the current mental model.

CTI has conducted a series of experiments for NASA with commercial airline pilots. One of these experiments varied the size of the time window available for thinking about an important decision. More experienced pilots adapted their thinking processes to the available time. Less experienced pilots took about the same amount of time to decide regardless of the time window.

If time is available, proficient decision makers make use of that time to reflect on and improve the results of recogniton. Shown on right side of diagram, critical thinking involves looking for qualitatively different kinds of uncertainty in a mental model and dealing with each kind of uncertainty in appropriate ways. In critiquing and correcting, the decision maker may find gaps (not enough information to chose one conclusion or the other); conflict (reasons to chose both conclusions); or unreliable assumptions (reasons subject to alternative interpretations, which depend on unexamined premises).

Critiquing and correcting are iterative processes in which solving one problem can lead to identification of a new problem as understanding deepens. For example, filling gaps in information leads to conflict if the new information is inconsistent with current beliefs; resolving conflict can require potentially questionable assumptions about the trustworthiness of sources or the priority of competing goals. Decision makers continue to improve their understanding of the situation and plans until the time for action.