Information agents as guides in cyberspace
© Copyright 1994-2002, Rishab Aiyer Ghosh. All rights reserved.
Electric Dreams #10
2/May/1994

Suppose you want to arrange your appointments for the week. Normally, you might come up with suitable timings, and actually contact each person you want to meet, for confirmations. If you're busy, that could get hectic. Most offices have secretaries for precisely this purpose. Information technology tends to provide automated solutions for such tasks, which, apart from reducing the need for a secretary, bring you much closer to the workings of your office. A wired solution to your scheduling problem would be to send electronic mail, automatically, for every appointment you enter in your electronic diary. The only problem is that you'd have to process confirmations and alternative proposals on your own. A large number of your ordinary electronic information needs pose this problem, of personal verification and processing. What will save us from this flood of information is the technology behind some new innovations in cyberspace -- agents, also known as knowbots (for knowledge robots).

The problem with information on demand, instantly available on any subject you could possibly want to know about, is that you have to demand it. Every time. You have to specify, in a detailed way, what it is that you are looking for. After all, the advantage of digital data, its organization, remains one only if it is used to search for information. You can't just throw a question into cyberspace: "I want material on the uses of grass in tribal mathematics" and expect a reply (unless there are others who share your particular interests). You have to know how to break down your needs into morsels according to the structures prevalent in information organization. What to look for, where, and how to put it all together. For any requirement but the most specific, this is not an exciting task. It may even result in you ignoring the huge banks of data "out there" because they are not user-friendly.

Now suppose you could just pose the question like that. Presumably you have been doing some other research on the uses of grass in tribal mathematics. You create an agent, and give it the task of travelling around the world looking for information you may need. The agent, a piece of software, knows the major sources of data. It has a vague understanding of what you need, and proceeds, on its own, to break down your question into queries to different databases, or requests in different on-line forums. How it does this will differ from agent to agent. Essentially, an agent will use a combination of your description of the problem, with some level of 'intelligence'. After collecting large amounts of data, the agent will hopefully digest the whole thing into a coherent summary.

General Magic is a company that emerged out of Apple Computer, with the support of Sony, AT&T and others, to develop a standard interface for agents. The user interface with the cute name of Magic Cap, will be available on Sony devices later this year, along with the scripting language, Telescript, to aid in the description of problems to agents. Much of the complexity of the operations will remain transparent to users. To tackle our original scheduling problem, under this interface your diary will automatically create Telescript agents to go and confirm or reschedule appointments. These agents will of course negotiate with those of your clients, which raises the issues of security and verification of identities. These are solved through mathematically secure digital signatures and encryption techniques, perhaps making cyberspace not into a well-planned city, but at least a jungle with experienced guides.




  • Electric Dreams Index
  • dxm.org Homepage