On the Use of Agents Considered Harmful

Like many traveling business people, I often find myself on the Hertz bus returning to an airport on Friday afternoons. During the short trip, with my eyelids beginning to droop, I love to eavesdrop on week-ending philosophical insights from one business person to another. The following event took place on the way to O’Hare.

Two thirty-ish guys were returning to the airport together; one worked at the home office of a high tech company, call him Homes, the other in the field, call him Fields. Fields says to Homes, “Does your staff get such-and-such magazine anymore,”? Homes says, “Nope. We don’t get any magazines or journals for the staff. We found that they spend too much time reading.”

“In fact,” Homes goes on, “we don’t have subscriptions to any magazines or journals. We subscribe to a service. This service asks us what types of articles are of interest to us, which we specify as rules, like a boolean or keyword query; the service then creates “agents” which apply these rules to magazines and journals, electronically ‘clips’ articles, and delivers them to us on our network as files! That way people don’t waste time,”. And they don’t learn anything either, I thought.

In a way, this is another version of “garbage in, garbage out”. If the only things worth knowing are the broad categories you can specify logically, then the largest chunk of knowledge, the things you don’t yet know, will be hidden from you forever: “stuff you know in, stuff you know out”.

In a deeper sense, however, this vignette points to the weakest and most dangerous part of computers: they do what you say, not what you mean. Computers can only carry out a sequence of unambiguously specified logical instructions. We cannot even guarantee, in general, that this sequence will not go into an infinite loop (A. Turing, the Halting Problem).

It is impossible for me to specify logically all of the things of interest to me in a magazine. No computer can scan a magazine and free-associate. In fact, the act of browsing is it’s own reward: a new product announcement here, an ad with great layout helps to design an input screen there, a subject which never before generated interest catches my imagination.

Imagination! I guess that’s the real problem. Logic never made innovation. Insight, serendipity, and fortuitousness are the paths of innovation. To determine, via logical “agents” working in cyberspace, what will be of interest to you, will cut you off from the future. The richness of information on a page of WIRED, or even COMPUTERWORLD, cannot be logically specified for retrieval.

Cyberspace is a black box which can only be penetrated by logical query. The field of display, a 14″ to 20″ screen, is just not big enough to present things easily. Too many thoughts and physical manipulations are necessary to find information, let alone compare several pieces of information at the same time. Those who don’t learn the limitations of cyberspace are doomed to live in it.

Soon, Fields’ and Homes’ conversation drifted to the financial woes of their company. It seems that their company’s stock price has dropped below 16, a mythical lower barrier, and that their founder’s stock is next to worthless. Maybe they missed what is coming next.