Productive Failure in Medical Education: Developing visual expertise for differential diagnosis

One key competence for future physicians is differential diagnosis. For many diagnoses, visual expertise – which means reproducibly superior visual skills (Gegenfurtner et al., 2017) – is one central component. Visual expertise is crucial, for example, to classify new disease patterns in medical fields such as dermatology and histology, and to diagnose variations and appearances of diseases that one has never seen before. In traditional medical education, students usually learn a large amount of theoretical medical knowledge. The instruction is typically a memorization- and rule-based knowledge acquisition approach. However, if students memorize knowledge, they tend to forget substantial parts of the knowledge and have difficulty to transfer it into clinical practice (Aldridge, Maxwell, & Rees, 2012). Recent research in the learning sciences offers scientifically robust solutions that can help to address these challenges of traditional instruction. The aim of this project was to use principles of “Productive Failure” (Kapur, 2008) and “Desirable Difficulties” (Bjork & Bjork, 2011) to design for increased learning of visual expertise in service of the fuller competence of differential diagnosis. A special focus was laid on examining whether the learning interventions are not only innovative, but if knowledge is also better transferred and retained by students.

We developed an online tool, which included interactive learning modules for developing visual expertise. Latest learning research, such as productive failure and desirable difficulties, was used to optimize retention and transfer of visual expertise. Consistent with the principles of productive failure, the online tool was available for lecturers in their introductory courses to engage students in visual expertise problem-solving experiences before students received the corresponding formal instruction. For example, students were asked to categorize sets of images without knowing the underlying medical basis of doing so. The student-generated categorizations, even if incorrect or sub-optimal, could then be used to consolidate and assemble the correct knowledge they needed to learn. The interactive online modules were available for students during the whole course of their studies and could be used as a training tool for exam preparation as well as for repetition in clinical practice phases. A first version of the tool, which focuses on one important skill of future physicians, namely the distinction of harmless and suspicious skin lesions, is available at https://dermotrain.lse.ethz.ch.

As intended, we managed to establish a fruitful collaboration between learning scientists and medical experts to a) design an innovative learning tool for the development of visual expertise for differential diagnosis, and b) test the tool and evaluate its effectiveness. Our studies showed that the developed learning modules had positive effects compared to more traditional trainings and learning occasions. In particular, we found a beneficial impact of our approach on long-term performance, on transfer abilities, and in difficult differential diagnosis tasks. We presented our results and the learning tool at an international conference and we are currently working on journal articles to publish our findings.
We think that a key success factor for the project was the multidisciplinary collaboration. The involvement of experienced medical lecturers ensured that the online materials was aligned with the lecture contents, and the involvement of the learning scientists ensured that the material was designed along proven learning techniques. We are convinced that also future teaching projects could profit from bringing together experts from a variety of fields.
Finally, we would like to share some student feedbacks:
“I think I learnt a lot in here”
“Very helpful! would be nice to have an app you could practise it more!”
“Was very cool!”
“This is very useful”
“I found it to be more helpful than expected. It was nice to see the personal progress in numbers in the end, even if my detection skills are not still expert level at all.”
«Cool stuff 🙂 Learning done easy»