The obvious one: MOOCs usually involve small or no payment and are not typically part of a degree program. Everyone who has attended a traditional college has taken one or several classes where they disliked the material, the format, or the faculty but kept taking it anyway because it was a required course, they needed the credit hours to stay in good standing, or they had already paid nontrivial tuition and/or fees for it. For most MOOCs, if you are even mildly disappointed, you can just drop out with little remorse.
The second, less obvious one: on Coursera, I bookmark courses (including ones I have only tenuous interest in) months ahead by enrolling in them. Then when the course starts, I judge whether I still have the time or interest (I usually don't), and if not, drop out. I don't know if this is common, but if it is, then it would have an impact.
I'd also be curious to know how pacing impacts completion rates. Personally, while some platforms treat it as a selling point, I find it very difficult to complete self-paced courses. On the other hand, some people may miss one or two deadlines in a non-self-paced MOOC and simply give up.
I think the incentives you mentioned are very relevant. College students don't want to waste the high tuition they've paid, nor do they want to lose their academic status they've worked for. But you left out another huge factor. In the current hiring system, too many employers value where a skill was learned more than the level of proficiency.
I believe that, as more companies transition to audition-style interviews and make more accurate assessments of ability, the value placed on educational credentials will diminish. When this happens, there will be hope that educating oneself via moocs will provide a fair chance to be considered for skilled employment. When students believe that most companies genuinely value what you can do more than where you learned how to do it, I think there will be far more incentive to both enroll in and complete moocs.
IMO, there needs to be a better way to measuring value in an online course other than "did the user complete the entire course".
A MOOC is essentially an extended web interaction, not a college course. Very short web interactions which require minimal levels of work often have conversion rates in the 5~20% region. Think like "Fill out a survey" or "Make free bingo cards."
MOOCs require many hours of work, stretched over months.
Physical college courses carry a heavy precommitment to attend with gigantic costs associated with them. They also have a low completion rate in many circumstances, for example when the college hasn't already done an absolutely brutal filter based on a combination of socioeconomic status, IQ, and ability to take complex tasks to completion. MOOCs don't typically pre-filter based on those criteria, which means their completion numbers are affected in the predictable fashions.
1 - Sign ups alone require nearly no resources from the provider.
2 - Attendance and therefore popularity could be much lower. While that could be a good thing (less people and more money -> more support to each student), it would probably never be enough to cover a course's cost.
In the end, people are learning. I haven't completed any course yet (attended about 4, attending one now), but I'm thankful for them and I don't think the resources I used harmed the MOOCs.
I have signed up for updated versions of courses I had finished earlier simply because I wanted to show three minutes of lesson 4.2 to someone, and I've told my students to "sign up" for course X because its materials at week 4 explain a particular concept very well, and I'd recommend them to watch that - but the materials aren't available if you don't sign up.
In real life, if someone takes a brochure of a course because they're interested but never show up afterwards, you don't count those people against your completion rate (you might count them when evaluating your marketing funnel, though). Similarly, for a MOOC it makes no sense to treat the number of "signups" as anything more than that.
If you want a reasonable completion rate, you divide the number of graduates against the number of participants that (a) showed up to the course - actually listened to more than 20 minutes of lectures and (b) actually wanted to complete the whole course, as opposed to wanting only a single topic/part of the course.
Yes, I really wish Coursera et al would have "lecture archives" which didn't require login or, if they did, didn't require joining a class. This would help boost their completion rates, so it would be mutually beneficial. I suppose a few of the partner universities might like having restrictions on their content, though.
I think it's all just human psychology. We overestimate the amount of effort we'll give to a achieve a certain goal. And sometimes just having the gym membership/MOOC purchase is enough to satisfy our feelings of progress.
I struggle with side-projects as much as I do MOOC's, for the same reasons I think.
IMO (speaking personally here) - I find this approach to be Much more beneficial than the traditional college system. I learn the material better and have a deeper understanding of the entire subject. I've officially "failed" every MOOC course I've taken, because of missed due-dates. But when I look through the actual submissions, I usually receive near 100% on every assignment. This probably wouldn't apply to every topic, but the MOOC courses give me a structured way to learn that I wouldn't be able to come up with on my own, but also provide enough flexibility for family life, work, etc..
In the end the completion rate of a MOOC is a fairly meaningless statistic. It's completely reasonable to buy a technical book and only read a few chapters that cover what you want to know. If someone gains more from participating in 3 MOOCS vs finishing 1 then a low completion rate might be a sign that MOOCS are more useful than traditional classes not less.
The reward level is also low. You might get a certificate but nothing like credits that can transfer. You mainly just get knowledge.
I think it is a lot like going to the library. I've got three books checked out right now. I've browsed one of them and decided it is pretty useless and not quite what I expected. One looks really good but I haven't started it. It is by one of my favorite authors but it wasn't on my radar, I just picked it up because it was on the shelf. One I am halfway through and I intend on finishing it but it is taking about three times longer to finish because Hofstadter's Law applies to my reading list, too.
But my completion rate is the wrong metric. Over the past year I've probably read a dozen or more books that I wouldn't have otherwise if it weren't for my local library. And one lead me to buying the other six in the series, and a couple I've bought as reference.
I feel like MOOCs should be judged in the same way. There are a lot of people gaining education through MOOCs that they would not have otherwise.
The punchlines are that completion rates correlate with course duration and that the absolute numbers of people who complete a course are a quantum change.
The point of comparison in Grossman's view is that MOOC's are like textbooks more than traditional courses.
(1) Content. From what I've seen of MOOCs
in some areas where I know some things,
usually the content looks poor just on
the subject itself and/or its presentation,
i.e., too many of the lectures are just poorly
done in any common sense of good lectures.
E.g., start in the upper, left corner of
the board and write clearly, clearly enough
so that a student can take essentially perfect
(2) Purpose. Commonly the courses do not have
a sufficiently clear and meaningful purpose
for enough candidate students.
E.g., why study calculus? What MOOCs provide
a really clear answer meaningful even to
most of the people who consider a course in
E.g., why'd I study calculus?
In., in grades 1-8 I got dumped on by the
teachers; those teachers wanted to
teach to the girls with their
better handwriting, better verbal
talent and skills,
better clerical talent,
into fictional literature, and much
better classroom decorum. I
wanted, strongly beyond belief, to
know how things worked, and the
girls were little masters at knowing
how the teachers worked!
In grade 9 I discovered I could
do math, i.e., algebra and fairly easily
could lead the class. Essentially all the
girls in the class, all of whom had
effortlessly blown me away in grades 1-8
now were struggling while I was having
a great time, easily.
I got sent to a
math tournament. That success continued
in grades 10-12. In the 11th grade, I
saw the same for physics -- could usually
have my head down in class, resting, and
still be one of the best students in the
class -- of mostly 12th graders,
three of whom went to Princeton and
ran against each other and some fourth
sucker for President of the freshman
class. I got sent to an NSF math and
physics summer program.
So, in college, I was hot to go in
math and quickly went through the
course catalog and planned all the
math courses I would have for all four
years. I really liked math!
But to save money, I did my freshman college
year at a state school, inexpensive, I
could walk to. They forced me into a
course in college algebra beneath what
I'd already done in high school
(by far the best in the city, e.g.,
the three guys who went to Princeton;
MIT came recruiting; a guy I beat in
a shootout at the board in trig class
went to MIT). A girl told me when
the tests were, and I showed up only for
But, I knew that as a freshman
I was supposed to be learning
calculus and was torqued that I
was not. So I got the recommended
calculus book and dug in. In grades
9-12 I'd learned mostly just from the
book so also was able to do well working
through the calculus book, totally
I did well: E.g.,
I had quite nice college board scores,
especially in math and physics,
so, the next year (we had a little more
money) I got into
a good four year college.
So, where to start there in math?
Okay, I asked to start with just their
sophomore calculus. They said that
they couldn't give me credit for
first year calculus, but that was totally
fine with me. The text they used was the
same as Harvard used; it was good.
I did fine -- made As, loved the material,
had fun, didn't work at it very hard.
Right: I was a math major but never
really took a course in freshman calculus!
Also blew away everyone in their freshman
physics course -- first test had four
questions; I got all four; no other
student got more than two; tree counted
as 100; so I got 133; and I never missed
anything for the rest of the semester
so ended up with 33 points over a perfect
score. The second semester, missed only
one test question.
Went on and got honors
in math, got a Ph.D. in applied
math (stochastic optimal control),
have had math much of my career, and now
have some original math I derived as the
crucial core technology for my search engine
start up. Lesson: I've had a
purpose in mind learning math.
So, I did learn calculus, just on my own
for freshman calculus. So, why? Sure:
(a) Motivation. I really, really, really
wanted to know the material, not just a little
bit but a lot. (b) Resources. I had plenty
of time to do the work. (c) Preparation.
The four years of math I took in high
school gave me about everything one could
want in prerequisites for calculus.
(d) While I didn't know much about just
what I would do with calculus or math
after college, I did believe that they
should help my career, and I knew that,
for math, calculus was one of the biggies.
Okay, for the MOOCs, if
have (a)-(d) and some good materials, then
lots of students should do well.
If are not having many students are doing well,
then look first at (a)-(d) and the
materials. From what I've seen,
too commonly the course materials
are not very good.
E.g., for calculus, from what I've seen,
it appears that the courses want to
imply the a person can learn calculus
easily as a spectator (sport) watching videos
instead of doing the work studying good
materials, say, a good text. I doubt
that many students could learn calculus
as a spectator sport.
There really can be opportunities for
people to learn outside a classroom,
and online materials can help.
But need good materials, likely including
a good, traditional text book, and
also need, say, (a)-(d) above.
By the way, as a math grad student, I
taught calculus successfully. As an
MBA prof, I taught more in applied math
successfully. In my career, I continued
to learn on my own: E.g., for some weeks
I carried Blackman and Tukey, The Measurement
of Power Spectra to dinner at a seafood
bar in Silver Spring, MD, and then one week,
mostly on my own without my company
knowing about it,
wrote some corresponding software which
was a big help in our company winning
a competitive software development
contract -- my work on power spectra
estimation had in effect given our company
Lesson 1: Math can
Lesson 2: It is possible
to do the work of math making money
mostly via independent study.
I doubt that any of the MOOCs will
cover how to measure power spectra, say,
with the fast Fourier transform (FFT)!
How about a continuous
time, discrete state space Markov process
subordinated to a Poisson process?
Make money with that? Did that once.
Covered well in a MOOC? I doubt it!
I used to look at MOOCs, wanted to learn
say, stochastic processes in continuous
time, say, like in Gihman and Skorohod or
Lipster and Shiryayev
Karatzas and Shreve.
Gee, I didn't find much!
There is high irony here:
This is Hacker News, and my view is that
nearly all the learning crucial for the
current US information technology
industry, especially the software part of it,
has been from essentially just independent
study. So, the irony is that the
readers of Hacker News should be
about the best audience for MOOCs.
E.g., for a while in my career,
when I was fairly deep into computing but
before I'd ever had a course in
computing or computer science,
I taught several sections
of computer science at Georgetown
University. Later in grad school,
I was pushed to take a course in
computer science -- not much past
what I'd taught at Georgetown! At
one point, sure, the course tried
to cover quicksort. Okay, worthwhile
for such a course!
But I'd long since learned quicksort,
heap sort, Shell sort, bubble sort,
merge sort, etc. from some original papers and
Knuth's The Art of Computer Programming,
programmed all those that were in-place sort
algorithms, compared their performance,
programmed some special purpose versions,
etc. So, during the course, I just
watched and smiled. Gee, they neglected
to mention that in the bit reversed
permutation from some of the
versions of the fast Fourier transform,
Shell sort does nothing at all until
the last pass at which time it runs
in O(N^2)! Gee, why didn't they
mention that one!
Lesson: It's possible to
do well in computing and computer science
Surprise: One of the profs in that
course didn't like me and gave me a grade of
C. I appealed to the department,
pointed to some code I'd submitted
in my application to the department,
the code I'd written to schedule the
fleet at FedEx, the code I'd written
for power spectral estimation, etc.,
and the next year that prof was gone!
Lesson: If the Hacker News audience is
not doing well with MOOCs, then blame
the MOOCs, not the students!