This chapter distinguishes three types of performance levels. Slips and mistakes arise from different mechanisms. Slips from failures in execution (often occurring because of automatic procedural routines also called action schemas) and mistakes from failures in planning. There is a problem in this categorization because some errors possess properties from both.

Which errors?

Two big errors that illustrate a combination of slips and mistakes

Oyster Creek: The annulus level and the water level are usually the same, but this time they weren’t. They mistook the annulus level for the water level, which was very close to the fuel elements. Even though an alarmed sound 3 minutes after the error, it was not discovered until 30 minutes later.

Three Mile Island: The operators did not notice that a switch of the pressurizer was open; the panel light showed that it was closed. They only looked at the panel and did not think about the switch being broken or stuck.

The wrong appraisal of the system state looks like the property of a mistake, and the selection of the strong-but-wrong interpretations are more close to a slip-like failure. These type of error can be categorized as an inappropriate diagnostic rule, specifically the If (situation X prevails) then (system state Y exists) kind.

Using the by Jens Rasmussen distinguished three performance levels; skill-based (SB), rule-based (RB) and knowledge-based (KB). Using these three levels, there are three distinct error types created: skill-based slips, rule-based mistakes, knowledge-based mistakes.

How can we distinguish the performance levels?

The core distinction is whether or not the actor was engaged in problem solving. Skill-Based (SB) level involves no awareness of the current problem and is completely automatic and habitual. Routine-Based (RB) and Knowledge-Based (KB) are only triggered when the actor is aware of the problem he/she is facing. When the problem has an accessible solution it works on the RB level, but when it is harder to form is work in the KB level. When the problem is unfamiliar it is usually dealt with by trail-and-error.

One feature of SB and RB mistakes it feed-forward control. When it is a feature of SB it is a performance based on feed-forward control and depends upon a very flexible and efficient dynamic internal world model. When it is a feature of RB it is a performance that is goal-oriented, but structured by feed-forward control through a stored problem- solving rule. In may look like: If (X, Y, and Z are present) then it is (a B situation) or If (A) then (do C).

Control at KB level is mostly through feedback. This is necessary because at the KB level one can’t rely on their mental schemas and habituations anymore. It can be achieved by setting local goals, initiating actions to achieve these goals and then reflect if the action was successful or not.

There is a relationship between the predictability of error and the level of expertise. The more skilled an individual is, the more likely it is that their errors will take the strong-but-wrong form on a SB or RB level.

Humans are better at solving problems than computers. That is because if we exhaust our problem-solving skills we search for analogies that can help us solving the current problem.

The best example of coexistence of these three levels

Driving is the best example of the coexistence of the three levels. For example steering, changing gears and speed control are operated on a skill-based level. Dealing with other road users is an example of a rule-based operation and the knowledge-based level only shows if you for example unexpectedly have to change your route because of work on the road.