28 October 2009

Dave Pritchard on physics education

I've always found the topic of teaching and learning fascinating. Yesterday, I went to a talk by Professor David Pritchard of MIT. He helped develop the Mastering Physics software that is used in many American universities today. Back in the early 2000s, it wasn't called Mastering Physics, rather it was Cyber Tutor and piloted in the MIT freshman physics classes. I took notes during Prof. Pritchard's talk.
  • Students spend the most time and learn most from homework (education experts and parents agree on this point!). However, homework is the bottom priority for most professors.
  • Cyber Tutor acts like an expert physics tutor. Prof. Pritchard showed statistical data that proves the software is as effective as a real-person tutor. If a student is completely lost on a problem, he or she can ask for a hint. If a student answers incorrectly, the software can provide feedback such as "check your units."
  • A goal of software like Cyber Tutor is to teach students multiple representations of information and multiple approaches to solving problems. Experts know all of this, but they usually don't communicate this knowledge. Rather, they focus on the one fastest way to the answer.
  • Cyber Tutor can track learning trajectories. Each action can be logged: FA = first attempt, SA = second attempt, NF = no feedback to wrong answer, F = feedback to specific error, H = hint, S = subtasks, FS = failed subtasks. An example trajectory would be H → FA → F → SA.
  • The software is a treasure trove for data mining. In addition to assessing a student's skill, data can also be used to fix badly written problems.
  • Prof. Pritchard posted his "cheaters never prosper" plot. I didn't understand any of the statistics terminology, but the graph proved that students who copy the most do the worst on the exams. The cheaters are detected with the following criteria -- 1) Response is under one minute, 2) Response is correct.
  • Interestingly, copying had very little effect on conceptual learning. This is probably due to the fact that class attendance was required and the majority of class time was spent on conceptual learning. Of course, copying had a huge negative effect on analytical learning.
  • In another example of how detailed data mining can get, Prof. Pritchard showed a plot of percentage of homework completed vs time before homework's due date. As one would expect, the copiers did very little work until close to the deadline.
  • Men cheated more than women, and business majors cheated more than other majors. I didn't get the exact statistics unfortunately.
  • Conclusion: It is clear that copying has a large negative effect on learning outcomes. Therefore, professors should discourage copying. This has in fact happened at MIT, partly through eliminating second semester pass/no record grading.
  • The last part of the talk was called "what should we teach?" Unfortunately, I had to leave and go to another talk, so I didn't find out the answer.

No comments:

Post a Comment