The
Kraal

Academe

Models and representations

Of course, if the thing is not structured it can still be represented. This may be one of the more difficult areas. The words we use to describe mental constructs are themselves vague. The definition of a model used above, for example, is only used for a mental model by analogy: the simple presence of the phrase "in three dimensions" might otherwise cause us concern.

If a mental model is not truly a model then what is it? "Analogy", "representation", "metaphor", "image", or the even more formal "simulacrum" are all weasel words, hard to pin down.

Mental model for a process operation

Bainbridge 88 states clearly what she believes to be the requirements for a process operation:

Two types of information organization, for:

• The structure and function of the process
• The goals and working memory of the operator

Two modes of representation:

• Predicate (language-like)
• Pattern

She begins by listing the supposed types of knowledge that the brain can deal with, and suggests that graphic information depends on pattern recognition and cued memory. The display need not be photographic (if representing a physical object rather than conceptual information) because people can recognize features not used in the display definition. She uses the example of a "square", defined as: four lines of equal length, joined at right angles (of course this definition is deficient - it equally well describes a cross or a zigzag), and says that observers could then see that the diagonals are equal length, and so on

She then talks of cues:

seeing a resistor, or a symbol for an 'and' gate, evokes a memory of its function

But of course this is also not necessarily true. Recognition depends as much on context as on the cue itself, and also, as she says, on training and experience, although she appears to believe that a simple breakdown into naive and experienced operators would be sufficient.

She then talks about how information can be represented. This is nothing to do with the operator's experience of the information. In particular, these logical groups enable the perpetrator to describe the whole of the connection, and to accurately predict the result of adding new nodes or connections. Human information processing just does not work like that. So what Bainbridge is talking about is a non-representational model, with severely restricted application, of operator's behaviour. She is not actually dealing with mental models at all.

She mentions that:

short-cuts ... appear to be an important aspect of human knowledge of cause-effect relations

arguing that missing out links in the cause-effect chain still demonstrates use of that logical chain: whereas I would be tempted to say that this is fact shows that operators (like most humans) are very bad at strict cause-effect chaining and prefer to rely on apparent, easily observed relations. This is partially acknowledged later:

Observation suggests that some people find it easier to remember and use interrelations in complex knowledge networks if they 'make sense'

that is, if some other simple relation is found. And although:

...reversing the cause-effect reasoning is not sufficient to account for the goal-directed behaviour that people show

she continues to attempt maps and diagrams that explain behaviour.

Later, she remarks that

It is evident that, with all these different types of interconnections between networks and networks of networks, the structure og knowledge is very complex

Again, the structure of her representations of knowledge are complex, but the thing she is attempting to represent is not necessarily complex but may be large combinations of simple processes

The best form for a representation gives cues to what the observer can be expected to remember or infer about the general properties of the situation

I would argue that the cues must cover all non-general cases. A non-general case is anything that would involve reasoning, because human reasoning is not merely bad, but it is only brought into play when all other relationships and short-cuts have been tried. Because it is laborious! I would suggest that the mental model follows the same form: it is a collection of simply-related parts that can no logical structure: in fact, it is not a model in any useful sense, but a hotch-potch of semi-conscious sensory impressions with no internal structure.

Continuing with her explanation, Bainbridge characterizes the skilled operator as knowing:

• Efficient methods for meeting the goal
• Optimum slots for working memory in a particular task
• Efficient categories: groupings of items and actions with common properties relevant to the task
• Short-cut links around the knowledge base
• Appropriate attribute values: knowing, for example, what will be the result of any method...

I have never encountered such a skilled operator, and I doubt that anyone has. To take each point in turn:

1. It is a truism that people stick to what they know. Having found a method that is congenial, a person will stick with it despite evidence that it is not 'efficient', ie. takes longer than it could, produces results that are worse than could be achieved. (Kieras & Bovair?) have a cost/benefit view of what we would in everyday life simply call conservatism
2. I cannot believe she has even said this. It is absurd in principle.
3. This is perhaps one of the two half-believable principles. People are determined to create simple relations wherever possible, insisting on them even when evidence and good sense points to the contrary. 'All blacks are lazy', 'all policemen are honest', 'all doctors are clever', and so on.
4. This is a piece of wishful thinking on Bainbridge's part. What she interprets as short-cut links, I would suggest are a variety of the groupings mentioned in 3.
5. Again, another variety of supposed inference from cause to effect, that can be explained by habitual groupings.Propositions have to be encoded and decoded... Central to this theme is attention. What guides it? How is what is attended to, encoded?

It seems to me that you can't begin to talk about modes of representation without discussing these fundamental problems first. Yet that's exactly what most people who use the terms "mental model" do.

Intelligence requires stepping away from signalled reality - to create internal representations, which can be richer and more useful than the available sensory data. (Gregory 95, p.13)

It was assumed that subjects' mental models were identical to an engineering model of the controlled system. (Rouse & Morris 86, p.357)

These two quotes show the start and end of speculation about mental models. For some, the mental model is a Platonic form with infinite possibility; for others, it is a convenient way of ascribing performance data to a person.

/Next Conclusion