[an error occurred while processing this directive]
"Smart agents are here to stay. Once unleashed, technologies do not disappear" (Norman, 1994, p. 71).
We agree fully with these sentiments which is why we believe agent technology is not a passing fad. We have now overviewed a broad range of work which goes under the banner of ëagentsí. We have outlined their various promises as well as their challenges. However, apart from such technical issues, there is also a range of societal (i.e. social) and ethical problems which are looming. Donald Norman writes:
"Probably all the major software manufacturers are exploring the use of intelligent agents. Myths, promises, and reality are all colliding. But the main difficulties I foresee are social, not technical. How will intelligent agents interact with people and perhaps more importantly, how might people think about agents?" (Norman, 1994, p. 68).
We disagree with Norman as regards the major technical hurdles ahead; as shown in the previous section, there are some extremely demanding technical issues to be resolved in most of the agent classes reviewed. However, we agree that he poses an extremely pertinent question. There are issues which society would have to grapple with through various legislations and they would be very thorny. They include the following:
However, such issues are not that critical immediately, but would be in the medium to long terms. In the short term, we expect some basic agent-based software to be rolled out e.g. some basic interface agents such as mail filtering or calendar scheduling agents. More basic mobile agent services would also be provided in the short term. We can also predict comfortably that many vendors would claim that their products are agent-based even though they most certainly are not. For example, we are already hearing of ëcompression agentsí and ësystem agentsí when ëdisk compressersí and ëoperating systemsí would do respectively, and have done in the past. As Guilfoyle (1995) warns:
"there is a danger, however, that customers may be disappointed by the gap between colourful press reports about smart agents handling half the work, and the reality of ëif-thení rules for message routing", p. 139.
In the medium term (3 to 5 years), some more decent agent applications would be rolled out for most of the classes of agents overviewed in this paper. Perhaps, there would be collaborative agents and/or integrated heterogeneous agent applications doing limited, but real workflow or air traffic management or controlling real telecommunication networks say, etc., rather than just simulations. Useful, but limited, interface agents should be available which perform roles including the following: eager assistants, WWW guides, memory aids, entertainment and WWW filters/critics. Indeed, the HOMR/Ringo system is already being marketed as Firefly by Agents Inc. - a Boston-based agents company. Open Sesame (Caglayan et al., 1996) - another interface agent which employs a neural network learning technique, is already on the market from Charles River Analytics (Cambridge, MA). More mobile and information agent applications and languages would soon be rolled out by vendors. We expect reactive and/or hybrid agent technology to start delivering some real everyday industrial applications during this period. Furthermore, during this medium term, we expect the WWW to be commercialised, to some degree, enabling agents of different classes to play a role in paying for services and performing some restricted buying and selling on our behalf, as Kasbah agents propose to do. We would also start seeing a proliferation of specialist agents conferences: agents in the aviation industry, agents in law, etc. We also envisage that the new domain of agent-based software engineering will grow from strength to strength. In the long term (> 7 years), we would expect to see agents which approximate true ësmartnessí in that they can collaborate and learn, in addition to being autonomous in their settings. They would possess rich negotiation skills and some may demonstrate what may be referred to, arguably, as ëemotionsí. We caution against the usage of such words as the latter, not least because the agent literature abounds with such vocabulary which have subtle and complex meanings in the human context (Smith, 1996a). However, it is also at this stage society would need to begin to confront some of the legal and ethical issues which are bound to follow the large scale fielding of agent technology. In the long term too, agents would also provide another design approach to constructing complex pieces of software.
Indermaur (1994) writes:
"they canít fly yet, but intelligent agents and smart software are beginning to walk" p. 97.
Our view is different: using Indermaurís metaphor, we argue that limited software agents are just to about to ëcrawlí out of research laboratories; apart from a few interface and information agents, there is still a very long way, yet, to them ëwalkingí, yet alone to their adolescence.
In summary, we believe, like Greif (1994), that agents can have an enormous effect, but that this will appear in everyday products as an evolutionary process, rather than a revolutionary one predicted by much press coverage. For example, the Ovum figures quoted in Guilfoyleís (1995) paper suggest a seven to eight fold increase in the agents market, in Europe and USA alone by the year 2000 (i.e. in four years time) - this suggests a revolution! Greif notes correctly that agents would initially leverage simpler technologies available in most applications (e.g. word processors, spreadsheets or knowledge-based systems). Then agents would gradually be evolved into more complicated applications. We hope the over optimistic claims about agent technology are moderated. It is a sad fact that the only jarring voices in this hymn of confident approbation about agent technology come from those doing the real research, and who know what some of the real technical, social and ethical challenges are. [an error occurred while processing this directive]