BLOG@CACM
Education

Computing Education Must Go Beyond Intuition: The Need For Evidence-Based Practice

Posted
Mark Guzdial
Georgia Institute of Technology professor Mark Guzdial

The highest-quality education practice is evidence-based education defined by the US Department of Education as "the integration of professional wisdom with the best available empirical evidence in making decisions about how to deliver instruction" (see website). The movement towards evidence-based practice has already swept through medicine (see here), psychology (see here), and other fields.

Computing educators’ practice would dramatically improve if we drew on evidence, rather than intuition. Raymond Lister has been writing for years about the problem that computing education practitioners do not engage with evidence and the research on education (see his 2012 Inroads article). Raymond calls the knowledge that computing educators use in making teaching decisions "folk pedagogy" (see this article).

Folk pedagogy encourages what is believed to be best practice, but it cannot validate best practice.

We have evidence that computing teachers don’t use evidence. Davide Fossati and I studied 14 computer science teachers from three different institutions (see our SIGCSE 2011 paper). Davide asked them about times that they made a change in their teaching practice — why did they make the change and how did they know if it was successful or not. They used their intuition, informal discussion with students, and anecdotes. Not a single teacher used evidence such as class performance on a test problem or homework.

Without evidence, teachers rely on their intuition, informed by their experience. Sometimes that intuition may be informed by many years of experience. Sometimes that experience is not at all relevant.

The March 2015 issues of Inroads (linked here) has a special section on "The role of programming in a non-major, CS course." That’s a topic of great interest to me, since I’ve been developing and studying a programming-based approach to teaching introductory computing that we use with non-CS majors at Georgia Tech for over 10 years (see summary paper here). The first article (see here), by Richard Kick and Frances Trees, described the AP CS Principles course, its research-based development and experience at the pilot sites. The article by Steve Cooper and Wanda Dann (see here) presented findings from early research on teaching programming to children and evidence from courses at Stanford and CMU using Alice. The rest of the papers use no evidence.

Henry Walker’s first article (see article here) argues that programming should not be a priority for a non-CS majors course.

Development of student problem-solving skills (with programming) requires time, so the inclusion of programming within a non-CS majors CS course comes at a price: significant other topics must be dropped.

The claim that "significant other topics must be dropped" is an empirical claim. Since there are many non-CS majors CS courses with programming described in the ACM literature (like this one at CMU and this data-centric one at Boston University and this one at Kalamazoo College and the one I developed at Georgia Tech), one could clearly consider those courses and identify what significant topics were dropped in favor of programming.

Henry doesn’t consider these contrasting cases, but he does offer recommendations of what a non-major student ought to know about CS. Are CS faculty the right ones to define the learning goals for students who do not plan on a career in computer science? When we designed Media Computation, we formed an advisory board of non-CS faculty to tell us what their students needed to know about computing. For example, a professor in Architecture wanted his students to understand the difference between Photoshop and a CAD tool for manipulating a diagram. While both can "extend a line," Photoshop manipulates pixels while CAD tools manipulate a vector representation that can later be used to control automated devices like milling machines. I hadn’t realized that that was an important data representation learning objective for Architecture.

Michael Goldweber makes two arguments in his paper against including programming in a non-CS majors course (see paper here). His first argument is like Henry’s in that he offers what should be in a non-CS majors course. He suggests an "embarrassment model of course and curriculum development" which he describes as "one enumerates the topics for which, if their students did not know about/have experience with, one should feel embarrassment."

I assert that this topics list does not include programming, not because programming is without merit, but because the inclusion of programming requires a time commitment that then leaves an insufficient quantity of time for the key "embarrassment" topics. These include algorithmic problem solving, the use of abstraction to tame complexity, and the limits of computation…It is my position that it would be an embarrassment if a student coming out of a non-CS majors course could successfully program a solution to Selection Sort or write a program that draws a fractal tree, but have no idea of how to apply algorithmic problem solving to real-world problems.

Michael suggests an example real-world problem that he believes non-CS majors should be able to solve after their introductory course to CS.

Given a graph representing cities and connecting highways, some of the cities house a Red Cross warehouse while one other city experiences a disaster; describe an algorithm for locating the closest Red Cross warehouse.

My suspicion based on research evidence is that few CS majors or non-CS majors would be able to solve this problem, even after several classes (with or without programming). Transfer of knowledge is hard to achieve, and students are particularly challenged to recognize that a real-world problem is related to another knowledge that they have learned (see the chapter on transfer in How People Learn). In the decades of studies that have tried to find such transfer, the research evidence is that computing courses do not help students develop general problem-solving skills (see paper reference here). However, that’s just my suspicion. We could gather evidence to see if students get closer to achieve Michael’s goal with a programming course or a non-programming course. Michael doesn’t offer any evidence.

Michael’s second argument is about the cost of programming in time and tedium. He dismisses the use of languages such as "Pascal, Java, Python" in which students engage in "wrestling matches with compilers over a missing or misplaced semi-colon or squirrely bracket." He disparages graphical programming languages ("Scratch, Alice, Kodu") as a "children’s introduction to programming."

Should the learning outcomes for a child’s introduction to the field be what we want for one’s singular university-level non-major’s course?

My colleague Betsy DiSalvo recently published a paper about a study with African-American teens who learned both Python and Alice (see blog post about her study). There wasn’t a clear winner. Rather, Betsy found that the preference depended on the careers that these students desired. Those who were interested in computer science as a career preferred Python. Those who were interested in media and design preferred Alice because they liked what they created, but also because the graphical nature made the high-level structure more evident. What one participant told Betsy in her evidence-based study sounds much like what Michael and Henry want students to achieve:

I like the top-down design (in Alice projects). We are able to break down the bigger problems into smaller problems and then even smaller problems, so it’s a simpler way to file through. I think you could probably apply top-down design to anything in life.

The biggest concern I have for making education decisions without evidence is who is making the educational decisions for whom. Most CS faculty (reflected in the authorship of these articles) are white and male. Because we are CS faculty, most of our experience is rooted in being students in STEM fields.

Our intuition is most likely wrong about non-CS majors. Most non-CS majors are not students in STEM. Non-CS majors are much more diverse. We don’t know their experience, their goals, or their desired careers. We should not assume that we know how to teach non-CS majors. We should not assume what non-CS majors need to know. When we want to know what CS majors will need in their careers, we ask industry advisors (as the CS2013 curriculum process did, see article here). We know much less when non-CS majors might need to know in their careers. We need to do the research and gather evidence to determine what non-CS majors need to know about computing.

We need the humility to recognize what we don’t know, and we need evidence to inform our decisions. You don’t want your surgeon to apply the best practices of folk medicine on you. Our students deserve the best educational practices informed by evidence, not folk pedagogy informed by intuition.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More