RE: Concept of "user"

From: Jim Pitkow (
Date: Thu, Jul 29 1999

Date: Thu, 29 Jul 1999 17:35:20 PDT
To: "Lavoie,Brian" <>, "'VJ Cothey'" <>
From: Jim Pitkow <>
Cc: "''" <>
Message-Id: <99Jul29.173808pdt."362221">
Subject: RE: Concept of "user"

I agree that while convenient to a non-expert, using "user" at the server
side is dangerous.  For the server side, "client", while better, still
leaves many holes, as one of our recent studies (WWW Journal) highlights
the problems of treating each IP as a separate user.  So user != cookie !=
client != IP. For the server side, basically what we try to get is a
"uniquely trackable entity (UTE)"  When possible, we use cookies.  If not
present, we use heuristics that use IP address, time, user agent, referrer
information, and the site's topology to determine a UTE and their
corresponding path.  IN certain circumstances, we can get persistent UTEs
(like persistent cookies), but the problem is detecting and handling
cookie-non-compliance in such studies.  Most of these are modified
definitions off the WWW6 "In search of reliable usage data" paper I wrote a
while back.

At 08:38 AM 7/29/99 , Lavoie,Brian wrote:
>Thank you for your comments. I see the point you're making, and the
>potential for ambiguity in the "user" definition. However, rather than
>adding more terminology, I think the remedy may be to rephrase the existing
>definition to make some of these points more explicit.
>The user definition was intended to represent "human" agents explicitly
>operating clients (browsers, etc.) in order to interact with servers. In
>this sense, the human agent can take multiple forms: a person, a household,
>etc. The definition needs to be re-phrased to make this more explicit. The
>bottom line, however, is that a user represents some form of direct human
>control of client-side requests.
>In my opinion, the user concept is not easily translated to the server-side
>perspective. The server interacts with the client (see definition), not the
>user, although obviously the client is an intermediary for the user. But the
>two are not synonomous: consider that servers cannot always determine if
>requests from a client originate from explicit human direction, or from the
>client itself (e.g., an automated agent like a spider). Because of this
>problem, I would suggest that the term "user" be reserved for analysis that
>addresses client-side activity. I.e., in order to preserve the human element
>of the user definition, client-side analysis should talk about users, while
>server-side analysis should talk about clients.
>These are just my opinions. Perhaps others in the group have an alternate
>-----Original Message-----
>From: VJ Cothey []
>Sent: Thursday, July 29, 1999 10:45 AM
>Subject: Concept of "user"
>I am engaged in a large ( > 1000 ) longitudinal ( two years )
>client side investigation of information seeking activity which is
>essentially web user characterization.  Local network configuration
>provides instrumentation to consistently track individual users over
>collections of daily client side traces derived from browser history
>files.  My goal is to discover how patterns of activity evolve as users
>become more experienced.  My personal background is library &
>information science so please excuse my ignorance of the finer points of
>CS :-)  
>So much for background.
>I applaud the mission and work to date.  In particular the draft
>terminology & definitions and metrics which is impressive.
>However in this context, my worry is that, in practice, the definition
>of "user" is likely to be corrupted and for example will refer to
>different entities depending on its client or server side usage.
>Whilst this may be overcome by in each instance clearly defining what
>surrogate entity is being used to represent the user, my suggestion is
>that additional terminology be employed especially in respect of server
>characterizations in order to refer to the server side "user" concept,
>eg an individually identifiable client.  I don't rule out, in a
>particular study, there being an equivalence between two diffent types
>of user entity, but it would be the responsibility of the researcher to
>establish that equivalence.
>As an example of my concern in practice, consider the Huberman et al.
>(1998) paper ( Science, 280:95-97).  AOL members are uniquely identified
>and referred to as "users", so possibly several individuals in a
>household would all appear as a single entity.  A few paragraphs later the
>term "user" is employed in relation to the Catledge & Pitkow client side
>study.  Here user definitely has the intuitive meaning of individual person
>( with attributes such as age or gender ).
>Do you see the problem?
>btw apologies if this is a mis-post.
>Viv Cothey, Graduate School of Education, University of Bristol