Five questions to ask if you think teaching problem-solving works

By John Sweller

Every few decades there is a campaign to include general problem-solving and thinking skills in school curricula. The motivation is understandable. Everyone would like our schools to enhance students’ critical thinking and problem-solving skills. Because it is so obviously important for students to have such skills, these campaigns are frequently successful in including thinking and problem-solving skill modules in the curriculum. Unfortunately, success in introducing thinking and problem-solving into curricula has not been matched by successful educational outcomes. Across the world, there is a consistent failure to actually improve problem-solving performance. We now know enough about human cognition to know why attempts to teach general cognitive skills such as problem-solving will always fail. Here are a few questions that those advocating for the curriculum to include problem solving in areas such as mathematics might want to consider.

  1. What sophisticated, learnable and teachable problem-solving strategies do you personally use when solving a novel problem? If you cannot describe the strategies you use, what hope do you have of teaching them? At least consider the possibility that there are no learnable and teachable general problem-solving strategies.
  2. Ignoring the problem that no novel, general problem-solving strategies have ever been devised, what evidence is there from randomised, controlled trials that teaching general problem-solving strategies improves problem-solving performance? If, after dozens of years attempting to find a body of evidence for the efficacy of teaching general problem-solving strategies, no such bodies of evidence exist, we must at least consider the possibility that they will never exist.
  3. The relevant randomised controlled trials have been run. Within a cognitive load theory context (Lovell, 2020; Garnett, 2020; Sweller, Ayres and Kalyuga, 2011) dozens of experiments from around the globe have compared learners solving classroom problems as opposed to studying a worked example demonstrating the solution. For novice learners in an area, the results overwhelmingly indicate improved performance by the worked-example group over the problem-solving group. Why? Humans are amongst the very few species that have evolved to obtain information from other members of the species. We are very good at it. We can obtain information by problem solving but it is a slow, inefficient technique. If available, novel, complex information always should be obtained from others during instruction rather than attempting to generate it ourselves.
  4. There is evidence that problem solving can be superior to studying solutions but it only occurs when students are already knowledgeable in the area. They need to practice problem solving. There is no evidence that knowledgeable students are better at solving novel problems outside of their areas of knowledge. Why does practice at solving problems only become effective once we become reasonably knowledgeable in the relevant curriculum area? Cognitive load theory provides an answer that is beyond the scope of this statement (see also Martin and Evans, 2018 ).
  5. Ignoring the lack of evidence from randomised, controlled trials, why do correlational studies on data from international tests consistently demonstrate that the less guidance learners are given when learning, the less they learn? (Note, problem solving is associated with minimal guidance.) (Oliver, McConney and Woods-McConney, 2019; Jerrim, Oliver and Sims, 2019)

As indicated above, cognitive load theory provides one answer to this set of questions. The theory uses our knowledge of human cognition and evolutionary psychology to devise novel instructional procedures. It explains why (a) when dealing with novel, complex problems, studying worked examples is superior to solving the equivalent problems, (b) why solving problems is superior to studying worked examples when levels of expertise have increased in a particular domain, and (c) why attempting to teach non-existent problem-solving strategies, by taking time away from teaching subject matter, reduces students’ performance on international tests.

So, how can we increase problem-solving skill? By increasing domain-specific knowledge. Expecting anyone to engage in sophisticated problem solving and critical thinking in areas where they have minimal knowledge is futile. Lots of domain knowledge allows critical thinking and effective problem solving to occur naturally and automatically. Attempting to teach general problem-solving skills rather than knowledge, does not.

John Sweller is the emeritus professor of educational psychology in the School of Education at UNSW.

Republish this article for free, online or in print, under Creative Commons licence.

2 thoughts on “Five questions to ask if you think teaching problem-solving works

  1. 1. There a few problem solving strategies, e.g., trial-error & iteration, is a simple general strategy that can be easily investigated with students that, in my experience, fosters engagement and investigation, the MATHS300 lessons designed by Charles Lovitt et al., are good examples. I use the lesson ‘problem dice’ every year for Y11 & 12 students as an intro to probability.
    2. I’m not sure about RCT but there are plenty of Meta-analyses that find problem solving can improve student achievement – I can give u many more but here’s 5 – https://drive.google.com/drive/folders/1bASPvttFMbFe_RBrvFo-bRaV8DN1VHOa?usp=sharing

    Also, i think it a bit unfair to use the PISA correlation analysis, when requiring RCT for any PBL you would consider. Correlation is low quality evidence & meta-analysis would be considered higher quality.
    3. Cognitive load is really useful for teachers, but there are still problems with research, e.g., how do you measure cognitive load? The studies i’ve looked at use a 5 point scale for each student to SELF assess. Even though the experiment maybe RCT, the way cognitive load is measured is very unreliable.
    3b. Gijbels et al. (2005), looks at the different effect sizes derived from different tests. They propose since PBL is developing problem solving skills, test which measure these skills will show higher effect sizes than facts tests. I’m think most of the cognitive load studies use fact tests.
    3c. Hattie suggests (I know his research is dodgy, but you used PISA) the Jig Saw method, which is a collaborative & problem solving technique, is the BEST strategy for on going learning.
    4. “There is evidence that problem solving can be superior to studying solutions but it only occurs when students are already knowledgeable in the area.” I agree, but you implied in point 2 there is no evidence?
    5. There are a range of other issues for me – A) student engagement &motivation. B) how do you run a sequence of lessons, say over 5 weeks? Can you use a range of strategies -PBL(experimenting, trial&error, iteration, jig saw, etc & reducing cognitive load strategies? If so, in what order? Has research delved into this complexity ( I have seen 1 study suggesting Explicit teaching first then PBL, but there needs to be much more)?
    c) There are a lot of anecdotes, which I know are not high quality evidence, but all the same, are illustrative. I’ve been fascinated by the Australian Art critic Robert Hughes. I read his biography, and what really struck out to me was Hughes describes being taken to a Sydney Art gallery as a student,.

    His teacher noticed that Hughes scoffed at an impressionist painting and Hughes exclaimed this is not Art!

    The teacher retorted, “then what is Art, Robert?”

    Hughes said that, this one question motivated him to investigate & write about Art for the rest of his life!

    I think CLT is gr8 and useful, but i’m not ready to put the nail in the coffin of PBL yet.

Comments are closed.