Artificially Intelligent Bargain Hunters

This provides an interesting counterpoint to one of the classic anti-free-market arguments, which is that the free market economics is based on the approximation that customers have a lot of information about a product and can successfully use that information to negotiate price. The counterargument is that we do not have such perfect agents creating our system. The counter-counter argument is to observe that it often works even so.

(As you can probably guess from the shortness of that paragraph, I'm not actually making those arguments right now, just setting up some context.)

Having truly pervasive electronic agents in the market working on people's behalf changes that argument chain. Maybe I can't study everything about, say, a given brand of hard drives, but my agent can learn a lot more on my behalf. How does this change the way the market works?

Honestly, I have no idea, I'd need way more time to think about this than a single HN post can be given. I just thought I'd toss this out there, because I think it's a very thought-provoking question, to the point that I'd say that if you think you can figure it out in 5 or 10 seconds I feel pretty confident you're probably off base. (For one thing, there are clearly second-order effects that will emerge if this ever becomes pervasive.)

You can look to the stock market as an exemplar, but there are a lot of potentially-relevant differences between the stock market and an economy full of pervasive use of agents. Price aggregators are another exemplar, but there are also potentially-relevant differences there. I think a large population of people using these agents for a lot of consumer goods would have a new effect of some sort.

The article describes these agents as "negotiation ninjas" based on "a series of simple rules - known as heuristics." That's far from artificial intelligence. Nothing in the article suggests that the agents do any kind of learning.

The problem with heuristics is that the best ones make very specific assumptions, so they usually don't scale to new problem domains, and don't adapt to changing market conditions.

Automated method of finding the correct price of something? It's called the stock market.

But I do look forward to having an agent go out and search not just on lowest price but some fitness function (% of being ripped off, chance of insane s/h fee's, etc) besides absolute price.

I've heard a lot about software agents over the last decade or so (they were hot when I was studying AI in grad school), but I have never seen a particularly compelling application.

I'm curious to see how well these shopping agents (probably the preeminent practical example of the technology) actually function in the wild, and how they're adopted.

If the use of shopping agents actually became a market force, I suspect many online retailers would need to change various business practices to adapt, much like websites have adapted by optimizing for search engines.