Friday, June 16, 2017

Teach yourself to code using online tutorials…but which ones?

There’s an almost embarrassing plethora of “teach yourself to code” tutorials online—including reference sites like W3schools, interactive “try it in a browser” sites like CodeCademy and CodeHunt, beginner-friendly step-by-step activities such as those featured on Code.org, and many more. And that’s not even counting the dozens of online courses, both free (edX, Coursera) and paid (Udemy, Lynda)—devoted to some aspect or other of learning to code.

But which of these are actually likely to be effective in producing learning? Researchers who study how students learn have developed a rich body of well-tested ideas and recommendations around evidence-based teaching and learning—that is, the kind where you can unambiguously measure whether learning is taking place—but which if any of the online teach-yourself-to-code resources actually follow these principles?

My colleague Prof. Andrew Ko at the University of Washington and his student Ada Kim have produced a very nice paper addressing this question, which they presented at SIGCSE 2017, the largest conference focused on computer science education. Kim and Ko studied 30 free and paid highly popular online resources for teaching yourself beginning coding skills.

They have numerous findings, but the one that struck me most was that while many of these sites cover most of the same material (with varying degrees of beginner-friendly organization and quality of presentation), most focus on how to understand and practice a particular concept in coding, but provide little practice on identifying when and why to apply that concept in real programming situations. While Kim and Ko stop short of making “hard” recommendations, they find that the tutorials from Code.org, and online games such as Gidget, Lightbot, and CodeHunt, do the best job at incorporating elements of evidence-based teaching that have been shown to promote learning:

  • Provide immediate and personalized feedback to the learner
  • Organize concepts into a hierarchy with clear goal-directed for learning each concept
  • Opportunity for mastery learning, i.e. to practice a concept repeatedly until understanding is complete
  • Promote meta-cognition—knowing how and when to use a concept, not just how to use it
With everyone trying to make a buck (or just get visibility/sell ad space) claiming to teach beginners to code, a critical look at these resources through the lens of Computer Science Education Research is a welcome breath of fresh air. Read their 6-page paper here.

No comments:

Post a Comment