I had gone through this before when I was an Instructor.
Back in the late 60s I sent a class home; it was Random Process Theory, 6.573,
at Christmas with a Take Home Quiz. But I made a mistake, or it may have been
our secretary but I take the blame. I asked the class to prove an inequality
and I put the inequality backwards. Thus it was impossible to prove it. Now all,
but one student, proved the wrong inequality. One guy, I forget who, caught my
mistake and proved it the right way! I often wonder where he went. His approach
was that there was a right answer and he found it and the other class members
were just silly for not getting the right answer.
But the lesson was plain. Never send a class out with not
lifeline and an exam with possible errors. I apologized but I found out how
many hours people spent trying to solve the unsolvable. No one ever questioned
that I may have been wrong.
The second lesson is to try to understand how others will
read your question. Wording counts. In some of the 7.00X questions one wondered
just what they were asking. In addition it often took longer to read the
question then to answer it. Why? I suspect it was the attempt to set the
question up to be answered using the computer interface. The technique has yet
to be mastered. I can write questions that require a written answer. I would
find it more challenging to get the student to answer a computer solved
question. Ambiguity reigns. In addition by trying to be a bit tricky in answers
to challenge the student one further increases the ambiguity level.
The third observation is that one should not try to answer
the questions the way I may have done so say 60 years ago. Namely do not try to
be fast and try to get close; there is no close in computer graded answers. The
grader has no sense that you got it but just left out a hyphen. After all why
put those hyphens in anyhow. But alas, read the directions. Last time they did
not want hyphens and this time they do. They are not testing course materials
but reading skills. That to me is one of the problems with the MOOCs and
computer grading.
The other observation I made was that in the Discussion
sections the students seem to have that computer age chatter urge. Nothing is
thought out just typed in half-baked and expecting responses. They are never disappointed;
there are always others who will respond. The net result, however, is that the
Discussions are useless.
Finally, Lander was truly professional for where the state
of the art is today, but in another Harvard Med course they had the wrong
answers and then tried to patch it up. Prior planning prevents poor
performance. This is the risk of MOOCs; it is like my old MIT days, but multiplies
it by thousands. You have to be not only clear but correct. There is little
room for corrections.
Thus after almost two years looking at these are they
getting better? Yes, some more than others. But I still believe that have a way
to go.
In a recent NY Times piece the write points out[1]:
And perhaps the most publicized MOOC experiment, at San
Jose State University, has turned into a flop. It was a partnership announced
with great fanfare at a January news conference featuring Gov. Jerry Brown of
California, a strong backer of online education. San Jose State and Udacity, a
Silicon Valley company co-founded by a Stanford artificial-intelligence
professor, ...., would work together to offer three low-cost online
introductory courses for college credit.
Mr. ..., who had been unhappy with the low completion
rates in free MOOCs, hoped to increase them by hiring online mentors to help
students stick with the classes. And the university, in the heart of Silicon
Valley, hoped to show its leadership in online learning, and to reach more
students.
But the pilot classes, of about 100 people each, failed.
Despite access to the Udacity mentors, the online students last spring —
including many from a charter high school in Oakland — did worse than those who
took the classes on campus. In the algebra class, fewer than a quarter of the
students — and only 12 percent of the high school students — earned a passing
grade.
One may ask why? I think that the explosion of courses has
been useful but the platforms need fine tuning, and in some case a total
overhaul. The Lander 7.00X course is a brilliant example of how to get there.
The lectures are classic MIT. The Problem Sets are also classic. You learn more
from the sets than from the lectures, but that in no way demeans the lectures. The
Exams are a bit different and the problem again was classic MIT. Problem sets
are NOT Exams, so be careful. But like the old days, there is always a Hidden
Curriculum in MIT classes and once you discover it is all downhill from there.
The key is finding the Hidden Curriculum, the way the Prof thinks and what he
wants.
I would not like to see this abandoned, just the
expectations balanced. There could be an asset here, just not a goldmine yet.
Posts
