Models of Operators as Components in Complex Systems

The following description is primarily limited to the role of the operator as a system component. First it is helpful to examine the operator’s ability to fulfil the goals of the system. Of course, many other factors influence the operator and thus the operator’s work. The operator as a contributor to, as well as a part of, the complex process control system is only one aspect of the possible influencing factors. As a social being, the operator is influenced by background, leisure, and family circum­stances, for example. This more complete view interacts in different ways with the individual’s role as a system component (this interaction is discussed in Chapter 8).

The classical way of describing human information-processing abilities often starts from a model such as that shown in Figure 2.17. This type of model is based on descriptions by, for example, Welford (1968) or Ivergard (1981). The opera­tor uses mental processes to convert the various sensory impressions and signals from the environment. These signals are structured in what is known as perceptual

Models of Operators as Components in Complex Systems

FIGURE 2.17 Human-machine system with a highly-simplified description of the human operator where, for example, the various forms of feedback mechanisms have been omitted.

organisation whereby the primary signals are converted into some form of mean­ingful whole, for example, letters, pointers, or what they indicate. Based on these structured signals, the operator makes different types of decisions. The decisions are taken on the basis of various forms of stored ‘set values’ (memory) of what the operator feels the signals should actually be in each particular case. The decisions have an effect on the ‘motor organisation’ in the cortex. This forms a ready-made and stored programme in the human brain capable of carrying out various types of motor activities under different circumstances. The motor programmes then activate groups of muscles in predetermined sequences.

Using the above model as a starting point, it is possible to differentiate diverse types of skills. The ability to control some parameters directly ‘online’—for exam­ple, to hold a car on a winding road—is called sensorimotor skill. A typical example of a perceptual skill is the ability to understand a three-dimensional world when represented on maps or radar screens. Perceptual skills also include many forms of inspection work in industry, for example, inspection of electronic circuit boards, as well as engineers listening for faults in a diesel engine. Many types of perceptual skills involve what is known as vigilance work situations. This means that the sig­nals to be detected, such as faults in the process, are very few, unclear, and occur randomly. Vigilance skills are thus a special form of perceptual skill.

It is typical for all types of skills that the ability to carry out a task correctly is developed through training. However, in order for the skill to develop, training must often be of a practical nature. It is not sufficient just to know how to do something, but one must also be able to do it in practice. It is possible using written instructions to explain, for example, how a radar picture should be interpreted. On the other hand, the ability to read a radar picture correctly is only developed after many years of work actually reading radar pictures. Another typical example of a perceptual skill is that of lookout on a ship. Practically always, the older seaman will see other ships earlier than the inexperienced one. This may be partly explainable by differ­ences in visual ability, but is probably primarily attributable to a better developed perceptual skill.

When information processing is mainly concerned with decision making—that is, cognitive tasks—this is referred to as cognitive skill. Here, the degree of diffi­culty of perceiving the incoming signal may be relatively simple. The difficulty lies in manipulating the incoming signals suitably, putting these in the correct context, and drawing the correct conclusions from them. This becomes a question of inter­action partly between different parts of the long-term memory, and partly between long-term and short-term memory. This is the case, for example, where an operator in a control room has to interpret correctly a particular reading on an instrument. A change in a reading may depend on many parameters. A skilled operator has a picture of how the process works stored in his or her long-term memory. Faced with incoming signals about changes in the process, the operator continuously updates the mental picture. In any situation, there may be many sensory inputs that contrib­ute to this updating. For example, there may be signals from different instruments in the control room, or the operator may actually have noticed at firsthand the various changes that occur in the process (such as increased wear, beginnings of leaks, and so forth). Skilled operators build up a detailed and accurate mental picture (pro­cess model), which allows them to understand clearly the meaning of the different changes shown on the instruments and to make appropriate decisions accordingly. Planners in different fields work in a similar way. A subtle difference is that, instead of having a model of an existing reality which is continuously being updated, they have a model of a conceived, planned reality which may become modified in differ­ent ways. The work of planners is often very abstract. However, as human beings have a very limited ability to work solely on an abstract plane, so planners often get help by relating the newly planned system to some known and already existing system. This limitation in the human ability to work with abstractions is a barrier to the creativity needed to find new and unprejudiced solutions to problems. As human beings we find it much easier to extend ideas from existing concepts than to invent new concepts.

In classical manual crafts, such as those of blacksmiths, shoemakers, welders, glassblowers, and so on, motor skills are of the greatest importance. It is typical of a well-developed motor skill to have a very well-developed programme of different patterns of movement stored in the motor organisation mentioned previously. In high levels of motor skill, it is usual not to have to depend on seeing the effect of what one does directly but to have some form of stored feedback mechanism, which goes straight back to the motor organisation centre and corrects and adjusts the movement pattern to suit the actual circumstances. A typical example of a motor skill is that of typing on a keyboard. At a low level of skill, a person looks for each key and then presses it. The motor organisation in this case is fairly simple. The typist has a men­tal model that directs the fingers to the correct keys. At a very low grade of skill, the eyes are used to steer the finger onto the correct key. As motor skill increases, less visual support is required in order to find the keys. At a highly-developed level of motor skill, a person develops a ready-made mental model to press not just individ­ual keys but whole groups of keys. Relatively independent of the bodily position of the operator, automatic matching of the muscles occurs in order that they will move correctly and hit the correct key or groups of keys. It is often said that skilled practi­tioners can do their skill ‘blindfolded’. Highly-developed feedback mechanisms are found, for example, in welders or in potters where mental models are converted into physical movements.

Development of a high grade of skill is a necessity if a craftsman is to work rapidly. If there is total dependence on perception and the perceptual organisation and its decision-making process, it would take the craftsman longer to complete the task, and the movement patterns would lack precision and efficiency. A human being is commonly regarded as a one-channel system, and since decision making is a sequential mechanism and only one decision at a time can be handled, this creates a bottleneck. However, the motor organisation with its attendant feedback mechanisms can handle several simultaneous operations in parallel—despite a human being only being able to make one decision at a time.

Rasmussen (1980) suggested that there should be a similar mechanism for paral­lel information processing on a human’s perceptual side before the decision mecha­nism is engaged. Rasmussen developed a very useful model where he describes the control room operator as consisting of a conscious part together with an unconscious process with particular emphasis on the importance of the unconscious dynamic world model. The model is shown in Figure 2.18. Rasmussen considers (in common with Crossman, 1965), that operators have a dynamic mental model of the actual process stored in their long-term memory.[4] This mental model is updated uncon­sciously by external signals and internal changes, that is, by the actual process. When monitoring the process, some form of synchronisation continues throughout; that is, the operator uses an unconscious plan to check whether the mental model agrees with how his or her perception interprets the actual process. If there is any form of error in this synchronisation—that is, that the interpretation of the real-life model does not coincide with the mental model—a signal will be sent to the conscious part of the brain. Based on this model, which has certain superficial similarities to the traditional one of the human as an information processor (see Figure 2.19), Rasmussen and Lind (1982) developed different categories of human behaviour or performance level (sometimes using the word ‘behaviour’ and sometimes ‘level of performance’):

1. Skill-based behaviour

2. Rule-based behaviour

3. Knowledge-based behaviour

Simple Actions —- *-

Models of Operators as Components in Complex Systems

FIGURE 2.18 Schematic map of the human processing mechanism. (Rasmussen, 1980. With permission.)

There is no complete congruence between these different types of ‘behaviour’ and the different forms of skills presented earlier. As a rough guide, what Rasmus­sen calls skills-based behaviour corresponds to motor skill and sensorimotor skill; knowledge-based behaviour corresponds approximately to cognitive skill; there is no direct equivalent to rule-based behaviour. On the other hand, Rasmussen discusses a form of skill that is efficiently used to update and synchronise the mental model of the real-life process. This function is probably dependent on a highly developed form of perceptual skill, perhaps in combination with some form of cognitive skill. It is difficult to see the psychological/physiological basis for Rasmussen’s division of human behaviour into different groups. Existing bases for division of behaviour in different types of skills are probably a better basis for classifying behaviour.

It is common in connection with job training to talk in terms of factual training (which leads to knowledge) and skills training (which leads to skills/abilities). To a certain extent, these two types of training could be seen as stages in the development of skill, where one first obtains knowledge of the particular situation—for example, the rules and norms (or knowledge training)—and then engages in some form of

Models of Operators as Components in Complex Systems

practice in readiness to use these skills. There is agreement to some extent between these two bases for division and Rasmussen’s division into skill-based behaviour and rule-based behaviour. Skills-based behaviour therefore corresponds to a form of behaviour that is achieved after a longer period of training including readiness training, while rule-based behaviour is achieved after only knowledge training. A wider discussion of these different forms of training can be found in, for example, Ivergard (1982).

Rasmussen (1979a) used the division of behaviour into different groups to some extent as a basis for building up a decision ladder (see Figure 2.20, which is a sche­matic plan of the sequences of different types of processing from initiation to manual action). This decision ladder has been found to be a valuable aid in structuring and describing control room tasks. Rasmussen and his colleagues have used the method, for example, in collecting data with the aid of verbal questionnaires.

In a project on Computerisation in the Process Industry carried out by Ergolab (Ivergard, Istance, and Gunther, 1980), an attempt was made to use this form of deci­sion ladder, or information-processing ladder, retrospectively in order to describe the various tasks and their parts that were observed. In itself the decision ladder proved to be a useful tool, but it was not possible to use this to generalise between different types of control room work. The ladder was similarly of little use for describing the task that takes the longest time for the control room operator—namely, monitoring the process and updating his or her own mental model of the physical reality. These cognitive processes are in many ways the most important part of the job, as they enable the operator to prepare and create the basis for coping with disruptions in the process. Neither method describes ‘skills-based shortcuts’.

Models of Operators as Components in Complex Systems

FIGURE 2.20 Sequence of operator mental activities. (Rasmussen, 1980. With permission.)