 
|
|||||||||
              
|
Shop till your bot drops Chaos, not order, may be the hallmark of a computer-driven consumer economy Sunday, July 26, 1998 By Michael Newman, Post-Gazette Staff Writer
Ever since the first caveman compared dinosaur cuts at the corner supermarket, humans have proved to be natural shoppers. Now we are asking computers, which have done so well with such chores as spell-checking and arming nuclear warheads, to do to our shopping for us.
But the gazillion-dollar question isn't whether computers will be good shoppers (most seem to think they'll do just fine), or even whether people want computers to do their shopping (they won't have much choice in the matter). For deep thinkers, at least, the question is this: How will all these shop-till-they-crash computers affect the economy?
One of the more hopeful scenarios is this: Instead of paging through endless catalogues or dealing with surly clerks, we humans will simply sit at home. Our computers will do our shopping for us over the Internet, comparing products and sifting through prices. Let them fight the crowds at the holidays.
"It makes a lot of sense," says Ken Cassar, an analyst at Jupiter Communications in New York. There are already several shopping programs, called "shopbots," on the Internet. Tell them what you want, and at what price, and they will venture into the wilds of the World Wide Web to find it for you.
It all sounds too good to be true - which means it probably is. The immediate problem is that shopbots simply do not work very well, at least not yet. The more long-term problem is that while shopbots make the market more efficient, they also make it more volatile, and possibly less friendly.
Jeffrey Kephart is an IBM researcher who has created a model of a "software-agent economy." He basically put some "consumer agents" (think of them as shopbots) in a market with a "source agent" (a wholesalebot), some "broker agents" (retailbots) and stuff to sell. Then he stood back and watched.
The market was soon reduced to bedlam, marked by vicious price wars and poor selection that would have driven human retailers bankrupt and frustrated human consumers. "You just saw these agents chasing one another's tails in a horrific fashion," he says. "It was very upsetting to watch."
Shopping for shopbots
If shopping is the great 20th-century American pastime, and the World Wide Web is the great late-20th-century American invention, then it was only a matter of time before they found each other.
"We're fairly bullish on the potential of the Internet as a shopping medium," Cassar says. Jupiter expects $6 billion in online shopping revenues this year, ballooning to $41 billion in 2002. And much of that $41 billion, according to Jupiter, will be "spent" by computers.
Right now, there are relatively few shopbots on the Web, and those that do exist are not exactly perfect.
To begin with, most shopbots ask for a wide range of information, from your Zip Code to your bellybutton type. Often they ask for too much at once, Cassar says, forcing users to fill out a long form before they head to the cybermall. "The interfaces I've seen haven't really blown me away," he says.
Then, once they start working, the shopbots aren't very discerning. At www.shopfind.com, for instance, a search for "underwear" returns an unwieldy list, including 581 different kinds at an online store called www.theoutdoor.com. The prices are expensive - except for a $5.95 pair of "edible underwear" offered by Condomania Online, they're all at least $11 - and some of the selection is, to say the least, bizarre.
At junglee.hotbot.com, another shopbot site, 215 pairs of underwear are available. Along with prices, the site lists availability and shipping costs.
More unusual items are perhaps easier to find on the Internet. A search for the "Empire State Building," for example, discovers a veritable treasure trove, from an Empire State Building 3-D puzzle to Empire State Building salt and pepper shakers. These items aren't readily available at the local Wal-Mart.
Most of these sites do best with computer software, which is sold all over the Web. Excite's Product Finder gives a price range of $10 among dozens of retailers for a popular word-processing program. It's also quite good at finding rare Beanie Babies, including Trap the Mouse for $2,500.
Another feature about these programs that may not be readily apparent: Some merchants pay to have their site included in a search. And some sites are not included in searches because their operators won't let these shopbots in. It's the Web equivalent of keeping teen-agers out of the convenience store.
"Merchants love these agents and they hate them," Cassar says. They love them because they increase traffic to their site, generating advertising revenue. They hate them because they encourage users to shop based on price alone, ignoring such crucial factors as service, ease of use and cool graphics.
Shopbots "generally make merchants' sites a lot less important than they'd like to be," Cassar says.
Agent (mis)behavior
But then, shopbots aren't really designed for merchants. One of the great promises of the Internet Economy, so its proponents say, is that it will empower consumers, giving them access to information never before available. Armed with this data, they will be able to shop more efficiently.
"Any time information improves in a marketplace, buyers gain advantage," says Sam Kinney, co-founder and executive vice president of FreeMarkets Online, a Downtown company that holds industrial auctions over the Internet. It's certainly been the case for FreeMarket's buyers, who pay an average of 17 percent less for products purchased online.
But wait: It gets even better. In another of those too-good-to-be-true scenarios, not only do you save your money, but you're also doing your part to improve the economy.
By enabling more accurate and frequent comparisons, the Internet forces companies to be more efficient and responsive. And since shopbots are more ruthlessly logical than humans - since they will not be swayed by a salesman's pitch or a product's pretty color - they will force merchants to pay attention only to things that matter, such as price and quality.
Of course, that's the way it's supposed to work. In reality, no one really knows what an "agent economy" will look like.
But research at the Institute for Advanced Commerce at IBM's Thomas J. Watson Research Center in suburban New York have raised some intriguing questions. Jeffrey Kephart - whose official title is manager of the Agents and Emergent Phenomena Group - designed an experiment that looked like this:
He created a model with three types of agents. One published information, one was an information broker and one was an information consumer. He then gave each type of agent some rules to follow: The publisher was to earn money by selling information to brokers. The brokers, who resold the information to consumers, were programmed to maximize their profits. The consumers were given preferences for certain kinds of information and designed to purchase it as cheaply as possible.
Think of the publisher as Time-Warner, and the brokers as your local magazine store. The store owners know that, say, 51 percent of all consumers prefer Time, while only 30 percent want Fortune. People magazine is listed as a preference by only 3 percent.
Given this scenario, and the fact that they can buy only one title at a time, every store owner makes the same decision: Buy only Time, since it has the biggest potential market. This results in a vicious price war, as everybody sells Time at increasingly lower prices. Many stores go bankrupt.
But others try selling Fortune - though the market's smaller, the margins are higher. Then, as soon as all those other owners see how much money they can make by selling Fortune, there is a mass migration into the Fortune market. Another vicious price war ensues.
Meanwhile, all those People readers (there may not be many, but they need their weekly fix of Leo news), are left high and dry. And, for a while anyway, even the majority of Time readers are ignored while the Fortune melee continues.
Paradoxically, it is one of the very qualities that makes the Internet so efficient that also makes it so volatile, Kephart says: It's the Internet's "general lack of friction."
"There are a lot of things that tend to create friction in the real economy which are greatly reduced in the electronic world," he says, such as advertising and transportation costs. This is usually seen as part of the Internet's allure - "we tend to think of friction as a bad thing," Kephart says - but it is also problematic.
Moreover, he says, logical behavior on an individual level can seem "stupid from a societal point of view." It may be that, as the agent economy becomes more sophisticated, agents will learn to become more socially conscious, for lack of a better term.
It's easy "to look at these results and really get scared," he says. Yet it's only research. The agents "are doing really bad things inside our simulations ," he says. "I hope to confine that to our simulations."
Bots, agents, whatever
The question, then, is not merely how shopbots will affect prices, or how merchantbots deal with price wars, or whether little greeterbots will be positioned at the front of all the online stores. ("Hi there. Nice protocol!") The question is how all these various bots , or agents, will get along.
"What is the world going to look like when there are a lot of agents running around doing things for themselves?" Kephart says. "It's very exciting to think of analogies between the human free-market economy and what you might be able to mirror with agents. However, you can't assume it's going to work."
For example, he says, advertising currently supports almost all Web sites. "But agents aren't going to be watching ads," Kephart says. "Some of the incentives for using the Web are going to be changing."
As for the Big Picture - who knows? Ken Lang, who as the founder of Oakland's WiseWire - now part of Lycos - designed agents to help people find news and information on the Web, knows a thing or two about how they behave.
"I think the important thing is that there be coordination between agents," he says. Agents can be designed to supervise, delegate or cooperate, he says. There's no reason to doubt that they can also be designed to foster both economic efficiency and stability.
"There's always the problem of something going weird," Lang says. "That's usually a kind of bug and can be fixed." On the whole, he says, agents "will only screw up as much as a human would."
|
||||||||
