Why Simulation

Angelo Cangelosi and Domenico Parisi, who together devised the lovely mushroom experiment into altruistic communication I tend to reference, have collected relevant research papers into a book called Simulating the Evolution of Language. I've read some of them before, but not Kirby-Hurford's Iterated Learning and a promising one on symbol-grounding (an area I need to rethink).

Anyhow, their introduction involves an analysis of simulation as a scientific tool. It's an interesting question, one I find my approach founded upon. Where will my results be relevant, and where will they find limitation? Can I mention evolution of language if I don't actually address it?

The nascent discipline of artificial life has yielded hope that computer simulations will produce complex grammars and natural language through emergent behavior. A key criticism of this approach is that even if successful, this insight cannot be directly correlated with the development of human language in esse. Notwithstanding, there are tangible benefits blahblahblah

My proposal thus avoids the question, but here I've found it properly addressed.

First, the appearance of language is a one-time event. We won't see it again, developing from linguistic nothingness. Second, we'll never be in a position to satisfactorily recreate the original conditions from genetic and/or fossil evidence. It was a long time ago, and there are too many potential factors to consider. Social conditions, genetic disposition, initial intent, cultural dispersion... Research from subjects like physical anthropology might endorse particular theories, but I don't see them driving the debate.

"Simulation," however, in the words of Cangelosi and Parisi, "is the implementation of a theory in a computer...

First, if one expresses one's theory as a computer program the theory cannot but be explicit, detailed, consistent, and complete because, if it lacks these properties, the theory/program would not run...
Second, a theory expressed as a computer program necessarily generates a large number of detailed predictions because, as we have said, when the program runs in the computer, the simulation results are the predictions (even predictions not thought of by the researcher) derived from the theory...
Third, simulations are not only theories but also virtual experimental laboratories. As in a real experimental laboratory, a simulation, once constructed, will allow the researcher to observe phenomena under controlled conditions, to manipulate the conditions and variables that control the phenomena, and to determine the consequences of these manipulations..."

I really like this idea. Equally well-put, however, are some of the criticisms of simulation-based resesarch. Simplification is dangerous, but as George Box said, "all models are wrong, but some are useful." Careful simplification is a scientific sine qua non. Trouble with external validation sucks, but in this case, the research isn't yet at that point. Maybe when enough work has been done, we'll start looking for supporting evidence in fossil finds, but right now the research is more about machines than people.

Also mentioned is arbitrariness of assumptions and experimental parameters. This, for me, is the crux. After much thought, I still can't put myself in the abstract position of my agents; is there too much to consider, will they be overwhelmed? Or starved of stimulation? In this respect, you can see why researchers turn to robots; the information is tangible and the details accessible. And assumptions! In pencil, we connect dots we draw ourselves.

Hand-holding was the bane of my proposal, and now it blossoms into the bane of my project.