Friday, 30 April 2010

Learning: Learning Models

Continuing my series on software developers learning, here is a brief look at a couple of very interesting and useful learning models. (See also learning styles on Wikipedia for other interesting material).

There are a wide variety of interesting and applicable models, and so this investigation is naturally not an exhaustive survey of what might be applicable to us, as software developers.

This material has been adapted and expanded from my March 2010 CVu Professionalism in programming magazine column.

Learning models

There are a number of very illuminating models of learning that have been constructed by educational psychologists. The Dreyfus Model of Skill Acquisition is a particularly interesting example, postulated by brothers Stuart and Hubert Dreyfus in 1980 whilst they were working on artificial computer intelligence [A Five-Stage Model of the Mental Activities Involved in Directed Skill Acquisition. Stuart E. Dreyfus; Hubert L. Dreyfus. Storming Media. (Available from:]. After examining highly skilled practitioners in fields such as airline pilots and chess grand-masters, they identified five specific levels of understanding:

  • Novice A complete newbie. Novices want to get results fast, but have no experience to guide them in doing do. They look for rules they can follow by rote, and have no judgement to tell whether those rules are good or bad. Given good rules (or luck finding suitable resources on Google), novices can get quite far. Novices have no knowledge of a topic (yet).

  • Advanced beginner At this level, some experience that has lead to learning; you can break free from rules a little and try tasks on your own. But perception is still limited, and you'll get stuck when things go wrong. At this level there is a better understanding of where to get answers (you know the best API references, for example) but you are still not at a level where you can comprehend the bigger picture. The beginner can't focus out irrelevant details, as far as they're concerned everything and anything could be important to the problem at hand. Beginners rapidly gain explicit knowledge – the kind of factual knowledge than can be easily written down and articulated.

  • Competent This stage sees you with a mental model of the problem domain; you've mapped the knowledge base, and have begun to associate parts and understand the relative importance of different aspects. This big picture view allows you to approach unknown problems, and plan methodical routes into those problems rather than diving in and hoping rules will get you to a solution. This is a good place to be.

  • Proficient people move beyond competency. They have a much better understanding of the big picture, and are frustrated with the simplifications that the novice needed. They can correct previous errors and reflect on their experiences to work better in the future. At this point you can also learn from other's experiences and assimilate them into your body of understanding. Proficient people can interpret maxims (as opposed to simplistic rules) and apply them to a problem (e.g. they know how and when to apply design patterns). Now it is easy to identify and focus only on the issues that really matter, confidently ignoring irrelevant details. Here we see the person has gained significant tacit knowledge – knowledge that's hard to transfer by exposition, that is only gained by experience and deep understanding.

  • Expert This is the pinnacle of the learning tree. There are very few experts. They are authorities on a subject; they know it completely, and can use this skill interlinked with other skills. They can teach others (although they probably teach competents better than novices as there is less of a disconnect). Experts have intuition, so rather than needing rules they naturally see an answer, even if they can't articulate why it's the best solution.

Why is the Dreyfus Model interesting? It's a revealing framework to understand where you currently stand in the mastery of a topic, and help determine where you need to get to. Do you need to be an expert? Most people are competent and this is satisfactory (indeed, a team of experts would be far too top-heavy, and probably dysfunctional).

It also illustrates how you should expect to be solving problems at each stage of your learning. Are you looking for simple rules to apply, hungrily gathering maxims to draw experience from, or intuitively sensing answers? How much of a “big picture” view do you have of the topic?

The Dreyfus model is also a very useful aid for teamwork. If you know where a colleague sits on the notice-expert spectrum you can better tailor your interaction with them. It will help you learn how to work with other people – should you give them some simple rules, explain some maxims, or leave them to weave new information into their broader understanding.

Note than the Dreyfus model applies per skill. You may be an expert in a particular topic, and a complete notice in another. This is natural. And should also be a source of humility – even if you know all there is to know about Behaviour Driven Design, you may no nothing about the Joe Bloggs Test Framework. It should excite you that there is something more to learn that may enhance your expertise in BDD, whilst keeping you humble that you aren't an infallible expert in every subject! No one likes a know-it-all.

Shu-Ha-Ri is another interesting model of learning that originates in Japanese Martial Arts. It is a set of three terms describing the stages a student goes through to to achieve mastery in their discipline, roughly translating to Learn, Detach, Transcend. In the Shu phase, the student studies under one teacher; he does not understand enough to mix concepts from different streams, but aims to imitate the teacher precisely. He must build up experience in the field, absorbing the teacher's explicit and tacit knowledge. Once he has a full understanding of the techniques, the student reflects on them in the Ha phase. The knowledge he has learned will not change, but he can interpret, reflect and explore the deeper meaning. Finally, the student progresses to Ri, where he moves beyond studentship; he has original thoughts, and becomes a master himself. He now has the skills and experience to innovate.

Heed this model. Don't try to transcend and innovate until you have completed the Shu phase of your learning (remember to be aware of unconscious incompetence) and have completed the Ha phase of understanding the knowledge.

In my recent conference presentations on this subject, I've also looked at the fashionable "craftsman" model, and seen how the historic guilds system of the pre-industrial revolution applies (and doesn't apply) to our learning. It's an interesting model that bears comparison to shu-ha-ri.

No comments: