The term “Double Loop Learning” was coined jointly by Harvard Professor Chris Argyris and MIT Professor Donald Schőn. They said that when when we take an action that doesn't achieve our desired outcome, we typically respond by changing or modifying our action. They called this Single Loop Learning.  

It is what we do most of the time and it usually works pretty well. We take action that does or doesn't produce a desired outcome. If we don’t get the outcome we want, we reflect and take a different action or adopt a different strategy. If we get the desired outcome then we may have learned what works.

Our actions are, however, made within a context of drivers, assumptions and constraints that Argyris and Schőn call 'governing variables'. These, they argue, guide our strategies and our actions. Double loop learning occurs when we revisit those 'governing variables' ... or to use another term 'key assumptions'.

Argyris and Schőn used the analogy of a thermostat controlling room temperature. The thermostat takes the action to turn the heating up or down, to maintain a specified temperature. This is analogous to Single Loop Learning. The thermostat has no capacity, however, to question whether the specified temperature is suitable for the people in the room. That would, they say, be double loop learning. 

We tend to do a lot of single loop learning in projects. For example, on more than one occasion I have been asked to help with a project that is struggling to achieve its target date or is running over budget. The single loop question is: how can we do things faster or more cheaply? The double loop question is: what is the business value that we are trying to deliver and how best can we achieve it from where we are now? 

An example of planned double loop learning would be to continually focus on the key assumptions that underly the business case, not just focus on delivery budget and cost. This is a key topic within Gary's book, Business Leadership for IT projects.