Jeff Young has a great piece in the Chronicle of Higher Ed today about what he calls the “bandwidth divide” and how most MOOCs require learners to have persistent high-speed internet access. When we created the Mechanical MOOC course, we built it on existing open resources mostly because we though it was the most efficient and cost effective way to do it–by leveraging the investments already made in creating MIT OpenCourseWare, OpenStudy and Codecademy.
We realized very quickly that a lot of additional flexibility came with leveraging these resources. Because they were from mature projects focused on openly sharing their resources and functionality, they had developed alternate modes of delivery to address bandwidth issues:
- the 6.189 course used an open textbook that was downloadable
- the 6.189 course materials (assignments, notes) themselves could be downloaded in a single zip file
- the 6.00SC videos used were downloadable from iTunes U and the Internet Archive
- OpenStudy was launching a beta mobile interface just as the course kicked off
And our learners downloaded the materials is large numbers:
Beyond that, we were able to also leverage the deep investments made in translating these open resources. The text is available in a dozen languages, and the course materials have been translated into Chinese. By building our course on open resources, we saved money and leveraged the work that these projects have already put into reaching audiences working without persistent internet or in other languages. A win-win-win.
- How big is the typical MOOC? – while an enrollment of 180,000 is often cited as the largest MOOC so far, 50,000 students enrolled is a much more typical MOOC size.
- How many students complete courses? – completion rates can approach 20%, although most MOOCs have completion rates of less than 10%.
- What factors might affect completion rate? – the way that the course is assessed may affect completion rates; the completion rates of courses which use automatic grading range from 4.6% to 19.2%, while the rates for courses which use peer grading range form 0.7% to 10.7%. This may present a greater challenge for teaching MOOCs in certain subjects.
- Do more students drop out if courses are longer? – there does not appear to be a negative correlation between course length and completion rate, which is interesting as you might expect fewer students to ‘keep going’ and complete longer courses.
It’s great to see some data on completion rates, and this will certainly stir up more debate on the topic.
But one issue not addressed in the current discussion is who really cares about MOOC completion? Certainly the groups offering them do, and educational researchers do. A fair guess that many non-profit funders do as well. Interestingly, though, some of the data coming out of the Mechanical MOOC Python course suggest that in the absence of extrinsic carrots like credit or certificates, learners may not.
In the eighth and final week of the class, we asked the 5,775 learners who signed up for the first iteration of the Python course a series of end of course questions; we received 21 partial and 61 complete responses. Assuming a survey completion rate of 3% (typical of what we see for MIT OCW surveys) and 5% (really good for an OCW survey) that would suggest a rough engaged population of learners (that is, still reading the e-mails we were sending out to structure the course) of between 2,733 and 1,640 people during the last week of the course.*
One question asked which was the last week of the course out of the eight they had completed. Here’s the response:
At the point of the survey, midway through the eighth week, 12.1% indicated they had completed the course and 13.8% had completed week 7. If we assume 25% attrition from those that completed week 7, maybe 10.4% of the 13.8% would be expected to finish the course. So in very rough numbers, 20.5% of the survey respondents might be expected to finish.
Apply that number to the estimated engaged population of learners above, and we can get very rough numbers of estimated completers: 560 – 336, or 9.7% – 5.8%. or somewhere in the mid to low range of MOOC out there, which might be expected, since we weren’t offering a certificate or other incentive for finishing. Now there are plenty of places to take issue with the above numbers, and since our course set up doesn’t have a solid way of counting course completers, this really should be taken for the back-of-the-envelope analysis it is. But…
What is really interesting to me here is the distribution of learners across the weeks completed. There is a large cohort of students (68.9% of respondents) that reports most recently completing weeks 4-7, which is to say they progressed significantly through the course but most of them were not positioned to finish the course “on schedule.”
How do they feel about this? Apparently pretty good. Granted the n’s are painfully small here, but if you ask how successful they felt they were in the portions of the class they completed, most report being completely or mostly successful:
Further, if you ask whether they feel prepared for further study based on what they had learned so far in the class, they likewise responded largely that they were very or somewhat prepared:
The data’s a little thin, yes, but this would seem to at least suggest that while MOOC providers and higher education commentators wring their hands about the completion rates of MOOC, the learners may not really care that much. If they are learning for the sake of learning, they may be quite content to fit in what learning they can given the constraints of their lives and be happy with wherever they finish up.
There’s a great deal of excitement (and fear) over whether MOOCs will replace parts of the current higher education system, but right now I suspect most of the activity with MOOCs (as has been the case with OER more generally) is in extending educational opportunity beyond the current higher education system. If this is the case, we may need some better metric for understanding student success and satisfaction than completion rate.
* Correlating data point: The week 8 assignment e-mail recorded 1,929 opens through our e-mail system.
Innovative model combining open education resources drew 6,000 registrants for first offering
CAMBRIDGE, MA, November 8th, 2012 — In August of this year, a group of leading open education projects announced the launch of a massive open online class (MOOC) that had no instructor or central learning platform. This so-called “mechanical” MOOC combined the offerings of three leading open education projects—MIT OpenCourseWare, OpenStudy and Codecademy—loosely linked together by an e-mail list to create a free and open course on introductory Python programming. The initial offering of the course, managed by Peer 2 Peer University, attracted 6,000 learners from around the world. A second round of the course will begin in late November, even before the first has concluded.
Sign up for the next offering of A Gentle Introduction to Python at http://mechanicalmooc.org
The vision for the course is that the e-mail list helps participants to keep moving through the materials together, but also provides flexibility supporting the differing pace of individual learners. ”There’s no penalty for working faster or slower than the e-mail schedule,” notes P2PU Executive Director Philipp Schimdt. ”Since the resources are all openly available, our job is more about creating community than enforcing rigid schedules.” The e-mail list driven model also means that new rounds of the course can be started at any time, allowing multiple cohorts of students to move through the materials and even support one another.
“There are tremendous advantages to our approach,” comments OpenStudy’s co-founder and CEO Preetha Ram. ”The learners from the first round are now four weeks into the course. They’ll be a tremendous resource for the learners starting in the second round. And for students in the first round who are struggling, they have the option of dropping back to the second round.” P2PU’s Schmidt also notes that the Mechanical MOOC model supports very rapid iteration and improvement. ”We’re folding what we learn from the first round directly into the second round six weeks later. We don’t have to wait for a semester to end or course to conclude.”
The Gentle Introduction to Python course combines content from MIT OpenCourseWare’s 6.189 A Gentle Introduction to Python class, with a study group supported through OpenStudy and instant feedback and practice projects from Codecademy. Learners earn badges demonstrating mastery through Codecademy and earn recognition of collaborative skills through OpenStudy’s SmartScore.
The Peer 2 Peer University is a grassroots open education project that organizes learning outside of institutional walls and gives learners recognition for their achievements. P2PU creates a model for lifelong learning alongside traditional formal higher education. Leveraging the internet and educational materials openly available online, P2PU enables high-quality low-cost education opportunities. P2PU – learning for everyone, by everyone about almost anything.
OpenStudy is a social study network where students can ask questions, give help, collaborate and meet others. Founded by professors and students from Georgia Tech and Emory University, and funded by the National Science Foundation and Bill and Melinda Gates Foundation, OpenStudy believes that students can teach other students through collaborative learning. OpenStudy believes in making the world one large study group where students can work together in a single place regardless of their school, country or background.
Codecademy is the easiest way to learn to code. Since its launch in August of 2011, Codecademy has been used by millions of users in more than 100 countries. Users learn to build websites, create web applications, and to understand the fundamentals of computer science through an innovative, interactive interface. Codecademy is funded by top tier investors like Union Square Ventures and Kleiner Perkins.
About MIT OpenCourseWare
MIT OpenCourseWare makes the materials used in the teaching of substantially all of MIT’s undergraduate and graduate courses—more than 2,100 in all—available on the Web, free of charge, to any user in the world. OCW receives an average of 1.75 million web site visits per month from more than 215 countries and territories worldwide. To date, more than 125 million individuals have accessed OCW materials.
Peer 2 Peer University
My first job out of grad school was as the coordinator of Emerson College’s adult degree program. Without a doubt, the highlight of the job was the daily contact with the amazingly motivated and persistent adult students who were overcoming tremendous challenges to complete their degrees. Their enthusiasm for learning and pride of accomplishment was absolutely infectious.
There are lots of great things about my current gig for sure, but close contact with motivated learners has not been one of them–until now. We are a few days away from the official start of our Mechanical MOOC Python course, and already I am thoroughly enjoying the opportunity to interact with the learners participating in the course.
The enthusiasm they show, the extent to which they are already working ahead through the material and helping each other, the Twitter stream and blog posts about the course all bring back that feeling of really helping people to do something they care about.
No doubt that OCW does this too, but my daily connection to it is abstract, often discussed in numbers. Nice to reconnect more firmly with the people.
I’m working with a group of testers to run through the initial draft of the course sequence for the upcoming Mechanical MOOC Intro to Python course, and I have to say, I am really loving the unplatform aspects of it. I live in one of the more wired cities in the US, and I still spend a fair amount of my time outside of WiFi range. I tried to complete the Udacity Stats course this summer, but one of the challenges was that I always had to be connected. My biggest blocks of free time are during my train commute, when theoretically I have wireless service (from AT&T) but practically I have at best spotty cell coverage (from AT&T). This meant no working on the Stats course during the ride.
Because the Mechanical MOOC depends on existing open content outside of an enforced platform, I have other options. MIT OpenCourseWare helpfully provides a course download option, so I have the 6.189 course installed locally. The text for the course is an open resource downloadable as a PDF. The videos from 6.00 are available through iTunes U, so accessible offline on both my laptop and phone. As an added bonus, OpenStudy just released a mobile interface, so I can even ask and answer questions without a WiFi connection. Codecademy even seems to be functional on my iPhone at some level, though I doubt I’ll try to complete those lessons on that platform.
By not creating and enforcing a single platform, the Mechanical MOOC gives up the opportunity to harvest lots of tightly integrated data about the learners, but it allows us to take advantage of all the hard work that the content and community providers have put into making their environment accessible and inviting. Hopefully this model is going to allow us to meet the learners where they live.