I recently considered simulating meme propagation. I wondered what constitues a meme and how to encode it (to simulate it). I considered the following parts:
- Locations - A meme may say that something is/was at a certain place.
- Individuals - These are the actors/agents passing on memes. A meme may assertain something about an actor like that he is a at a certain place.
- Objects - Or Ressources. This is what actors are after I guess. We could say, that an actor must consume some type of object every some time to "survive" or "reproduce".
- Names - A meme may introduce a name for one of the above categories or for a meme itself. Before this is done, the above categories are identifiable only at the place of the actor. With this I mean, that I may know you (an individual) and I can introduce you to someone else and he will recognize you later, but I can only tell you about him later if I tell you his name (or sufficient properties).
- Movements - Individuals and Objects may move (optionally together). Memes may refer to such movements.
These parts can be related in the usual higher order logic ways (existential/all quantification, conjunction). As well as associated with a trust indication (individuals can pass on that some actor trusts some meme).
I suggest you start with a simpler set of primitives. BreitenburgVehicles? are quite instructive here - they can express a meme through a simple matter of sensor placement and latching. The primitives above are way too high level to be ontologically null.
- I don't think so. I agree, that these 'primitives' are high level, but thats intentionally so. I assume here, that these elements are sufficiently distinctive in e.g. the human mind (we have a spatial memory, a separate 'module' to recognize faces and people, and a mind to work on named concepts). This is no model to show how these distinctions come about, but what follows from them.
I postulate a length/complexity limit for a meme as well as a maximum number of memes held simulatneously (this could be set quite low for a simuation).
What memes can we build from such parts?
- object X is at Y
- object X is named Y
- object named Y is at place named Z
- the meme (object X is at Y) is named T
- if meme T is true, then individual Y is at place named Z (see time below)
- individual named X trusts, that individual Y trusts meme Z.
We see, that there is no need to explicitly introduce the concept time, because it is possible to refer to processes (someone or something reliably moves objects someplace). I assume, that a sufficiently complex simulation would quickly evolve time memes (as our ancestors did).
Why do you assume that? See http://ucsdnews.ucsd.edu/newsrel/soc/backsfuture06.asp for an obvious counterexample.
- Huh, thats no counterexample, but exactly proves my point. I must have badly explained my point: Time concepts are not ingrained (then I would have made them primitives) but memes themselfes (i.e. cultural products). I wouldn't have expected it at such a low level though. I had in mind something like solstitial events like midsummer, moon phases or equinoxesm, that can be recognized and named.
Now what are examples of these kind of memes?
If your memes began at a much lower level - down in distinctions (WhatsaDistinction) - then you might have something here. But as it is you just have some high level definitions defined in terms of some other high level definitions imho.
- Memes, that relate and name locations are called geography.
- Memes, that relate and name individuals are called genealogy (maybe).
- Memes, that relate and name memes for cause and effect actions of individuals are called law (simplified).
- Memes, that relate and name trust of individuals are called politics.
- Memes, that relate and name memes for processes of the simulation are called science.
- Memes, that concern trust are called religion (very simplified).
- add more here
- Actually thats the whole point of it. What else would you expect? It looks to me like you want to test/prove that memes exist, whereas I want to show what follow if they exist.
These memes can be correct (the statement made is correct) independent of there usefulness (a meme may say, that an indivudual trusts X which it doesn't, but if many individuals trust this mem it may be useful as a group effect).
How would you design experiments to test these hypotheses?
I would program a simulation environment, that has a limited set of locations (organized in some graph or fully interconnected). A very limited set of objects (say one type of 'food', one type of 'weapon', possibly also a predator or prey (to test coordination of individuals). The individuals have some hardwired set of behaviours which should include a desire to eat and maybe reproduce (we don't want to simulate coevolution of memes and genes yet). The individuals may hold a very limited set of memes or rather there should be a limit of the total meme size. Individuals keep only those memes, that they trust most. To track the trust we also have to associate each known individual with a trust value. This value is updated if memes got from that individual turn out to be useful or not. Also the trust in a meme is composed of the passed on trust assignments and the relative trust of the mentioned individuals. Trust is initialized on start or by reproduction (say). Individuals have an inference engine to derive actions from desires, memes and current environment (location, held objects, other individuals at this place). Obviously there should be some energy source (providing new food) and some way for mutation (otherwise the simulation may converge too fast).
Given a sufficient simulation of this kind I guess that
- Time memes turn up (because they enable coordinated action).
- Complex memes turn up (because they enable coordinated action too). I think, that they will not get as complicated as 'religion' or 'science', because that would require a much to expensive simulation will all those inference engines running.
- Lies turn up (deliberate wrong memes), because they enable parasitic behaviour e.g. on untrusted individuals (that are of no use otherwise).