Modelling the real world:
the uses and benefits of simulation
What are we to make of Richard Dawkins' computer programs to support his evolutionary ideas - in particular his biomorphs? What about the more sophisticated programs, such as Tierra (New Scientist, 22 February 1992, 36-39)? The results of the computer models appear to show the feasibility of evolutionary change. The most recent modelling work to gain publicity is that of Nilsson and Pelger on the evolution of the vertebrate eye, again apparently showing that these changes are not the problem that Darwin thought them to be. As someone who has taught courses on the theme of simulation modelling to graduate and undergraduate students, I have felt it necessary to consider more carefully the significance of the evolutionary models. This short article is to summarise one of my conclusions.
Techniques for modelling, or simulating, have transformed the study of complex systems. In most cases, simulation provides a means of investigating dynamic behaviour patterns: how the system varies over time. As computers have provided greater processing power, simulation techniques have been able to tackle more demanding problems, and many applications have been demonstrated. For example, in industry, large investments in machinery may be modelled to satisfy the purchaser that the installation will perform satisfactorily. In research, models may be created which simulate, for example, the evolution of populations of animals over time.
Of course, models can never be perfect representations of the real world. All contain simplifications which limit their accuracy. This must be the case, because even our appreciation of reality is imperfect - so computer models capturing some aspects of our understanding must be incomplete. Consequently, a major issue for modellers concerns validation: establishing confidence that the model is able to represent adequately the real world.
The validation process is illustrated diagrammatically below. The real world system is distinguished from the particular conceptual model of it which exists in the mind of the modeller. The simulation model is constructed as a further simplification of the conceptual model. A good modeller is alert to the simplifications built into the simulation, knowing that these constrain its applications. However, there are possibilities for the sub-conscious incorporation of simplifications into the model and also the introduction of erroneous ideas about the real world. Consequently, it is vital that validation work is undertaken - to establish correspondence with the real world. Validation is not checking the model against the conceptual model - as erroneous perceptions may remain undetected - but looking at model behaviour in the light of real world behaviour.
+---------+ \ +------------+ \ +------------+ | REAL | ========> | CONCEPTUAL | =======> | SIMULATION | | WORLD | / | MODEL | / | MODEL | +---------+ +------------+ +------------+ / / / -------------------------------------/ < V a l i d a t i o n < \ -------------------------------------\ \ \The problem with all the computer models used to portray evolutionary change is that the crucial step of validation is deficient. People are not asking: `Does this correlate with the real world?', but `Does this represent the theory?' The models portray conceptual models quite effectively - but they have not been validated! Evolutionary models have helped to visualise the theoretical concepts of Darwinistic evolution but, to date, they have made no contribution towards refining our understanding of the real world.
David J. Tyler (1995)
