COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	To:   Students in CS262, Winter 1985-86
C00017 ENDMK
C⊗;
To:   Students in CS262, Winter 1985-86
From: Robert W. Floyd
cc:   Nils Nilsson

Stanford's course evaluation system has benefits, but the anonymity of
evaluations is wrong. When I evaluate a student's work, I sign that
evaluation and am responsible for it.  If I make a mistake, it can
be corrected. If my evaluations reflect nonsensical beliefs (like
giving no A's because A means perfect and nobody's perfect), I can
be criticized for it.  If I evaluate maliciously, I assume there
could be legal redress. When students evaluate my work, I can't even
give a direct response to constructively intended criticism. I want
to respond to your evaluations of CS162.  I the only way I can do so
is broadcast.

I am aware that CS162 was not a pleasant experience for some of you.
Normally a course provides more of a support system through prompt,
frequent feedback, TA hours, review sessions, etc.  I had severe
bronchitis that was completely incapacitating for days, partially
incapacitating for a month, and still has debilitating effects after
more than three months. A course that should have had the services
of a TA for 10-20 hours per week had not even a grader.  The resources
were not available to do more.  I believed at each stage that I could
give a better course than could be given by a drafted substitute.
While I wish there had been a better support system, for me and for
you, I am not sure all of you took responsibility for yourselves as
graduate students can reasonably be expected to.

Replying to your specific comments directed to me:

	"Consider who IS taking your class 
	 not who you WISH was there."

The people who were taking my class are with few exceptions a select
(about one applicant in four) subset of an already impressive 
self-selected subset of college graduates (Our MS students have an
average undergraduate GPA of 3.72 and a 92nd percentile rating on the
quantitative GRE, compared presumably to a population mainly applying
to graduate school). They are voluntarily enrolled in the Software
Theory option, and intend to present themselves to the world as
having mastery of that subject - that is why it's called a Master's
degree.  They have taken a data structures course and a concrete math
course as preparation for mine.  My course is, in its area, the most
advanced they are expected to take.  I expect them to be able, with
guidance, to read a book generally regarded as a standard authority
on sorting, searching, and analysis of algorithms.  I expect them,
as graduate students, to take responsibility for themselves; to
allocate adequate time to study even if they are not forced to by
regular problem sets, to read and selectively to attempt solving the
distinguished set of exercises in the Knuth text, to respond 
imaginatively to the difficulties imposed by a diminished support system.
If that is not a description of the students I have, then I suppose it
describes the students I wish I had.  Even though I know of students
who, at the end of the course, can't read Knuth's assembly language
programs, I still expect students to do so on their own.  Even though
I know not everyone follows my recommendation of working on most of the
Knuth problems up to level 18 or so, and a few of the others to level
25, again I expect MS students to do this on their own, without compulsion.

	"Give SOME straightfoward problems 
	 that illustrate basic concepts."

If you mean problem sets, the Knuth exercises  are unsurpassed.
You just have to do them.  If you mean exam problems, each could be
solved by systematic application of methods covered in the course.
I suspect you said to yourself about information/entropy methods,
adversary arguments, enumeration methods, etc., "I don't need to
know this," but to be at the state of the art you do need to know them.

	"Too much work at the end of the quarter.
	 Not enough guidance at the beginning."

I agree about the latter.  The answers to the first homework probably
ran two hundred pages of mathematical scribbling.  For some weeks I
was not physically able to grade them properly.  The assignments
at the end were take-home problems by overwhelming preference of 
the class itself.  I think they were mostly routine one- or two-hour
problems.  To do five guaranteed an A.  If you hadn't been doing five
hours a week or so of problem solving all along, they might be a lot
harder, I suppose.

	"Be more enthusiastic about the material,
	 more clear in your explanations."

I am enthusiastic, I think the subject is beautiful, but it's hard
to seem enthusiastic when you can't breathe, when you have a coughing
fit if you more than whisper.  As for the explanations, oddly enough
I thought of them as clear.  If my thought is in fact muddled,
exhorting me to be clear will hardly reorganize my brain.  I'd say
that a course like mine gives you a chance to study as an apprentice,
to say "How does his mind work?  How can I incorporate his methods";
I am very good at solving problems of "software theory."  I attempt
to make my methods accessible by making them explicit, naming them,
drawing attention to which are being applied, etc.  If you are
accustomed to learning results rather than method, this requires a
shifting of gears. You checked off "Marginally contributed to my
skills in thinking."  Pity.  You missed a good chance.

	"Also DO NOT give exams that are much over the
	 ability of the average student in the class."

The exam questions, to repeat, were intended to be fairly routine.
A 50% performance is typically quite acceptable on my exams, and 
I made this clear beforehand.  If I give problems you are guaranteed
able to solve, the real world you go to work in won't be so kindhearted.
What will you do when an employer asks you to tune an algorithm and
you can't?  The average student got a B, and only three of twenty
got less than B-; is that so bad?  Or would you rather not know
how much you don't know?

	"The point is not to single out who is so good to
	 handle very hard problems; instead the learning of
	 the students should be your concern."

If some students learn to solve more problems than others do, isn't
it sensible for my evaluation to show that?

	"Try to learn how to generate interest."

Again you exhort me to change my sinful ways, without defining sin.
Now let's see how you like it.  Try to learn how to be interested.

	"Assign relatively small, but WEEKLY homework problems."

I normally do.  I would not have been physically able to grade them,
until the end when I was saturated with catching up on the exam
grading, the CS comprehensive exam, MS admissions, etc.  Do you
remember my early telling the class that it would be in your interest
to help me find a grader?  Did you try?  Did anyone?

	"(There is no reason to avoid giving homework that has
	  answers in the book, as long as you require the student
	  to show his work.)"

The university honor code requires professors to avoid creating
unusual temptations to cheat.  I would be very uncomfortable,
for my sake and that of honest students, giving homework with
available answers.

	"Grade and return homework promptly.  The homework gives	
	 students a chance to learn how to do the material."

See above.

	"The problems given for the midterm and final were good,
	 interesting problems.  It would have been better if they
	 had been distributed more evenly throughout the course."

Thanks, and agreed.  See above.

	"This has been a very useful course for me.  I think
	 your style is just great.  I am glad I took this course.
	 It has greatly enhanced my thinking.  Great Job!"

Gee, are we still talking about the same course?  Thank you; if one
person comes out with useful tools in his kit I am glad I taught
the course.

Some of you checked off the bottom (or middle, or top) answer to
every question.  I assume that expresses a feeling, or is intended
punitively, or something.  Did you really find that the course's
contribution to thinking skills "confused my thinking"?  Does your
confused thinking account for your giving such low evaluations?
Do you think you'll eventually get back to normal?  Are you going
to sue me if you don't? Seriously, this practice diminishes the
information content of evaluations and displays their popularity
contest role.

When the CS department first adopted mandatory course evaluations,
one visiting professor got by far the worst evaluations, across
the board, as I just did.  Next quarter he got no bad evaluations.
He was dead.  I feel lucky.

Let me know a few years from now whether you find useful what you
learned, or might have learned, from my course.  One trouble with
course evaluations is that there is no mechanism to measure subsequent
usefulness.  I see evidence that the effects of many courses sink
without survivors, possibly during the breaks between quarters.
I think the evaluations may be mostly popularity contests, that
unfortunately are often taken as the sole measure of the quality
of teaching.

I am sorry to be so longwinded; to answer each of you, I have to
answer all of you.