In the language of computer programming: You have a collection of “input variables” representing the state that the modeled person is in. These variables may be things like “hunger” or “thirst” etc. Then you have a collection of “output variables” representing the actions the person may take, like “buy food” or “buy drink”. Then the idea of the computer program (ABM) is to translate the set of input variables into output variables in a way that appears to match observed human behavior as accurately as possible. The model may contain multiple agents and may also continually adjust the input variables according to the actions taken. For example if the “thirst” variable was TRUE (or perhaps some value on a sliding scale) and the action taken was “buy drink”, then the model may now change the thirst variable to FALSE and now run the simulation all over again… etc. etc. Obviously this is just a cartoon example - in practice these models could contain all sorts of subtleties and complexities.
You can model what would happen if someone did prefer it or if they did not, but you cannot predict individual preferences precisely.
I’m not sure this is what mick is saying. ABM’s are not about predicting the internal states of the agents involved in a model, it is about presuming a given set of internal states and observing the system-wide outcomes when those agents interact. Andris put it well when he said;
ABM’s goal is to see what complex behaviors result from interaction of rather simple (or at least understandable) actors. So if you remember that behavior of individual actors is just an assumption, I think ABM is fully compatible with AE. The result of ABM would be not “market behaves like that” but “assuming actors behave like this and that market behaves like that”. And of course your model can have as many different actor types as you wish (and you simulator is able to handle).
Here is an interesting ABM to illustrate what I mean by the type of model that I believe is applicable to austrain economics. It operates in exactly the way Andris describes, and has a number of interesting emergent system behaviours that could not be predicted by thought experiment alone.
I just noticed I made a mistake - when I said “These variables may be things like “hunger” or “thirst” etc” - what I should have said is These variables may be things like “have consumed food recently” or “have consumed drink recently” etc. The idea of “thirst” or “hunger” are not strictly speaking inputs - but instead may be used inside the model.
I’ve been a fly on the wall in this forum for over a year now. I usually don’t have the time to discuss at length any issue broached in here, so I usually just listen to what you all have to say. My first impression when I began visiting here was that many of the people were very knowledgeable about Austrian economics. I’ve since changed my mind on that. That’s really beside the point; I’ve just wanted to get that off my chest for a long time.
Listen, agent-based modeling is very consistent with Austrian economics–and in my humble opinion, agent-based modeling from an Austrian perspective is a good way for Austrian economics to regain some respectability from the more technically-minded economic community.
“Personally I am a huge fan of ABM - but in my experience you won’t find any support for it on this forum.”
Agreed. But I think it could be positioned to get some support. For example if the purpose of modeling is not forecasting quantifiable moves in the economy but to set up a ‘model’ of a free market and see what various interventions by the state lead to, well that’s a model i think a lot of Austrians could get behind,. The purpose of the model can’t be to predict the future, but to study the affects of certain policies ‘in the lab’ so to speak.
Perhaps a better approach would be to use a MMORPG or some close variation so actual people were behind the agents and ‘ran’ the model so to speak. That would give you a continuous model with self motivated, unpredictable agents to study. Brief, very brief article on what this means: http://www.techdirt.com/articles/20060329/1846230.shtml
Or better, find a way to hotwire some animal brains into a huge network and see how they respond. A ‘Matrix’ of sorts for crows or mice, or some animal that has demonstrated some memory and problem solving ability. Thinking outside the box here, but it’s a possibility.
I meant to post this link in my above post but it takes so absurdly long for the “More” dropdown menu to load in Chrome that I decided screw it and made a new post.
You can add me to the pile of people supporting such a project. I’ve made a game where the user would interact with computer agents on screen. They knew how to join crowds and get angry, and they knew how to turn corners on streets and keep walking.
I really don’t care who is against such a project. I already see the merit in it. For one, it would help to shut up those mainstream economists who claim that the austrian school is “unscientific”
Personally I am a huge fan of ABM - but in my experience you won’t find any support for it on this forum.
Demonstrate that it is a) compatible with the Austrian method and economics (or else, demonstrate that this method/theory is flawed) and b) that it adds anything to our understanding of economics, and you might find support for it.
When I hear people saying they want to axiomatise AE or show that game theory is compatible with it, I say fine, go ahead. Until they demonstrate the utility of the endeavour, though, there’s no reason for me to be on board with it. A lot of things might be compatible with the Austrian framework, but strictly speaking add nothing to it. That’s why there is often a lack of enthusiasm.
BTW, I don’t think Austrians have any problem with modelling thought processes per se (whether they consider it part of economics is another story entirely.) They take issue with the idea that you can quantify/measure preferences.
For one, it would help to shut up those mainstream economists who claim that the austrian school is “unscientific”
Why feel such an obligation, though? Trying to make quantities out of the unquantifiable is itself unscientific, and so the onus is on neoclassical economists to show that their project is not ultimately a fatally flawed one.
I agree, though I have alluded earlier to perhaps some marginal gains from modelling to make certain demonstrations. There is no point trying to impress or compromise principles to court favor, to look scientific to a group of people who have patently misunderstood the scientific method.
A model is not a replica. A mannequin is a model of a human. It’s not exact but you can put some cloths on it and get a passing idea of how they might look on a human. As any clothing shopper knows, it is a bad idea to assume that the clothes will look as good on you as they do on the mannequin.
