What computer science can teach economics (2009) | Hacker News Reader

What computer science can teach economics (2009)

news.mit.edu

57 pointsbkudria22 comments

The article is mostly about game theory, which has been a staple of economics for quite a while.

That said, I would be very wary of jumping to such endeavors simply because a particular branch of mathematics appears to supply a useful vernacular for a particular economic problem. More often than not, it leads to a situation where instead of offering new perspectives on problems, you simply repeat the same Walrasian equilibrium analysis or whatever orthodoxy, but in a more pompous language.

In fact, by far the most mind-bending and mentally invigorating economists have been those who have almost no mathematical analysis whatsoever. These include, for instance, GLS Shackle and Ludwig M. Lachmann. Their entirely verbal analyses has expanded understanding of real-world economic phenomena far more than papers on game theory, I'd argue.

Mathematical economics like the Solow-Swan growth model, Cobb-Douglas production functions, Walrasian auctioneer markets and Paretian static equilibrium models have sowed great confusion and unrealistic assumptions as much as they may have enlightened. A lot of misunderstandings on economic policy originate from taking comparative statics models too literally.

That said, there might be some use of theoretical CS. I've had the impression that process algebras like Hoare's CSP for modeling concurrency can be used for some microfoundations and thought experiments. I don't think any economist has done this yet.

   The article is mostly about game theory, 
   which has been a staple of economics for 
   quite a while.

I suggest to read the article more carefully. It's about applying the study of computational complexity to game theory. When you approach conventional game theory from this angle, things become interesting: e.g. where conventional game theory says a Nash equilibrium exists, the complexity angle suggests that the market won't find the equilibrium, because the laptop can't (for example the problem might be NP complete), and the laptop is much faster than the market.

H. Simon was among the founders of artificial intelligence and a very important microeconomicist, and I have often thought that these disciplines are not fundamentally different. What problem does backpropagation solve? Credit assignment. What problem does the market solve? Credit assignment. How do they do it? Localization of computation.

I think it's more general than that. Both backpropagation and the market introduce a feedback loop. That's the source of the behaviour. You can see similar phenomena in other closed-loop systems. That's why I think control theory is one of the best things humans invented ;).

Positive feedback loop, specifically. Tesauro noted power-law times in BP learning and you can go off and look at highly skew distributions in BP weight assignments on your own. Simon had a little program of research on skew distributions in economics (you cannot throw a stone but you hit one: money, firm size, investments, trades, income, GDP, yadda yadda), and how they are best explained by positive feedback loops.

>Finally, he says, it may be that where the Nash equilibrium is hard to calculate, some approximation of it — where the players’ strategies are almost the best responses to their opponents’ strategies — might not be. In those cases, the approximate equilibrium could turn out to describe the behavior of real-world systems.

I often think about how death and pain affect economic decisions. Someone who is poor and must be paid next week or start to go hungry, for example, makes much different economic decisions than someone who still needs to work to eat but has a cushion -- they can go several months without pay without risking starvation hunger. Even though they are in similar danger over time periods that are not too different, the second groups act as if it is no danger at all.

So there is a time horizon dependency at the moment of decision. I wonder how the complexity of the likelihood calculation affects the falloff rate of the death or pain variable? I wonder if we are constantly making these rough approximations and have some intuitive sense of when the numbers get too fuzzy to be meaningful and at that point get ignored?

Does somebody know of any resources that may have dealt with the economics of desperation?

You're right, there have been several economics studies showing that poverty causes stress which makes it more difficult to make decisions that are "rational" in a classical economics sense, such as sacrificing in the short-term for a long-term pay off. Searching "behavioural economics poverty" turns up a lot of good articles, here's an overview:

http://www.economist.com/news/finance-and-economics/21635477...

That is interesting but seems to put too much onus on the poor individual.

>poverty makes people feel powerless and blunts their aspirations

What if it is more than a feeling? What if they are not irrational but seeing quite rationally what transactions await them?

Let's assume that two sides of a potential transaction are rational, and further assume that each side will maximize the value they have after the transaction. The poor person comes to the table with a value of 100 units but that they cannot eat and cannot be transferred directly for edible widgets -- lets call them time units. They must trade for 25 edible widgets or starve. You can buy each edible widget for one liquid currency unit. They sit across from a person that can pay them in liquid currency units for their time units knowing they can make 120 liquid units using those time units. That person knows they're poor and at risk of starvation if they do not trade their time unit for liquid units quickly. Being cold, rational actors they start the negotiation low. 1 liquid unit for all 100 time units. The poor person will starve anyway and thus refuses. Negotiation continues until they hit the magic number of 25 liquid units and an agreement is struck. The poor trades 100 time units to the person that can make 120 liquid units with it for 25 liquid units...they lose, but don't lose it all through death. The employer makes a profit of 95 liquid units.

Let's run through the same scenario with someone who's not desperate. The employer needs that employee knowing they can make 120 units with those time units in their possession. With no other players in sight, they might negotiate all the way up to 119 liquid units for those time units knowing they will profit. The employer's profit is 1 liquid unit and the employee gets 94 more than the poor person who was facing a complete wipe out if they refused the offer.

Competition changes things obviously and this is a simplistic model but the fundamental point is that one is negotiating against loss and is willing to take any loss above total loss while the other is negotiating to increase value. The second can afford to have aspersions and has every reason to feel like they have a good measure of power -- while the other is truly powerless and they live at the whim of others and have no rational right to have aspirations because they have no way to improve their negotiating position nor any expectation for that to change since every transaction they enter into is to reduce loss rather than increase their possessed value.

Yes, a non-poor person has a better BATNA, therefore they can demand more from the other party. But I think the studies show that the poverty has an effect even beyond that.

I guess my problem with studies that show the other effects of poverty is that they seem to paint the poor as just hapless, irrational, creatures; subhuman and deserving of their plight since they can't reason their way out of it.

And, who knows, they may be, but I'm going to have to see a lot more proof than what I've seen so far which tend to be of the vein 'well, look at their circumstances. You wouldn't make that decision that led them there. I wouldn't make that decision that led them there. Obviously they are not rational like you or me'.

It is very strange that everyone in the economic model of the world is assumed to be rational but the poor.

While your concern is valid - patronizing and disrespecting the poor is rampant - I don't think that's a consequence of observing this effect; there's usually a prior belief. After all, the evidence is actually against the "subhuman" hypothesis - it shows that the reduced quality of the decisions is specifically a result of the environment, not an innate deficiency of some individuals.

Also, the assumption that everyone is rational is no longer unquestionable, ever since behavioral economics started gaining prominence.

Economics is a highly inbred field, which makes it very difficult to transmit cross-disciplinary information. The recent popularity of machine learning in the field might be a rare exception.

Economists only cite other economists. One of my colleagues put it this way: if you write a paper that is interesting to economists, and an economist finds it and does some follow-up work, that paper will get all the citations from economists.

> Economics is a highly inbred field, which makes it very difficult to transmit cross-disciplinary information.

I disagree. Economists pull in work from many other fields: statistics, math, computer science, political science, sociology. Kahneman and Granger are recent Nobel Prize winners that didn't have a PhD in economics.

> The recent popularity of machine learning in the field might be a rare exception.

Not really. Economists were applying neural networks in the 1990's, just as one example. The problem is that these ideas often fizzle out because they weren't designed for economic data and it turns out that they don't provide much value.

> Economists only cite other economists.

That's the real issue. The field of economics is very closed. It's even worse than you describe. Not only do you have to cite papers written by economists, you have to cite the right economists, or the referees will stop reading and write a negative report. Getting new approaches into economics is easy if you have connnections within the editorial process that rejects more than 90% of submissions.

> Kahneman and Granger are recent Nobel Prize winners that didn't have a PhD in economics.

I don't know about that--it was really a chore to christen the field of behavioral economics, previously considered to be "psychology".

My favorite example of this kind of thing (from medicine): someone re-discovers the trapezoidal rule for integration, names the technique after themselves, publishes in a Diabetes journal, and gets hundreds of citations [1].

[1] https://www.reddit.com/r/math/comments/1xfa8p/medical_paper_...

Economists have been investigating these concerns for decades. There is work dating from the late 1960s on the computability of general equilibria under uncertainty, and concerns about computability have played a role in the influential literature on bounded rationality. A small sub-field of the discipline, usually called "computable economics", began to appear in the mid-1990s, when K. Vela Velupillai [0] began to publish papers on the topic.

Velupillai is a central figure in computable economics, and you will find references to much of the work that's been done in the area if you search for his name or browse Google Scholar for his papers [1] (and the papers that cite his).

0. https://en.wikipedia.org/wiki/Vela_Velupillai

1. https://scholar.google.com/scholar?q=KV+Velupillai

i love how CS people always assume everyone in every other field is just slacking off waiting on the programmers to assist them with their elite knowledge about things no one else has considered

Actually economists are easily the worst offenders of any field - they even gave a name to it, "economic imperialism"

Computer Scientists in my experience are quite respectful of a number of other fields, including mathematics, most engineering disciplines, physics, even philosophy (logic) and psychology (big in stats), and most recently, biology. Computer Scientists have a "no-bull" mentality, because ultimately the output of their endeavours has to work on an unforgiving and brutally honest machine, unlike economics, where the output need only persuade others in the field, who have a vested interest in perpetuation of credibility of the discipline, leading to the suspicion, by others, of alchemy.

I think it has to do with how integral computers are to every other field. Like if everything deeply depended on woodworking to function, carpenters would be offering input on everything.

   assist them with their elite knowledge about 
   things no one else has considered

Did economists consider how computationally hard it was to find Nash equilibria before CS folk?

There's a third option when shooting a penalty kick: aim straight for the goalie. No matter which side the goalie goes for, the shooter always wins.

Reply on news.ycombinator.com