IT-workgroup of Finnish Society
for Information Services arranged a
discussion meeting on agent programs. Here are some of the findings:
The term “agent program” has been used somewhat inaccurately in literature and in different marketing contexts. The accurate definition could be, for example: “An agent program is a computer application that executes tasks autonomously and thus supports and automates some steps of information acquisition and further processing”. Traditionally agents have been considered to be programs that automatically visit named www-sites and download pages that have been modified or added since the last visit, but also more traditional database alert-services and web-search engines (crawlers) can be considered as agent programs.
Agents have been discussed for
years, but they
are still not in a very broad use. The reason may be that in order to
take them
in efficient use, one must allocate resources to installation,
“fine-tuning”
etc. To deploy agents, a clear need must emerge. At least this far,
these needs
have been satisfied with more traditional and perhaps - in the short
term - as
efficient solutions.
One obstacle of broad use of
agents may lay in data
security issues. Some www-servers have been set to block out agent
programs
and, on the other hand, some agent arrangements (co-operation between
agent
programs) may in theory compromise the data security of the user.
Perhaps agent programs will not
be
successful as single, separate products but rather as additions to
search
engines or as parts of a KM software packages. It should also be
possible to
build a cost efficient substitute to ready-made, integrated KM
solutions by using
single agent products as
components.