How to be a terrible thesis advisor  (mostly for graduate work in AI)
--- -- -- - -------- ------ -------

Assign students thesis topics based on the section headings in your
grant proposal, or on the boxes of the flowchart for your master plan.

When someone brings up a research paper, tell anecdotes about the
author, his advisor, and his colleagues.  This will impress students
that who you know is more important than what you do.

When laying out your laboratory, give first priority to minimizing the
cost of cable, last priority to good workplaces for students, and no
priority to fostering interaction among students.

Read your students' papers at most once.

When honest differences of opinion arise, paper them over with words.
For example, say "well, we could talk about this forever, but I think
we're all working towards the same basic idea, let's call it a
`neologistic/noetic knowledge representation'.  Now let's move on."

Regarding other schools of thought, make sure students know just
enough to be able to point out the "fatal flaws" in each, and so can
be good foot soldiers in the crusade for your own approach.  A useful
phrase is "why do you want to waste your time reading that?"

Never visit the laboratory; learn about students' work only from what
they tell you.

Define your research aims with catch phrases ("dynamic X", "emergent
Y", "the Z problem", etc.).

Have students handle computer system administration, and let them
think it counts as research.

Mumble.

Assign older students to guide the younger ones.

Involve students in decision-making for unimportant things.  For
example, you can easily while away an hour of seminar deciding who
should be discussion leader for what chapter of the reading.

Share your most trivial thoughts with your students.  Better yet,
bring them up as seminar discussion topics ("in the shower this
morning, it stuck me that whitespace is really important.  Let's think
about whitespace from an AI perspective").

Avoid conflicts with your students; in particular, don't be too
demanding.

If a student reveals that he is confused about what counts as
meaningful research, ridicule him.

Take no interest in what courses your students are taking.

Pick up ideas from going to conferences, then bring them up in seminar
without explaining from whom you got them or explaining the context
in which they arose.

Plan for research seminars to last at least two hours.

Avoid meeting with students individually, do all advising out in
public, at seminars.

Never go near the laboratory evenings or weekends.

Always come unprepared for seminars; you're smart enough to fake it.

Never program yourself.  After all, you went through that once, and
now you're an ideas man.

Let your students see you rushing to meet deadlines.

Avoid critical discussions of research strategy.  A useful phrase is
"We'll do it this way.  Why?  Because I'm the professor and you're a
student."

Expect nothing much from your students, and subtly let them know this.

Give all your students the same research topic, but with slightly
different names.  If this is the same topic as your own dissertation
topic, all the better.

Let your students see your grant proposals and learn the art of
doublethink.

Enforce disciplinary boundaries.  For example, say, "that sounds like
the sort of thing that people in software engineering would work on,
so let's leave that topic alone", or "why do you want to worry about
that? that's a software engineering issue".

Never suggest your students contact other professors or other
researchers.

Let your students submit articles to third-rate journals.

If a student's work is not giving the results expected, belittle him.

Encourage your students to work on fashionable problems.

State your opinions loudly and frequently, so your students know what
to write in their theses.
--
nigel Ward  nigel@sanpo.t.u-tokyo.ac.jp
	    Mechano-Informatics, Engineering, University of Tokyo
	    7-3-1 Hongo, Bunkyo-ku, Tokyo 113 Japan