International Encyclopedia of Systems and Cybernetics

D. MEADOWS et al describe as follows the main qualities of computer models:

"Rigour. The assumptions in computer models must be expressed completely and precisely; no ambiguities are possible… (while) Mental models are usually fraught with inconsistencies…

"Accessibility. Because all assumptions must be written out in order to communicate them to the computer, critics can examine, assess, and alter computer models. Mental models are unexaminable and uncritizable; their assumptions are not even entirely clear to the modellers.

"Comprehensiveness: A computer model can contain much more information than any single mental model and can keep track of many more interrelations at one time.

"Logic: If programmed correctly, the computer can process even a very complicated set of assumptions to draw error-free conclusions… Different people may agree completely about a set of assumptions and still fight about the conclusions to be drawn from them, if they have only mental models to rely on.

"Flexibility: Computer models can test a wide variety of conditions and policies (which is)… much less costly and time consuming than tests in the real world. Different assumptions can be tested more easily and openly than they could be in a single mental model" (1982, p.12-13).

→ Assumptional Analysis

The same authors recognize however that: "Those are potential advangages of computer modelling over mental models, but they are by no means always realized. Computer models can become so supercomplex and be expressed in such esoteric jargon that they are less open to critical evaluation than mental models. They are not logical error free if they contain undetected errors in programming, typing or interpretation. They are not more comprehensive than mental models if their mathematical properties require the omission of things that non-mathematical people know to be important. They cannot provide easy tests if each run requires several thousand dollars, several days of computer time, six months of preprogramming and postprocessing. Computer modelling techniques are also based on deep, unexpressed, technical paradigmatic assumptions that are not congruent with mental models and often constrain what can be included"(Ibid).

The most important of these constraints are probably the need to use binary logic in the modelling process and the impossibility to take care of simultaneous change and thus of the sensibility to multiple initial conditions, proper to complex systems.

As a result, computer modelling, when applied to highly nonlinear global problems as in the various Club of Rome world models, are prone to many difficult problems and controversial interpretations.

For a critique of "acts of faith that regularly appear in global models" and their computer representation, see the same authors (p.285). Their deeply felt honesty is to be commended.