Does anyone know of the use of ABM’s in the area of Austrian economics? I am currently messing around in Netlogo to see if I can put together a model of a basic exchange economy, and was hoping someone may have already done womething like this.
From my experiances of AE, ABM would be considered garbage almost by definition. ABM’s attempt to model what is going on in the minds of the agents, but Austrians consider modelling people’s thought processes to be fundamentally impossible and so not even worth considering.
I do not agree with the Austrians on this point, especuially considering I am an artificial intelligence programmer!
In conclusion, I would suspect the answer to your question is “no”.
OK, cheers for the reply. It’s a shame if it’s true that Austrian econ has not looked into AB modelling. I am by no means a expert programmer, so would not be knowledgeable of how advances in artificial intelligence make modelling more possible, but it strikes me that the Praxeological approach is in essence about developing abstract models by assuming certain behaviours of individuals., and that this would not necessarily require a super intelligent agent, just basic suppositions about the rules people follow.
I mean, take the whole idea of the action axiom, and deducing from it the ordering of preferences. This indicates one avenue in building a model - that is to have agents that attempt to maximise the number of items toward the top of their preference list. Programming in those basic rules of behaviour should in theory produce results that are in accordance with what praxeology predicts. Maybe I misunderstand praxeology though.
BTW, artificial intelligence programmer sounds like an awesome job.
If you don’t agree on this point, then you probably do not understand it, because its not a matter of subjective opinion, but an epistemological fact.
“If you don’t agree on this point, then you probably do not understand it, because its not a matter of subjective opinion, but an epistemological fact.”
It may be true outside a deterministic universe that you can’t know what any one particular person is thinking at any given time, but being able to model it is still a possibility. Nature does it all the time; there’s more than one person in existence. While it may be impossible to figure out what everyone is thinking and plug that into a model, that doesn’t mean it’s impossible ot develop a model with agents that are reasonable to near perfect facsimilies of how people think and act. The problem is the model would have no relevance except to demonstrate of the effects of policy because once built it would go its own merry way and have nothing to do with the real economy we deal with every day. There’s no practical and likely no ultimate way to model our universe, that doesn’t mean there’s no way to build a model that parallels ours in governing laws and participating agents and see what happens.
ABM’s attempt to model what is going on in the minds of the agents, but Austrians consider modelling people’s thought processes to be fundamentally impossible and so not even worth considering.
I disagree. IMHO, 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).
I mean, take the whole idea of the action axiom, and deducing from it the ordering of preferences.
I tried to axiomatise AE mathematically. This is far from being obvious. E.g., preferences of a single actor probably form an order - but is it total or partial? What are the elements of this order? The full states of the whole world? Some observable subsets of these states? But do not forget the preferences change over time… And at the end, in the kosher AE preferences are not causing action, they can be deducted from it, so even methodologically this is suspicious.
BTW, an unrelated way I tried to model actors was by them holding beliefs (memes), and before trying to rationally deduce an action just consult the current set of beliefs (intuition/moral/social pressure). Of course, memes could be copied around, and even mutated - another idea was that actors may choose to teach/preach memes they do not believe themselves (like in real life). All in all, this quickly becomes complicated 
I am still hoping to make a sim economics game out of this.
I believe this approach may certainly be more fruitful in helping our understanding, than much of the “representative consumer” DSGE macro modelling currently favoured by the New Classical and Keynesian Schools. However, I would caution, that just as with modelling in general, it can only really help us in answering questions in a very specific context, and we must not overrate its utility in comparison with the development of “pure theory” vis a vis praxeology, which we could then use to feed in to help illustrate certain outcomes of situations with a computational model (and we must pay a lot of attention to the assumptions which we may “feed in” to our models).
One question which I have pondered pursuing, which ABM may be useful for would be to illustrate how Cantillon effects take place once we allow for asymmetric degrees of belief in a sample population concerning price movements in a situation whereby an economy is hit by localised monetary injections. Information theory could then be incorporated to help illustrate the spread of information triggered by the learning responses of individuals to this stimulus, hlping to illustrate both how prices change and the redistribution of wealth occurs across the board.
I also feel information theory may bear a very interesting relation to Hayek;s ideas about the use of knowledge. I’m not even sure the theory as it currently is would allow me to be able to answer the questions I’d like to, so at the moment this is probably one of the most speculative of my research ideas.
Upon what are you basing this insight?
Presumably, by your logic, you’d be happy to set fire to a passing stranger on the basis that you could not possibly know (in advance) whether they might quite enjoy it or not.
“I tried to axiomatise AE mathematically. This is far from being obvious. E.g., preferences of a single actor probably form an order - but is it total or partial? What are the elements of this order? The full states of the whole world? Some observable subsets of these states? But do not forget the preferences change over time… And at the end, in the kosher AE preferences are not causing action, they can be deducted from it, so even methodologically this is suspicious.”
That’s a kicker, the solution is you don’t need to know. The agent needs to know. It may be the study of the brains of some animals might inform the design of the agent. Crows for example show problem solving abilities that are surprising, using ‘resources’ to get at food, remembering and avoiding certain people, etc. I think a problem is Austrians assume the same types of models as other economists. Why do you need to know what your agents are thinking or prioritizing? You don’t. You need self motivated agents with the ability to choose. Which will give you what you need to some degree: unpredictability without randomness. I find it funny how often people think those things are the same.
Mick, I have two dreams for this world. One is world peace. The other is for you to understand the Austrian method.
Your attempt at a reductio is a non sequitur. How could we possibly know whether or not someone is seeking immolation? You can’t model it because it is impossible for you to model value preferences. You can model what would happen if someone did prefer it or if they did not, but you cannot predict individual preferences precisely.
Source?
are we all using the word ‘modelling’ in the same way? just askin…
Modeling, at least as I understand it, is essentially copying. So you would only ever be able to copy what humans have done in the past. And then only at random. It’s not any sort of intelligent being. You certainly enter into paradox when you( falsely) think your copy-catting human model is somehow going to become sentient.
Or what I fear most is people using these model’s(Random immitations of past human behavior) to explain future or present economic phenomena.
I have no particular source that I can remember - its just puttng together things that have been said on this forum in a variety of debates. Maybe I’m wrong. If you would like to tell me that Austrians are happy to model human thought processes then please do. I will stand corrected.
That’s interesting… why did you use the word “precisely”?.. can I surmise then, that you are suggesting that “you can predict individual preferences to some degree”?
If this is the case then we are in full agreement.
Correct, only that if you understand that your prediction is by accident and nothing more then coincidence. Like predicting that a coin will be heads or tails. To state otherwise implies your trying to see into the future or to put more bluntly, “Model the future” or “Model future events”.
I asked for the source more to understand the claim than to challenge you. As Nir alluded to, the word “model” in this situation doesn’t have an obvious meaning. I would need more words before I could come to a conclusion.
So it’s just an accident whenever I email one of my friends and expect them to respond at some point?
But don’t we do that all day every day?
Let me put it another way. If you modeled a man’s behavior for 30 years. And you had a 99% statistical possibility that if you emailed that man he would email you back. What guarantee beyond your historical experience do you have that the next time you email him he will reply?
More along those lines, what if you build a machine to copy that man’s behavior. The the machine is programed to respond 99% of the emails it recieves from you. Will this machine mimic the exact same behavior as the man? Will it respond to the exact same emails? Most importantly will this machine tell you which emails your friend will respond to in the future?
Ultimately this comes down to a useful application of empiricism or an abused application of empiricism. I don’t disagree with modeling, copying behavior, and developing expectations. I think it’s necessary to every day life. My point is its useful as long as the user understands it for what it is.
As I stated before my fear is:
Thats all really.