I want to pose a question for consideration. There is a great deal of debate and disagreement over what Computational Thinking means. For some it describes how computer scientists go about what they do, akin perhaps to the scientific method for scientists (Wolfram, 2002), and is applicable only to computer scientists. For others it is a skill set that has implications beyond the field of computer science, a set of generalizable skills of benefit to all (Wing, 2006). A third view is that it represents something of a new mode of thought capable of unique explanations (Papert, 1980) and knowledge building. In this sense it goes beyond a set of procedures, like the scientific method, and might represent a mode of thought distinct from the paradigmatic (argumentative) and narrative modes of thought proposed by Bruner (1986).
The paradigmatic mode represents knowledge founded on abstract understanding or conceptions of the world,. For example, I could explain why an apple fell to the ground by referencing the theory of gravity. This is largely the language and understanding of Science. The narrative mode of thought represents an understanding of the world founded in human interactions. I might explain why an apple fell by referencing a sequence of events in which my elbow knocked it off the table and I was not deft enough to catch it. Of course there is a continuum along which both modalities of thought intersect and interweave. So, my question is whether computational thinking represents a separate mode of thought in its own right, or simply new combinations of paradigmatic and narrative modes. If I were to model a world of apples, elbows and tables, my understanding of why apples fall might be based on a more complete understanding of how apples behave under different circumstances. The use of computational models allows for new ways of understanding the world, new ways of gaining understanding and knowledge. Chaos Theory, for example, emerged out of computational model building. Paradigmatic formulations of the world followed from computational modelling, rather than the other way round.
When we create a computational model of a weather system and run our algorithms through computers with slightly different inputs to make a hurricane path forecast, for example, or use machine learning algorithms to predict heart disease more accurately, are we deploying a new kind of thought which is somewhat different from both paradigmatic and narrative modes?
The need to ask this question rests, perhaps, on the rapid development of Machine Learning and how it threatens to disrupt our world. Machine Learning has brought us to a point where we might soon be farming most of our thinking to intelligent machines. And while probabilistic approaches to artificial intelligence allow human beings to trace back what the machine has done with our algorithms: neural networks, with their black box approaches represent thinking that is to a large extent opaque to us. It seems entirely possible then, that in the not too distant future machines will be delivering to us knowledge of the world, and we will not be able to explain the thinking behind it.
The idea of Computational Thinking (CT) has a history, and it is interesting to unpack some of it. The term was coined by Seymour Papert (1980) and popularised by Jeanette Wing (2006) and there is general consensus that it refers to the thinking skills employed by computer scientists when they are doing computer programming, derived from the cognitive processes involved when you are designing an algorithm for getting “an information-processing agent” (Cuny, et al, 2010) to find a solution to a problem. For some, information-processing agents should refer only to machines, but for others it could include human beings when they are performing computational tasks. Differences in how applicable CT is beyond computer science hinges on these nuances of understanding. I have often heard it said that getting students to design an algorithm for making a cup of tea represents CT and if students were to study designing algorithms through leaning to code they would therefore be improving their general problem solving skills.These claims are difficult to assess, but they are important because if CT applies only to the context of computer science, then its place in the curriculum occupies something of a niche, important though it might be. If, however, as claimed, it leads to benefits in general problem solving skills there is a solid case to be made for getting all students to learn programming. Equally, the case for exposing all students to some coding might rest on other claims unrelated to the transfer of CT to other domains.
Let’s start by looking at the claims made by the Coding for all lobby. Wing (2206) argued that CT skills have transferable benefits outside of computer science itself because they entail five cognitive processes, namely:
- Problem reformulation – reframing a problem so that it becomes solvable and familiar.
- Recursion – constructing a system incrementally on preceding information
- Decomposition – breaking the problem down into manageable bites.
- Abstraction – modelling the salient features of a complex system
- Systemic testing – taking purposeful actions to derive solutions (Shute, et al, 2017)
Wing’s claim has received a great deal of attention and has become the bedrock for the Computer Science for All movement, the idea that all children should be exposed to CT, by teaching them to code, both because such skills will become increasingly important in an increasingly digital world, but also because they equip students for the kinds of problem solving that is increasingly important. It is debatable, though, as to whether these cognitive processes are unique to computational thought. Abstraction and decomposition, in particular, might seem to be thinking processes shared by any number of activities. Wing’s thesis that computational thinking is generalizable to all other fields could perhaps be stated in the reverse direction. Perhaps general cognitive processes are generalizable to computation? This point is not trivial, but still might not threaten the thesis that learning to code or create algorithms is excellent for developing good problem solving skills applicable to other fields.
The question of the transfer of skills gained in one context to another is, however, fraught with difficulty. Generally speaking it seems to me that knowledge and skills are gained within the framework of a particular discipline, and that the application of knowledge and skills in other contexts is always problematic to some extent. There is a close relationship between knowledge itself and what we call thinking skills. It is hard to imagine, for example, anyone possessing dispositions and thinking skills in History or Mathematics without possessing knowledge in those disciplines. As Karl Maton (2014) has pointed out, all knowledge has both knowledge and knowing structures. There is the stuff that is known and the gaze of the knower. In different fields, knowledge structures or knower structures may have greater or lesser relative importance, but one cannot distill out something which is pure knowledge, or pure knowing. Therefore the question of the transfer of skills from one context to another, from one field to another, is not a simple one. Of course we do achieve this feat. At some point in my life I learned basic numeracy skills, within the context of elementary arithmetic classes presumably, and I have been able to apply this basic knowledge and skill set to other contexts, for example computer programming. But I am not so sure that the thinking dispositions I gained while studying History at University, and my appreciation for the narrative mode of explanation are altogether much use when thinking about Computational Thinking and what I ought to be doing as a teacher of ICT skills. I am painfully aware that there are limits to the general applicability of the enquiry and data analysis skills that I learned when training to become an historian. I did not train to become a computer scientist, and therefore I am very wary of commenting on how transferable skills in computational thinking might be to contexts outside the field. But I do believe we should be wary of claims of this sort. Peter Denning (2017) has argued that the idea that all people can benefit from CT, from thinking like computer scientists, is a vague and unsubstantiated claim. For Denning, the design of algorithms (algorithmic thinking) rests not on merely setting out any series of steps, but speaks specifically to the design of steps controlling a computational model. It is context bound.
My understanding from this is that the case for teaching everyone to code cannot rest solely on an argument that CT transfers benefits. This case has yet to be proven. It does not mean that teaching coding to all is not a good thing. I believe that learning to code represents a rigorous discipline which is good for the mind, has benefits because we are living in a world where computer programs are increasingly important, and because coding involves problem solving and this too benefits the mind. All in all I think the case for teaching coding to all is extremely cogent.
I also have this sneaking suspicion that the question I posed in my opening remarks is going to be raised more and more frequently as artificial intelligence gets applied, and if so, having a population trained in some level of competence with computational thinking is probably a really good idea.
Bruner, J. (1986). Actual Minds, Possible Worlds. Cambridge, Mass: Harvard University Press.
Cuny, Jan, Snyder, Larry, and Wing, Jeanette. 2010. “Demystifying Computational Thinking for Non-Computer Scientists,” work in progress.
Curzon, Paul, Tim Bell, Jane Waite, and Mark Dorling. 2019. “Computational Thinking.” In The Cambridge Handbook of Computing Education Research, edited by S.A. Fincher and A.V. Robins, 513–46. Cambridge. https://qmro.qmul.ac.uk/xmlui/bitstream/handle/123456789/57010/Curzon Computational thinking 2019 Accepted.pdf?sequence=2&isAllowed=y.
Denning, Peter J. 2017. “Remaining Trouble Spots with Computational Thinking.” Communications of the ACM 60 (6): 33–39. https://doi.org/10.1145/2998438.
Guzdial, M. 2011. “A Definition of Computational Thinking from Jeannette Wing.” Computing Education Research Blog. 2011. https://computinged.wordpress.com/2011/03/22/a-definition-of-computational-thinking-from-jeanette-wing/.
Maton, K. (2014). Knowledge and Knowers: Towards a realist sociology of education. London, UK: Routledge/Taylor & Francis Group.
Papert, Seymour. 1980. Mindstorms: Children, Computers, and Powerful Ideas. The British Journal of Psychiatry. New York: Basic Books. https://doi.org/10.1192/bjp.112.483.211-a.
Shute, Valerie J., Chen Sun, and Jodi Asbell-Clarke. 2017. “Demystifying Computational Thinking.” Educational Research Review 22 (September): 142–58. https://doi.org/10.1016/j.edurev.2017.09.003.
Wing, Jeannette. 2006. “Computational Thinking.” Communications of the ACM 49 (3): 33–35. https://doi.org/10.1145/1118178.1118215.
Wolfram, Stephen. 2002. A New Kind of Science, Wolfram Media, Inc. https://www.wolframscience.com/nks/