Computer Models as Thought Experiments

In his article, "Artificial Life as Philosophy" (Artificial Life, vol. 1, no. 3, 1994), Daniel Dennett writes:

There are two likely paths for philosophers to follow in their encounters with Artificial Life: they can see it as a new way of doing philosophy, or simply as a new object worthy of philosophical attention using traditional methods. Is Artificial Life best seen as a new philosophical method or a new phenomenon? There is a case to be made for each alternative, but I urge philosophers to take the leap and consider the first to be the more important and promising.

Philosophers have always trafficked in thought experiments, putatively conclusive arguments about what is possible, necessary, and impossible under various assumptions. The cases that philosophers have been able to make using these methods are notoriously inconclusive. What "stands to reason" or is "obvious" in various complex scenarios is quite often more an artifact of the bias and limitations of the philosopher's imagination than the dictate of genuine logical insight. Artificial Life, like its parent (aunt?) discipline, Artificial Intelligence, can be conceived as a sort of philosophy – the creation and testing of elaborate thought experiments, kept honest by requirements that could never be imposed on the naked mind of a human thinker acting alone. In short, Artificial Life research is the creation of prosthetically controlled thought experiments of indefinite complexity. This is a great way of confirming or disconfirming many of the intuitions or hunches that otherwise have to pass as data for the sorts of conceptual investigations that define the subject matter of philosophy. Philosophers who see this opportunity will want to leap into the field, at whatever level of abstraction suits their interests, and gird their conceptual loins with the simulational virtuosity of computers. ...

Artificial Life has already provided philosophers with a tidy batch of examples that challenge or illustrate points that have figured prominently in comtemporary philosophy. I anticipate that as philosophers acquaint themselves with the field and actively enter into its explorations, the philosophical progeny of the early work will multiply like fruitflies. After all, the field could hardly be better designed to appeal to a philosopher's habits: You get to make up most of the facts! This, as any philosopher knows, is perfectly kosher in a conceptual investigation.

Three crucial points can be emphasised here:

  • Computerisation enables a philosopher to investigate models whose working out requires calculation far beyond what would be remotely plausible without such assistance.
  • The philosopher, in implementing such a model on the computer, is "kept honest by requirements that could never be imposed on the naked mind of a human thinker acting alone".
  • Even thought experiments in which the initial assumptions are made up can be valuable, and even if these made up assumptions diverge significantly from reality.

The first of these enables us to move into entirely new territory of investigation by thought experiment, without getting lost in hand-waving and complexity. The second ensures that such a move, so far from leading to confusion and vagueness, is accompanied by greater intellectual discipline: computerised thought experiments – by their very nature – require explicit statement of all assumptions, and their outcome cannot be unconsciously moulded by wishful thinking. The third point, on the value of even unrealistic thought experiments, is discussed in the page on Computer Models as Existence Proofs.

Exploiting Our Own Evolved Intelligence

A fourth important point can be added to those made by Dennett. As investigators, we should be aware of our own evolved intellectual powers and limitations, which have been moulded by our evolutionary history of interaction with a harsh physical and social world, in which abstract thinking has until very recently played little role. The upshot of this is that we are "intuitively" far better at inductive thinking – spotting patterns in repeatedly observed phenomena – than we are at purely abstract, a priori reasoning, and we are likewise far better at "seeing" patterns literally – visually – than intellectually. Hence modelling thought experiments in a way that presents them as visual experimental objects, as things that we can play with repeatedly in real time and whose resulting behaviour we can see, has great cognitive advantages. This is closely related to Dennett's second point, in that without the computer, we have a strong tendency to lose the thread of complex thought experiments and unwittingly to smuggle in prejudices that influence our results. The complementary point is that this weakness in abstract thinking can not only be corrected by automating our thought experiments; it can even be reversed by exploiting our relative strength in more concrete, inductive thinking.

Daniel Dennett

Daniel Dennett

Philosopher of Mind, Artificial Intelligence, Language, and Biology