It's Not That Clever

Article excerpt

Artificial intelligence has not been a panacea for marketing, but expert systems herald a brighter technological future

Artificial intelligence. The name alone summons visions of dangerously out-of-control machines that are smart enough to generate lives of their own. The kind that futuristic cartoonists portrayed with vivid evilness in the 50s and 60s. Are market researchers ready for this technology?

Fast forward to the 90s. Market researchers are moving towards the use of AI in a lot of ways and the pace of new developments is picking up. But the first wave of "intelligent" products in market research are unlikely to spin out of control and they are more likely to be expert systems than artificial intelligence.

When talking to many market researchers, they often interchange AI and expert systems. "The confusion is understandable because they are so similar," says Edwina Dunn, director of marketing systems consultants Dunn Humby Associates.

"AI goes one step further than expert systems. The latter hold knowledge and decision patterns of one or more experts. These decision patterns or rules, are followed in the same order each time. In contrast AI allows the problem to be tackled using different sets of rules and the pattern the rules follow is changeable. The computer can run through a particular sequence of rules to make a decision, but the sequence can be changed if the computer finds a more effective route is found," she says.

In this sense, "computers can be programmed to 'learn' from the environment and then to use this learning to make 'intelligent' decisions," says Rory Morgan, managing director of Research International Technical Systems.

"It's a common fallacy that a computer can only do what you tell it to do. Clearly it's true if it means that the computer will faithfully obey the instructions given it; but it's incorrect if the meaning is that the programmer can foresee everything that the computer will do or that the computer will do exactly what the programmer intended," he continues.

If expert systems are more likely to proliferate before true AI systems, what are their benefits?

* They can communicate with users in English, or any other language (reports are actually written by the program).

* They have powerful reasoning facilities built in to solve problems.

* They can explain their reasoning to users.

One warning note. As expert systems can theoretically acquire the skills of an expert analyst and make decisions similar to those of the analyst, the decisions are only as good as those of the providers of the data. Bad analyst, bad expert system.

Another caution was issued by Louise Southcott, executive director of NOP.

"Market researchers mustn't get too clever about technology or else we risk confusing our clients," she says. …