Time: 4pm Tuesday, October 17
Place: 110 South Hall


IDENTIFYING INDIVIDUALS AND GROUPS IN COMPLEX ORGANIZATIONS

John B. Lowe
VP Language Engineering and Chief Linguist
Ask Jeeves, Inc.
jblowe@ask.com


Any large, distributed organization has several intersecting hierarchies and
groupings. For example, a company might have an engineering team, a testing
team, a marketing team, a management team, and so on. And these would be
distributed in offices around the country (or the world).  These groups (and
their subgroups) are formally identified and their membership is set (and
controlled) by the company.

Also, most companies need and therefore build cross-functional teams for
various purposes, and the tenure of these teams vary: some are permanent and
some last only for the duration of a project.

Membership in some groups is accidental (or perhaps incidental), though
these facts might be useful at some point.  Such groups include geographical
categories by office or home address (East Bay or City dwellers, Berkeley
Office members, etc.), age, gender, education level, etc.

In addition, there might be other "self-defined" groups such as Softball
Players, Bike Commuters, Java Experts, French Speakers, Bar Flies, and so on
(and perhaps sub-groups of these).  

At a minimum, the, each individual would have a set of fixed and variable
attributes defining their role(s), etc. within the organization.  Some of
these appellations would be immutable (either in fact or by fiat), while
some would be changeable by the individual,  and some quite transient (e.g.
"in the office today", or "in a meeting").

There are of course many instances of similar "communities": USENET, users
of a database system, and so on. I have chosen to focus on a particular one,
the "corporate model".

Of the many fascinating challenges managing such information presents, I'd
like to suggest we focus on just one or two:

1. This tangled hierarchy of communities could be viewed in a variety of
ways.

Suppose you wanted to build a retrieval system based on this information, so
that people could find out information like the following:

- Are any of the Ultimate Frisbee players in the Berkeley office are driving
to the league game tonight (I need a ride)?
- Is there anyone available TODAY who speaks Spanish and knows HTML and has
a few spare hours (we have a crash project and we're desperate)?
- Who is so-and-so's boss (I want to send a complaint/compliment)?

Assuming a browser-based user interface, how might this information be best
presented and queried?

Some specific subquestions:

- What does the "display" look like for a specific individual and specific
groups?
- What is an appropriate vocabulary for describing and cross-referencing
groups?
- How would you arrange to determine whether particular individuals or
groups meet certain criteria (e.g. would there be a text based UI which
permitted you to compute set intersections, or would something more
graphical be better).

2. What would be the best way to manage the composition of groups?

Of course some groups would be closed, with membership authorized by an
individual or another group.  But other groups might be created by
individuals, and the criteria for membership set on the basis of membership
in other groups. If the membership privileges are accorded by group members
how are their "votes" to be counted?  What needs to be decided here is if
there is a set of "group-making" procedures that can be applied generally.