You Say You Want An Education?

With the recent announcement of 17 new schools participating in the massive open online course (MOOC) site Coursera.org, I thought it would be an interesting exercise to see if it was possible to design a reasonable computer science curriculum using just Coursera courses, where “reasonable” is a curriculum that roughly mirrors the coursework required for a four-year university computer science degree.

I’ve looked over all of the available Coursera courses as of September 21st, 2012, and created a four-year curriculum. I’ve tried to follow the curricula suggested by real world colleges; in particular I’ve loosely based the approach on MIT’s course 6 curriculum (specifically, 6-3).

“Semester” is a loose term in this case, as the courses vary in length from 6 to 14 weeks. I’ve assumed it would be possible to take 4 core curriculum courses in the same semester, and that the student would take an additional course that is not computer science related.

My aim is a computer science education that gives the student a background in:

  • Algorithms
  • Control (robotics, other mechanical interactions)
  • Data structures
  • Debugging
  • Electrical engineering
  • Hardware
  • Programming language theory
  • Software architecture & design
  • Networking
  • Systems programming
  • Mathematics, including:
    • Statistics
    • Probability
    • Logic
    • Calculus

Little time is spent on specific programming languages; once the student is exposed to Python and some other languages in the first semester they are expected to be able to pick up other languages. That will take some work outside of the core course work but the Internet has so many good resources for learning the specifics of programming languages that I’ve chosen not to include classes on specific languages.

Below is the curriculum. I’ll discuss some of the reasons behind the choices later, but first take some time to click through to the course descriptions and see what you think.

Year 1, 1st Semester

Computer Science 101

Calculus: Single Variable

Statistics One

Learn To Program: The Fundamentals

Year 1, 2nd Semester

Introduction To Logic

Data Analysis

Learn To Program: Crafting Quality Code

Introduction to Systematic Program Design

Year 2, 1st Semester

Algorithms

Fundamentals of Electrical Engineering (first half)

Writing in the Sciences

Introduction to Computer Networks

Year 2, 2nd Semester

Networks, Friends, Money and Bytes

The Hardware/Software Interface

Probabilistic Graphical Models

Introduction to Mathematical Thinking

Year 3, 1st Semester

Fundamentals of Electrical Engineering (second half)

Algorithms Part 2

Control of Mobile Robots

Automata

Year 3, 2nd Semester

Programming Languages

Game Theory

Human Computer Interaction

Digital Signal Processing

Year 4, 1st Semester

Computer Architecture

Cryptography

Pattern-Oriented Software Arch. For Concurrent and Networked Software

Image and Video Processing

Year 4, 2nd Semester

Compilers

Machine Learning

Parallel Computing

Natural Language Processing

The curriculum above is purely focused on computer science and mathematics. I’ve left off the basic science and liberal arts classes, but Coursera offers many options in this regard and the student can pick or choose the classes that best fit the student’s interests.

It’d be awesome if Coursera had a course in discrete math. I think that would complement the other math courses in my curriculum nicely. Additionally, it’d be cool if Coursera had a course on operating systems development, or possibly another low-level programming topic like file systems or databases. Frankly, I’m amazed at the depths and breadth of the computer science courses available, and it’s a credit to Coursera that they’ve built such an impressive catalog. In 2012, a student can take the majority of a university computer science curriculum for free thanks to Coursera. That was not possible in 2011.

Learning Online

I’ve taken a few courses on Coursera and other MOOCs and for the classes I’ve taken the medium works well. Like in-person classes, the Coursera courses have lecture notes, slides, homework assignments, tests and quizzes, and online forums for the students to discuss the class. However, with a MOOC, the student can fast forward and rewind the lecturer. Online, quizzes and tests are primarily multiple choice, but Coursera randomizes the questions and the order of the answer (and often picks a set of multiple choices answers from a large pool of answers for a single question). Essay questions are peer graded, which is a process that I think is still evolving – in the class I took with peer graded papers, two of the papers I graded (out of 3) were verbatim copies of Wikipedia articles-which made grading those easy-but the 3rd paper was original content, and I’m sure I evaluated that content with less skill than a professor would.

In my Compilers class, the professor had a clever online tool that could be used to create proofs, and using this tool was required to complete the homework assignments. I think that more and more tools of this kind will need to be created to deal with assignments that aren’t multiple choice quizzes or peer graded essays. For the Compilers class our programming assignments were graded by a script that ran a series of unit tests against our compilers. Grading programming assignments this way felt natural to me – after all, at work my code has to pass a gauntlet of unit tests before the code can go into production.

On the other hand, MOOCs have a few challenges due to their scale (challenges which I’m sure people are working on solutions for). One of these is group projects. As a software engineer I work in groups all the time, and  as much as I hated group work in college it is an important skill for the professional programmer. Public speaking is another important skill that is easy to teach in a classroom setting but hard at scale.

There’s also a set of classes where students can be given the same assignment but produce correct but radically different work, or classes where the deliverables are too complex to be judged by software or other students. For example, MIT’s 6.033 class has several design projects whose deliverables are complex system designs. What is the best way to grade such assignments?

Coursera and its kin are evolving rapidly, and there’s been a lot of research into online education. I’m excited to see what learning online looks like next year.

Final Thoughts

I don’t think it will be long before Coursera, Udacity, EdX, or other MOOCs start grouping courses into curricula.  Khan Academy already has a knowledge map of exercises that build on another, only letting the student progress once they have mastered a required skill for the next level. I hope we’ll see curricula in MOOCs that is created and curated by professors or other education professionals.

What do you think of my computer science curriculum? What would you add or remove? Have you tried a class with Coursera or another MOOC?