Learnings from Statistical Thermodynamics MOOC: Part 2: Forum Engagement

Venkat Viswanathan
3 min readApr 18, 2016


In my earlier post, we discussed about the time delay between homework attempts. We found that majority attempted the homework immediately instead of going back to the content and re-learning the material. However, we did observe a trend that as the homeworks got harder, the time delay between attempts indeed increased.

In this post, we will discuss how engagement in the forums correlated with the final score. In this analysis, we will only track engagement through actual forum posts, forum upvotes or downvotes. It is a bit harder to track the amount of time spent in the forum.

Before we go into the quantitative analysis, I would have to admit the engagement in the forums was truly amazing. In fact, so much so that I decided to create a new office hours video just to answer some of the foundational questions raised in the forums. I list some of them below — some of them are truly foundational in nature:

  • The partition function approach is derived based on maximizing entropy under certain constraints. Does this require certain properties of the entropy?
  • Is entropy maximization a local concept or a global concept?
  • What are the limits of the equilibrium statistical thermodynamics that we discussed?
  • How should I think of the chemical potential? Is it a quantity like momentum, a conserved quantity?

Engaging in the forums was among the most enjoyable parts of my day for the 6–7 weeks when the course ran. I was quite intrigued to see how learners who engaged in the forums scored in the course.

We analyzed the distribution of the number of posts for score range 70–80, 80–90 and 90–100 and these are shown below. A large majority of the learners (~80%) that scored between 70–100 in fact never posted or engaged in the forums. Initially, this might seem a bit surprising but of course, this is typically how any class is. There are many that might never engage in a classroom discussion but do very well in the course.

Distribution of forum posts for learners that scored (left) 70–80, (middle) 80–90 and (right) 90–100.

This issue is pointed out by Chinmay Kulkarni, Michael Bernstein and co in a recent study. Although enrollments in MOOCs are high, our MOOC had an enrollment of ~6800, we did find that students’ interactions with each other are indeed minimal. As they point out, this isolation is particularly disappointing given that a global community is a major draw of online classes. They have built a system called Talkabout, and they found that globally diverse discussions boosted student performance and engagement: the more geographically diverse the discussion group, the better the students performed on later quizzes. Something for me to think about as I run my course again — starting next week.



Venkat Viswanathan

Associate Professor @CarnegieMellon University, Advanced Batteries, Electrochemical Devices