Information Overload: New Tools For Reading
Information overload is blighting many organisations. There is too much to read, too little time in which to read it, and no reliable way of working out what should take priority.
The blame must, in part, be pinned on the huge quantity of information that has become available electronically. But if technology is adding to the problem, it may also provide a solution. Organisations are experimenting with a number of tools, mostly based on artificial intelligence, in an attempt to lighten the load on their employees.
One such is Swiss Bank Corporation, which decided to address the problem after it monitored the readership of its own publications. Even though internal documents are a small fraction of the material demanding attention from its staff each day, it calculated that it would take several hours a day for its staff merely to keep up to date with its internal forecasts and reviews.
The response of its staff to being swamped with so much information was unsurprising: two-fifths of its publications were only read sporadically and a further two-fifths were hardly read at all.
These dismal findings encouraged SBC to explore a technical solution. It wanted to give staff a quick method of finding out what information is relevant to them; it wanted a way of delivering relevant information automatically to individuals; it also wanted to know which parts of which documents were best and least read. SBC asked its financial information engineering group, a team of researchers based in Basle, to find a solution for its private banking section.
An obvious part of the solution was ensuring that all its documents were available electronically. This did not present difficulties since SBC was already putting many of its documents on to its intranet. The more difficult challenge was finding a method of indexing these documents that would allow the investment bankers to
question and navigate the database quickly and easily.
The approach chosen for this project, which was known as the know-how pool project, was an artificial intelligence technique known as case-based reasoning. This technique makes use of an electronic database, which describes previous situations and the response to those situations. Presented with a new problem, the computer retrieves the most similar case
and, if necessary, adapts it to suit the circumstance.
The first step in applying case-based reasoning was creating an index of the documents. SBC achieved this using the CBR2 generator tool, which was designed by Inference Corporation of El Segundo, California. This summarises each document in the form of keywords; it also lists a number of keywords that are best suited for distinguishing one document from another. When a document is published or updated, the index file is automatically updated overnight.
One of the attractions of the system is its flexibility. CBR seems to be very easy to use, says Frank Block of SBCs financial information engineering group. For example, the user can type his or her request in ordinary language, such as I would like to have information about private banking.
Another attraction is that it can cope with typing errors and it can find dates close to those specified. The system also has a stemming algorithm, which allows it to match words that have the same stem.
The system, which has been developed both for a PC and for SBCs intranet, can help users define their interests. The user can select choices from seven boxes: region, currency, economy, branch, product, politics, strategy. The user can refine the search by answering a number of questions.
The know-how pool project also makes use of intelligent agents - software that can autonomously perform specific tasks. One of these is a news agent that informs readers about new information according to their specified user profile.
Another agent that SBC is working on is a document reading statistics agent, which counts how often and at what time documents are read. SBC expects it to yield quite detailed information, given that each document is split into chapters and sub-chapters. This would allow it to work out which sections are either read very frequently or not read at all. That could provide feedback to the authors about where they should provide more or less detail.
The pilot is being improved and extended, although Block is aware that it may be difficult to extend the operation throughout the organisation. The potential rewards - better investment decisions through faster and more accurate access to information - are high but hard to quantify. It is very difficult to give precise numbers on the return on investment, he says.
SBC is not alone in experimenting with CBR and agent technology for knowledge management. Halifax, the UK bank, used case-based reasoning to help its staff find the right information to deal with queries from members about products and shares in the run-up to its flotation. Boeing has co-developed the prototype of a knowledgeable agent- oriented system in a portable aid that provides training and support to customers on aircraft maintenance.
Organisations are using a range of approaches to help their employees find relevant information from an excess of data. Andersen Consulting, which tries to share knowledge between its 40,000 consultants across the world, has a repository of interfaces, or knowledge maps, which helps users find their way to the most relevant of its hundreds of Lotus Notes databases.
Automatic filtering tools are also proving valuable. The Price Waterhouse World Technology Centre in California, for example, is experimenting with an information extraction system called Odie - on-demand information extractor - to extract relevant information from news wires. As a result, users receive information on just a handful of relevant appointments every week.
The researchers working on applying artificial intelligence to knowledge management are aware that the subject is at an early stage. The American Association for Artificial Intelligence recently noted that artificial intelligence was relevant to knowledge management, but commented that most existing tools cannot be applied to the task in their present form.
Nonetheless, there is cautious optimism from researchers in the field. In our experience at Price Waterhouse, AI-based technology can play a key role in dealing with these difficulties in managing knowledge, says the PW team. SBCs Block is also upbeat about his companys pilot project. We have a good starting point, he says.