|
The first question we ought to ask is:
How do mental models differ from physical models?
However, if the world that we perceive is constructed by our brain using the limited data provided by our senses, and modified by previous knowledge, we will never know if there is something "out there" or of what kind it is. So we must go further back and ask:
What do we understand by the terms physical, mental, and model?
To talk of a physical model, is, at best, to talk of a representation constructed from perception and moderated by conceptions. How is this similar to and different from a mental model?
What is meant by "mental"?
Mental. a&n; done by the mind [Concise Oxford Dictionary]
The crucial distinction is that it occurs in or is associated with, a mental event. The arguments for the existence of mental events are many and complex, and there is no consensus. But since Broadbent began to publish papers on attention in the 1950's (Broadbent 1954, 1958), mental events have become a legitimate area of study.
Are there mental events and objects?
Not only are there mental events and objects, according to the SRK model, but they are describable (Wirstad 88).
Are all mental events conscious?
The SRK theory implies that mental events occur unawares, in the same way, presumably, that physical events occur unawares - the operation of my lymphatic system, for example.
So "mental" does not imply "conscious", which is lucky, because no one would suggest that people operate using a conscious model.
However, unconscious objects are inferrable and sometimes introspectable. This is the basis of psychotherapy, (and is implicit in most psychological theory?). As long as you are not committed to a thoroughgoing materialistic and deterministic explanation of human activity, there is room for the unconscious.
How then can we access an unconcious mental model?
1. by using tests and procedures that claim to measure or infer unconscious objects and events
2. by using external techniques for putting the subject into a state where unconscious motivations are directly reportable
3. by asking the subject to attempt to bring unconscious objects and events to ordinary consciousness.
Method 1 is the least desirable, since it relies on verification by the other methods.
Method 2 is clearly an extremely hazardous procedure, ethically and methodologically. Three ways could be considered: natural, mechanical, and operable.
- Natural methods include self-induced meditation and trance.
- Mechanical methods include the use of drugs, strobes or brainwave stabilisers.
- Operable methods include hypnosis and the use of conditioning apparatus such as isolation chambers
Method 3 is the most desirable, but is the least controlled. Many people have experienced the hesitation and uncertainty that sometimes makes it hard to answer even the least demanding of questions about internal events. Even "how are you" may cause problems.
The philosophical objections to introspection are discussed most clearly by Boring 50.
However, the psychological literature is more concerned with other aspects of introspection. Many psychologists have argued that introspection provides data that is at least as reliable as other methods. Duncan & Praetorius 88 even go so far as to say that since verbal reports are data, they cannot be said to be incorrect, only unreliable.
As discussed in section xx, a mental model can be distinct entity only if it has at least some of the qualities associated with the meanings of "mental" and "model". What is left?
In fact, what is left is the common or garden sense of model - a viewable, 3D object - whose only odd characteristic is that it is "in the mind".
The arguments set out above appear to confirm that this is the only aspect of the mental model that is available for study: the introspectable experience of "viewing" a model. This is the mental model that I am aiming to reproduce.
Given that we allow for the existence of mental events, are all mental events conscious? If they are conscious, are they available to introspection?
Whether they are simply mental correlates of physical events or an entirely different form of existence is not relevant since I am concerned with the experience only.
For my purposes I will assume that mental events are simply differentiable from physical events.
The point is that the model is an experiential entity whose existence is not a direct result of immediate sensory input.
What is meant by "model"?
Example 1
Model. n; representation in three dimensions of proposed structure [Concise Oxford Dictionary]
This definition is somewhat more fine than we need. It should be sufficient if we simply assume that a model must in some way represent a structure.
There are two diffficulties: the thing being modelled must be structured, otherwise it cannot be modelled; and the representation must be structured, otherwise it is not a model. However, "structure" is not observable: it can only be inferred.
This makes the content of the tests crucial. What can we do to ensure that we can determine if a mental model is structured?
Example 2
The use of models relies on the basic paradigm of science, where observations and experiments are used to construct theories by which real-world events can be understood and predicted.(Wahlstrom 88, p161)
A model is used as:
- A means of communication
- An aid to understanding
- A tool for prediction and control
- A device for training
But to do those things, the model must conform to the basic paradigm, that is, it must be able to be seen as "the operational formulation of a theory" (Wahlstrom 88, p162): it must have explanatory power.
And while users of the term "mental model" rely on its predictive and explanatory powers, they cannot, and arguably could not, deduce or construct the theory underlying the model, from observing the model. In experiment x, I try to do so myself, with the results as described in chapter x.
The construction of models follows the same general scheme regardless of the system under consideration...
- definition of modelling goals
- selection of modelling approach
- separation between model and environment
- identification of submodels, variables and their relationships
- construction of the model
- validation of the model
- use of the model
(Wahltrom 88, p163)
So how many of these steps apply to mental models (when they are models in the mind)? Who sets the goals, selects the approach, decides where the environment starts, organises the representation? Difficult. Whatever the experience is, it cannot be an experience of a model on this definition.
Example 3
This is an example of shifting ground. In an article about Rasmussen's work on mental models, Hollnagel talks about replacing
conventional engineering/mechanical operator models with another type which, for convenience, were called mental models...(Hollnagel 88, p264)
These mental models are explicitly external representations of a user's cognition, But a few paragraphs later, he states that
human beings... acted on the basis of their understanding of the context and th especific task. In other words, they had a mental model of what they were doing...(Hollnagel 88, p264)
And then he goes on to say that that it became necessary to model the operator's mental activities. Models of models? No, just confusion, as he later points out. And confusion that is widespread, As Grant 90 even says in his introduction,
Within the study of Human-Computer Interaction (HCI) there is a substantial body of literature which uses the phrases `mental models', `user models', `conceptual models', etc. Confusion starts because there is no generally agreed single term which covers these phrases, and other similar ones. For brevity, we will exploit the ambiguity between models of the mind and models in the mind, and here use the term `mental models' as a general term standing for all of the models referred to, and the general models implicit in modelling methods, techniques and theories...(Grant 90, section 2.1.1)
Grant in fact covers the metal model literature in detail up to 1990, and his conclusions are that
in general, since humans have no standard of consistent representation, formalisms cannot capture 'it'. This is equally true of analysis, and of making models of human perfromance, as of design (Grant 90, section 2.6.2)
A different angle
Let us start at the the beginning: if people have mental models, how do they experience them? What makes the experience, experience of a mental model? Is a descriptive label for a distinct type of mental activity, or is a convenient term for a range of different experiences?
Is a mental model, a model of the experience, or the experience itself?
What is the difference between action guided by habit (non-intelligent) and action guided by mental models (intelligent)? [Rasmussen's is a model of the experience].
Let us say that the world and its events are a mixture of perceptions and conceptions: that perceptions are not recordings of the external world; and that conceptions are not necessarily systematically derived.
Gregory 95 suggests that human perceptions are predictive hypotheses:
...muscle force is set according to whether a perceived object is heavy or light - yet its weight cannot be monitored by the eyes. (Gregory 95, p.13)
...perceptions are but indirectly related to stimulus inputs from the senses, and that perception involves betting on what they may mean. So knowledge is very important. (Gregory 95, p14)
An individual's world is his own creation. His world differs from the worlds of other individuals because the components of his nervous system differ in qualitative and quantitative ways from those of other individuals, and because he has access to knowledge and to rules and algorithms that differ from those of other individuals.
Unfortunately, conceptual understanding is slow
When all is going to plan, as predicted by the running hypothesis of predictive perception, there is minimal or even no awareness. Consciousness seems to be given by mismatches between [the best bet] and the resulting behaviour. (Gregory 95, p16)
Pederson (88) has put the problem succinctly. He begins by discussing the problems of evaluating internal states, the phenomenal subject. He uses the traditional arguments: my pain is not your pain; the experience of my pain is not reducible to its description or with the brain processes that give rise to it. He then outlines Rasmussen's idea that distinguishes between the objective and subjective states by his use of the terms cause-effect (external) and means-end (internal). Internal states are not rational.
"If our mental attitudes could be defined solely as behavioural, then the cause-effect relationships would be a sufficient conceptual framework." (Pederson 88, p94)
He discusses beliefs, and, quoting Searle 83, shows that beliefs are dependent on pre-intentional and pre-propositional attitudes, whose existence is demonstrated if we try to spell out each mental state in the network of beliefs. We soon see that much of the network is hidden in the unconscious; that we cannot individuate beliefs: for example, we cannot count them. These pre-intentional and pre-propositional attitudes are the "mental capacities that do not themselves consist in Intentional states (representations), but nevertheless form the preconditions for the functioning of Intentional states" (Searle 83). Finally, he situates the problem:
The phenomenon we want to model is a person who acts in an intelligent or rational way in a complex variable world. He has a mental representation of the world ... We may, roughly, define his intelligence or rationality as the ability to create internal models of the world on which he can base his predictions and decisions. Such models consist of systems of beliefs which are embedded in the network and based on pre-intentional capacities (Pederson 88, p96)
He continues by describing why researchers have based their models of human cognition on the computing metaphor. However, as
many of our decisions, perhaps most of them, follow immediately from our perception of the world without any deliberation
there is a certain amount of obfuscation, as he puts it, involved in the usual assumption that "mental processes are analogous to or identical with algorithmic processes" (Pederson 88, p96).
It does seem odd that,
given that the human operator is an inherent error-maker (Sheridan 88, p149)
authors still seem to insist that people have a coherent reference point for their opinions, beliefs and actions. Even Hollnagel 88, while on the one hand making fun of those behaviourists who deny non-mechanical apparatus in humans, goes on to suggest that
human cognition is intrinsically purposeful (Hollnagel 88, p263)
But how closely is the conception of purpose bound with that of mechanism. Purpose implies means, and also rationality. And a rational system is arguably a mechanical system. Fuzzy reasoning is built into mechanical systems so that decisions can be made on half-truths and semi-certainties: but still the same circumstances will always give rise to the same result. It is possible to argue that, in the case of humans, the possibilities of influence are so huge that the same circumstances will never arise twice. I would prefer to believe that a sane person could make different, reasonable, decisions: and would make different decisions, due to mental states.
What is it that wills the existence of models for which there is little or no evidence? Norman would argue that people are operating with a mistaken or defective or incomplete model: but where is there evidence for anyone with a true, effective and complete model? Or are models essentially defective or incomplete? What use are they then? You can see the point of modelling something as a rational decision process: modelling blood circulation in humans, for example. The model is then intended to perform a species of simplification in order to clarify understanding. But who intends a mental model? Who gives it a purpose? What is it for?
As human beings, we are bound to look for (ie. create) pattern and form in what we perceive. That does not, however, imply that the apparatus of perception is rigidly mechanical.
“A trained and motivated decision-maker is usually supposed to behave rationally" (Wahlstrom 88, p164). But not only do "human decision-makers ... not follow the strategies as defined by the formal theories" but they cannot follow them, since formal systems demand that all variables are considered. In human interaction, the possibilities of influence are innumerable.
Wahlstrom discusses different types of mental model:
- concepts, relations and categories
- decision-task oriented algorithms,
- methods and strategies
The first type contains essentially the knowledge inherent in language and some of the models can be visualized graphically by semantic nets. The second type of mental model is connected directly to the decision-making task in terms of utility function, allowed actions and the system model. The third type contains more general algorithms and methods for problem solving.
It is all very well to demand that: but is it even possible? Are we then doomed to inefficient decision-making? I will argue: no, it is not possible to have "correct" models in any reasonable sense; however, efficient decision-making need not rely on correct mental models.
Children's mental models
Children's
memory and identification of [a] document is based on references to physical things and colours like 'castle', 'black', etc. No relationship between phenomena is referred to, and no situational perception of actions occurring on the cover is expressed [when searching for books using cover information] (Pejtersen 88, p187)
This is what we would expect: representation without coherence. It is unlikely that if they described a book, you could generate that book. Children's descriptions of things are notoriously anarchic and unstructured. I shall argue that adult's descriptions of things as coherent are the result of language habits. Experiment x shows that, on closer questioning, apparent structure in an adult's answer to simple questions disappears. The results are shown in chapter x.
Next Uses of a mental model
|
