Lee Campbell
                                                       MAS837 hw1
                                                       2/26/97


I had the role of observer.  I am answering these questions for the case
of a human Assistant as in the class experiment.  My pertners were Mark
Foltz and Andy Dahley.


Q1: Did the Assistant provide effective assistance for the Searcher?

Yes.  Many times, the searcher would read a question aloud and then ask:
"where do you suggest I start looking?"  In such cases, the searcher is
deferring to the judgement of the assistant.  If the assistant had not
been effective, the search would not have gotten off the ground.

The assistant had alot of specialized knowledge about various sites on
the net.  For example, on the paris weather question, the assistant knew
the URL of the weather channel home page.  For the caffeine question,
the assistant suggested the Britannica, which produced an answer
quickly.  For the Alzheimer's question, the assitant suggested Yahoo for
sites devoted to Alzheimer's, which again produced a quick answer.



Q2: Were their interactions smooth, and communications between them
good?

Yes, although I believe they were strangers when we began, they
communicated clearly.  They jumped between two levels, sometimes
discussing the content of a question or a page, and sometimes discussing
the mechanics of searching or navigating.  Occasionally, they would jump
to a meta-level about whether to continue a particular search.



Q3: Were there times when one of them misunderstood the other?

No, I noticed no misunderstandings and not many requests for
clarification.



Q4: What could either the Searcher or the Assistant have done to
facilitate working together?

If they had more time, they could have developed a stronger sense of
what each other's strengths and weaknesses were.  Although the Searcher
was not a novice, he tended to rely on and defer to the assistant.  With
time, he may have found that on some topics he knew more than the
assistant.  Similarly, with time, the assistant may have learned to say,
e.g. "my first guess would be X, but you know this area better than me."

Ideally, they would have worked in parallel, in two browser windows,
talking to make sure they weren't duplicating each other's work, and
calling each other's attention to good leads and partial answers.

This approach is ideal because the bottleneck in these searches is
evaluating a page to see if it's really helpful or not.  This is best
done by a full human intelligence.  Since we had two of them, it would
have been better if they could have worked more in parallel.

-----------