Stanford AI class, some off-the cuff reactions, envisioning a future of technical learning online

January 25th, 2012
by jodi

I took the Stanford AI class. Overall I’m very impressed with the scalability and quality of the experience. I’m tempted to check Thrum’s TA roles for Udacity! While I don’t think that *every* kind of class should be taught in this way, for technical material with clear “right answers”, this is the (a?) right approach.

The class worked incredibly well overall despite a number of flaws. The best part was immediate feedback, on quiz materials and once the homework deadline had passed. When something is fresh in your mind, this is the real learning moment — so that’s invaluable for keeping student engagement. Even though so much of the class was rough at the edges, that, and a trust in the knowledge/expertise of the instructors, is the key aspect that would drive me to take this sort of class again. I also went in expecting to learn more about online learning (having both taken and TA’d other online classes); I did not expect to have such a powerful experience of the potential of this (relatively impersonal) approach.

“Education” means many things to many people. To the extent that education is about filling heads with core technical material, this is the future of education. In my mind, this pushes educators in several directions:

  1. pedagogical development / teacher training There are definite skills to be learned both in presenting course material and structuring courses for optimal learning.
  2. curriculum development and curriculum systems development
    There is incredible potential for increasing the personalization — for instance, there’s a kind of error that allows insight for the teacher, into how the student is thinking. For these kinds of errors, when a particular wrong answer characterizes a certain wrong way of thinking, you can provide specific feedback about the error, and even appropriate follow-up questions
  3. the tutorial model
    Formerly common in the UK — perhaps in Ireland, too?
    My impression is that this supplemented reading with personal interaction with a knowledgable person. Here the videos/class assignments provide structure, which could be supplemented as needed.
  4. exploring and justifying the need for liberal arts education, not as an alternative to technical training, but ideally in conjunction with and countering it.
  5. exploring and justifying what is the role of education in subsidizing research and stimulating researchers.

Here are some pros and cons, off the top of my head:
+
Material was chunked into short segments.
There were clear, regular assignments.
There was an active community of students, discussing in many places.
Discussion approaches improved over the course of the class (e.g. tagging questions to particular homework problems).
There were attempts at engagement (e.g. “office hours” where questions were submitted via
The students built tools for their own and others’ use; I relied heavily on the subtitling (much easier for skimming through the parts that I already understood).
The material was well-chosen.
Eminent instructors who know the field and are passionate about it.
The feeling of being part of a game-changing educational endeavor.


The classes were very video-focused:
–Other learning modalities were not well accommodated.
–Watching required a lot of time and for AI, no inherent speed-up capabilities were built in.
Feedback was not personal.
The schedule varied a bit more than necessary (DOS attacks, scalability issues, changes of plans)
Repetitive conversations in the online discussions made things hard to follow.
Lack of engagement in some ways.
Reliance on a number of external tools (e.g. google hangout for “office hours”, aiqus, reddit, …) made it difficult to keep up even with core discussion.
Difficulty inherent in the size of the class/scalability (e.g. google docs couldn’t handle the number of editors for course notes)
Assignments were not proofread in advance
Assignments could have made better use of the particular kinds of typical mistakes (“insightful errors”)
Class communication could have used improvement — for instance, announcements didn’t use RSS and important corrections didn’t always get shared.
There were two instructors with very different styles and pedagogical skills.
Some people complained about the “low tech” approach to videos — but I found it helped avoid sterility.
Class materials were sometimes not available at the moments when I had time.
Some material seemed simplistic.
Many students dropped out or lost interest (I haven’t seen statistics which were promised).

+/-
Assessment could be seen as varied or insufficient: several German testing centres were opened to allow students to prove their mettle in a timed environment.


The Stanford online classes (AI etc.) and Udacity came up in a DERI listserv discussion about the future of education. This is an answer to: “Care to share experience of the classes? How did it compare to a conventional lecture series? What were the pros and cons?”

Posted in higher education | Comments (2)

  • […] teaching is one of the ideas that could help sustain and justify small-group teaching as highly scaleable online learning becomes feasible and productive; MSNBC notes that, “Apparently even the Stanford students preferred watching the classroom […]

  • […] 10 invites if you want one). —- I’m taking a listserv post as the source of a blog post again; channeling jrochkind I […]