Devil's Advocate: Why machines can't bargain like humans

Though technology is advancing, people are still better at handling complex interactions than any machine. This is why, says Martin Brampton, it is taking so long to develop trading sites and other online marketplaces.

Internet-based trading sites were supposed to revolutionise purchasing. It turned out more difficult than the optimists of the dot-com era suggested. But now the Chartered Institute of Purchasing and Supply thinks a corner has been turned. Why was the process so hard?

Computer technology is complex, but interactions between people are a great deal more complex than any computer system. Business-to-business transactions are technically between corporate bodies that have legal existence apart from people. They only come to life, though, through the actions of people.

We've been finding ways to communicate complex interactions for a long time. When the Roman Empire spanned the then-known world, it built a web of military, commercial and legal links, even though it had only primitive means of communication such as messengers. It is now centuries since the introduction of the printed word multiplied the possibilities for the exchange of information.

The twentieth century saw a multiplication in communications channels. Radio, telephone and telegraph were established. Film and television became widespread. And by the end of the century the internet had become commonplace, with global access to information and almost cost-free communication through email.

Yet even before the Romans, trade took place across vast areas. It relied on a mix of capabilities such as established law, trading relationships and military might. Different factors predominated in different situations. Always prominent was the skill of the trader in understanding the situation, with failure leading to financial loss or worse.

Despite the growing bureaucracy of modern companies, buying goods and services has continued to depend heavily on individual knowledge and personal relationships. Naturally, these involve all the drawbacks and inefficiencies that are cited by the advocates of automation. But they are also extremely flexible.

And that is the root of a good many IT problems. Traditional databases contained masses of data, held in carefully designed structures. But typically the data on its own meant nothing because it lacked context. Only by reference to design documents could sense be made of the data and even then it was often exceedingly difficult to establish exactly what the data meant.

Slowly we are nibbling away at the problem. In the case of online marketplaces, a central issue is how to find a standard way to describe the goods and services to be traded. Only with carefully negotiated agreements about the precise meaning of data elements is it possible to achieve any uniformity in transactions. And even when the item to be traded has been fully defined, there is still the question of the terms and conditions of a sale.

Perhaps the most general question pertaining to this issue is how anything at all comes to have meaning. Long ago, when I first wrote computer programs, I was intrigued to find that computers could store whole English sentences. But it quickly dawned on me that storing the characters was not the same as storing the meaning. And that led to the question: What was meaning anyway?

Now that is a difficult and controversial question and without an acceptable answer there will always be a gulf between computer data and human communications. There is a good deal of scepticism about whether there actually is something distinct that is a meaning. Words are thought to carry meaning only because of their contrast with other words and the circumstances in which they are used.

All this implies that meaning is related to the experience of a human life, and not something that could be acquired by a typical computer. It seems that our optimism about computer solutions needs to be tempered by an awareness of the gulf between person and machine. We will continue to make progress with complex schemes such as marketplaces, but only slowly.