Communications of the ACM

Conclusion

Promoters of computer science have long believed computational thinking is good for everyone. The definition of computational thinking evolved over 60 years applies mainly to those involved in designing computations whether in computer science or in other fields. The promoters of computer-science-for-all, believing that “designing computations” is an insular computer science activity, sought a broader, more encompassing definition to fit their aspiration. The result was a vague definition that targeted not only designers but all users of computational tools, anyone engaging in step-by-step procedures, and anyone engaging in a practice that could potentially be automated. Teachers who find the vagueness confusing have asked for a more precise definition that also clarifies how to assess student learning of computational thinking.

My advice to teachers and education researchers is: use Aho’s historically well-grounded definition and use competency-based skill assessments to measure student progress. Be wary of the claim of universal value, for it has little empirical support and draws you back to the vague definitions. Focus on helping students learn to design useful and reliable computations in various domains of interest to them. Leave the more advanced levels of computational design for education in the fields that rely heavily on computing.

In the late 1990s, we in computer
science (including me) believed ev-
eryone should learn object-oriented
programming. We persuaded the Ed-
ucational Testing Service to change
the Advanced Placement curriculum
to an object-oriented curriculum. It
was a disaster. I am now wary of be-
lieving that what looks good to me as
a computer scientist is good for ev-
eryone. The proposed curriculum for
computational thinking looks a lot
like an extended object-oriented cur-
riculum. This is not a good start for
a movement aiming to define some-
thing greater than programming. Ear-
ly warnings that the object-oriented
vision was not working came from
the front-line teachers who did not
understand it, did not know how to
assess it, and could not articulate the
dence but have not found any. One
of the most notable studies, by Pea
and Kurland in 1984, found little
evidence that learning programming
in Logo helped students’ math or
general cognitive abilities. In 1997,
Koschmann weighed in with more
of the same doubts and debunked
a new claim that learning program-
ming is good for children just as
learning Latin once was. 20 (There
was never any evidence that learning
Latin helped children improve life
skills.) Mark Guzdial reviewed all the
evidence available by 2015 and reaf-
firmed there is no evidence to support
the claim. 14

Guzdial does note that teachers can design education programs that help students in other domains learn a small core of programming that will teach enough computational thinking to help them design tools in their own domains. They do not need to learn the competencies of software developers to be useful.

Finally, it is worth noting that educators have long promoted a large number of different kinds of thinking: engineering thinking, science thinking, economics thinking, systems thinking, logical thinking, rational thinking, network thinking, ethical thinking, design thinking, critical thinking, and more. Each academic field claims its own way of thinking. What makes computational thinking better than the multitude of other kinds of thinking? I do not have an answer.

My conclusion is that computational thinking primarily benefits people who design computations and that the claims of benefit to non-designers are not substantiated.

these professionals may become computational designers when they modify tools, for example by adding scripts to document searchers—but not everybody. It would be useful to see some studies of how essential computational thinking is in those professions.

Another claim suggested in the operational definitions is that users of computational tools will develop computational thinking. An architect who uses a CAD (computer aided design) tool to draw blueprints of a new building and a VR (virtual reality) tool to allow users to take simulated tours in the new building can set up the CAD and VR tools without engaging in computational thinking. The architect is judged not for skill in computational thinking but for design, esthetics, reliability, safety, and usability.f Similar conclusions hold for doctors using diagnostic systems, artists drawing programs , lawyers document searchers, police virtual reality trainers, and realtors house-price maps. Have you noticed that our youthful “digital natives” are all expert users of mobile devices, apps, online commerce, and social media but yet are not computational thinkers? As far as I can tell, few people accept this claim. It would be well to amend the operational definitions to remove the suggestion.

Another claim suggested in the operational definitions is that computational thinking will help people perform everyday procedural tasks better—for example, packing a knapsack, caching needed items close by, or sorting a list of customers. There is no evidence to support this claim. Being a skilled performer of actions that could be computational does not necessarily make you a computational thinker and vice versa. 13, 14 This claim is related to the idea I criticized earlier, that any sequence of steps is an algorithm.

The boldest claim of all is that computational thinking enhances general cognitive skills that will transfer to other domains where they will manifest as superior problem-solving skills. 3, 37 Many education researchers have searched for supporting evi- f If the architect were to specify how to erect the building by assembling 3D printed parts in a precise sequence, we could say the architect thought computationally for the manufacturing aspect but not for the whole design.