Communications of the ACM

sentation, and so forth, in order to get a model to accomplish certain work.

The other is the suggestion contained in the operational definitions that any sequence of steps constitutes an algorithm. True, an algorithm is a series of steps—but the steps are not arbitrary, they must control some computational model. A step that requires human judgment has never been considered to be an algorithmic step. Let us correct our computational thinking guidelines to accurately reflect the definition of an algorithm. Otherwise, we will mis-educate our children on this most basic idea.

Question 2: How Do We Measure Students’ Computational Abilities? Most teachers and education researchers have the intuition that computational thinking is a skill rather than a particular set of applicable knowledge.

The British Computer Society CAS description quoted earlier seems to recognize this when discussing what “behaviors” signal when a student is thinking computationally. 3 But we have no consensus on what constitutes the skill and our current assessment methods are unreliable indicators.

A skill is an ability acquired over time
with practice—not knowledge of facts
or information. Most recommended
approaches to assessing computational
thinking assume that the body of knowl-
edge—as outlined in Boxes 1–3—is the
key driver of the skill’s development.
Consequently, we test students’ knowl-
edge, but not their competence or their
sensibilities. Thus it is possible that a
student who scores well on tests to ex-
plain and illustrate abstraction and de-
composition can still be an incompetent
or insensitive algorithm designer. Teach-
ers sense this and wonder what they can
do. The answer is, in a nutshell, to direct-
ly test for competencies.c
The realization that mastering a do-
main’s body of knowledge need not
confer skill at performing well in the
c In 1992, Ted Sizer of Brown University started
a national movement for competency-based
assessment in schools. 31 He used the term “ex-
hibitions” for assessment events. I gave exam-
ples for engineering schools. 8 According to the
Christensen Institute, competency-based learn-
ing is a growing movement in schools. 34 In 2016,
Purdue became the first public university to
fully embrace competency-based learning in an
academic program in its Polytechnic Institute.

domain is not new. As early as 1958, philosopher Michael Polanyi discussed the difference between “explicit knowledge” (descriptions written down) and “tacit knowledge” (skillful actions). 28

He famously said: “We know more than we can say.” He gave many examples of skilled performers being unable to say how they do what they do, and of aspirants being unable to learn a skill simply by being told about it or reading a description. Familiar examples of tacit knowledge are riding a bike, recognizing a face, or diagnosing an illness. Many mental skills fall into this category too, such programming or learning a foreign language. Every skill is a manifestation of tacit knowledge. People learn a skill only by engaging with it and practicing it.

To certify skills you need a model for skill development. One of the most famous and useful models is the framework created by Stuart and Hubert Dreyfus in the 1970s. 10 They said that practitioners in any domain progress through six stages: beginner, advanced beginner, competent, proficient, expert, and master. A person’s progress takes time, practice, and experience. The person moves from rule-based behaviors as a beginner to fully embodied, intuitive, and game-chang-ing behaviors as a master. Hubert Dreyfus gives complete descriptions of these levels in his book on the Internet. 11 We need guidelines for different skill levels of computational thinking to support competency tests.

The CAS and K12CS organizations have developed frameworks for defining computational thinking that feature progressions of increasingly sophisticated learning objectives in various tracks including algorithms, programming, data, hardware, communication, and technology.d These knowledge progressions are not the same as skill acquisition progression in the Dreyfus model. The CAS framework does not discuss abilities to be acquired during the progression. The K12CS framework gets closer by proposing seven practicese—only three of which are directly d CAS: https://community.computingatschool.

org.uk/resources/2324; K12CS: https://k12cs.org e Fostering an inclusive and diverse computing culture, collaborating, recognizing and defining computational problems, developing and using abstractions, creating computational artifacts, testing and refining, communicating.

ACM’s Interactions magazine explores critical relationships between people and technology, showcasing emerging innovations and industry leaders from around the world across important applications of design thinking and the broadening ;eld of interaction design.