Heh, I am handy as a kind of party trick. But I often don't get invited back after I wreak my havoc. [grin]
I think the issue you're raising is fairly important, especially regarding zero intelligence agents. I tend to approach the problem with the following framework: Persistent processes build momentum by being historically dependent. Every iteration at every scale depends on the outcome of the last iteration at that and all the other scales. For any object that has a semi-permeable boundary (e.g. an animal), historical dependence builds up inside the boundary. That momentum allows it to barrel forward without regard to its environment. For example, an animal can plunge from a warm environment into a cold environment and maintain its warmth for quite awhile before its momentum bleeds off. The momentum often results in a bit of hysteresis. The mantra chanted by disbelieving members of your audience is akin to the creaking of a spring when you push or pull it too far. Systems with lots of little hysterical(?) processes can result in some interesting behavior even if there's no complexity or complicatedness within each little process. Hence, my recommendation would not be to start with anything complicated like tweaking the "quality of a decision process". I'd start with very simple, historically dependent processes. But there are 2 fundamental paths to take: 1) homogenous agents or 2) heterogeneous agents. I tend to think (1) is more interesting if you can provide an external source of variation. For example, place the homogenous agents in a varied environment so that their internal state is, initially, set by that environment. Then when they interact with one another, despite their being exactly the same process, their historical dependence will prevent them from syncing completely. It's interesting because with a tiny bit of, tightly controlled variation (and perhaps all determined), you can get strange behavior. But if your point is to _model_ some actual system, then you have to go with (2). There are simply too many sources of variation to tightly control where and when that variation enters the system. So, when modeling, you place enough of the right distributions to generate the phenomena you want to test against and then use validation data to guide the research. The language mismatch can enter with both (1) and (2). In (1), the mismatch is mostly in the sentences arrived at in the little formal systems inside the agents' skins. In (2) the mismatch can be very deep. Each agent can have its own formal system as long as the environment provides some sort of input into some part of the formal system, alphabet, grammar, or axioms. To search the archives, use the links Owen posted (nabble or mailman) or perhaps gmane: http://dir.gmane.org/gmane.org.region.new-mexico.santa-fe.friam And search for emergence, language, models, or whatever. Of course, that's all so ... fossilized. Near synchronous discussions are always more fun. But looking at the archives gives you a sense of the reaction you'll (probably) get to various hot-button words and phrases. Vladimyr Ivan Burachynsky wrote circa 03/22/2010 05:35 PM: > I wish I had you in tow at cocktail parties, I would pay to watch you argue > in public! What a hoot, it is like attending a Go tournament and having the > crap beat out of you with a style never seen before. I once got beat so bad > I laughed for weeks ( actually I still laugh 25 years later). > > Now personally I have often noticed that when engineering solutions to a > problem are required that I find discussion particularly frustrating. My > solution over the decades was to just build the damn thing and get it > running. After hearing how it was impossible, I would start the device or > run a simulation. However it only convinced about half the members of the > audience, the remaining half persisted in repeating the same mantra. This > reticence to believe the evidence before their senses always ticked me off > and certainly I was not amused and did not hide my contempt. > Some will believe others simply refuse. > > It is not simply language in this case. There is a twisted belief that the > symbolic statements uttered have more validity than the underlying concepts. > This seems to get into the realm of semiotics and I am just a dummy on such > issues. But these intransigent beliefs about meaning seem to condemn the > owner to disregard reality. Such a behavior seems absolutely contrary to any > evolutionary model. This makes me very curious indeed. > > Being a bit of a dummy, I used to play cards on weekends to pay for tuition, > I learned a lot observing the strange decisions players made about the > wagers before them. It struck me that all decisions are moderated not by > reason but by the momentary emotion dominant at that instant. A puff of > smoke in the eyes could make an individual switch from a good choice to a > bad choice. A buxom waitress could disturb choices. There was apparently no > method to weigh choices based on risk or returns. The choice appeared to be > made after an emotional reward had already been obtained. If a choice makes > one feel good then the individual will make that choice in real life and > damn the consequences. If emotions vary through a period of time the quality > of decisions also varies accordingly. Many individuals are satisfied with > their choice in spite of the wreckage at their feet. Conformal behavior > seems to fit the example, where the choice to conform is the reward while > the consequences are denied or are considered irrelevant. I don't ask people > what they were thinking anymore simply to avoid bad or degenerate arguments. > > My interest in Agents is in part about getting agents to vary the quality of > their decisions randomly or periodically. Hence my interest in stupid > agents. I suspect ( well more than suspect )that I personally have been > victimized by my own inherent stupid choices after a long life of tinkering > with machines and loose women. > > If we are all victims of flawed cognition then what do we do to protect > ourselves if even the brightest of us can not be trusted at critical > moments? As I got older I thought perhaps my judgment might improve but > clearly that was a vain notion and I still enjoy argument and women. Age > does not appear to lead to wisdom. > > Perhaps the flaw is deeper than language mismatch, perhaps there is no > reconciliation for an emotional biological entity and an abstract > intellectual fabrication or construct . I am just a dumb card counter, but > maybe the intellect is simply a delusion of a cunning mind. > > Knowing ahead of time I was going to be beaten in a game of Black Jack did > not make me feel much better about the loss of coin. Looking at gambling > addicts it always struck me as interesting how they persisted playing even > though the house was crooked. I have doubts that decisions are actually > based on monetary issues. > > The language trap might just be a symptom of a much deeper problem. The > lower intellect may not have an identity or self awareness and the concept > of self is constructed at a higher level based on language itself. Perhaps > the structure of the language/beliefs is the structure of the self identity. > > Back to agents, could an agent fabricate a belief system? Flawed or > otherwise... Would self awareness reside in the collection of disparate > belief systems? Is it conceivable that the self is nothing more than the > struggle between flawed belief systems in constant conflict? Perhaps the > Self is nothing more than the noise above a battlefield? > > As a footnote, the arguments about Language you mentioned, how do I find > them? > > Please excuse my manner, I get a little carried away with a good old > fashioned discussion, it makes me nostalgic for bad hamburgers and insipid > coffee at the campus dungeon. -- glen e. p. ropella, 971-222-9095, http://agent-based-modeling.com ============================================================ FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
