What is Double Loop Learning?
To understand double loop learning, we first have to understand single loop learning. While an oversimplification, a good place to start is that single loop learning is reactive, whereas double loop learning is proactive.
Believe it or not, you are already familiar with the concept single loop learning- even if you’ve never heard the term before. It’s the process of problem solving that we are all taught, and the one that we instinctually gravitate towards. We come up with a plan. We put that plan into action. We observe the results, analyze them, and make adjustments based on what didn’t go well. We revise the original plan. We put the new plan into action. Rinse and repeat until our objective is achieved.
This appears, on the surface, to be a good system. In truth it isn’t a bad one, but it is incomplete. The problem is that using single loop learning makes one big, important assumption: that you are starting in the right place when you make you initial hypothesis or plan. Worse, it doesn’t leave open the ability to change that basic assumption- only to tweak the details of it. What if the solution you are working on isn’t even for the right problem? Single loop learning doesn’t account for these kinds of questions.
The way that double loop learning addresses this problem is by adding a third step that occurs before the hypothesis: the assumption phase. It’s a step we all include, but because we do it subconsciously, we don’t recognize it as something we need to adjust.
When we create our hypothesis, we necessarily have assumptions built in. If those assumptions are incorrect, our hypothesis will be incorrect. We can change the hypothesis, but if we fail to recognize our assumptions we will be limited by them, and will only change the hypothesis in ways which still conform to our incorrect base assumptions.
By adding those assumptions consciously into the chain of the decision making process, we allow ourselves to make more significant changes, thus increasing the odds that we come up with the best, most efficient answer.
That isn’t to say that single loop learning is useless. It still allows us to improve, to increase efficiency or ROI, but only within the constraints of those initial assumptions. By questioning our assumptions and expanding the scope of our trial-and-error process, we can find solutions that are more right, more efficient, and move beyond efficiency to efficacy.
Nor is double loop learning in opposition to single loop, in fact it utilizes and includes the process of single loop learning. A hypothesis can be tweaked multiple times and retested based on a set of assumptions, and a satisfactory solution may be reached. What is missing in many individuals and companies is the ability to recognize when that answer eludes them, when it is necessary to question how they arrived at the current process, and to change their initial assumptions in response.
Unfortunately, none of this is as easy as recognizing that single loop learning on its own is problematic and accepting that double loop learning is an improvement. While it is a wide held belief that recognizing a problem and accepting that it needs to change is enough to initiate that change, this is actually yet another example of single loop learning rearing its ugly head- we don’t understand that our unrecognized assumption is that we know enough to fix this problem once we understand it, when in fact we don’t.
A case study by Chris Argyris, one of the originators of double loop learning, showed exactly this. A president at a company was not getting the results he wanted from his VP. He worked on the issue with other executives at the company, and determined that he himself was the cause of the unsatisfactory performance, because he was pushing the VP too hard. He responded as most people would, and decided to back off. Other executives pointed out that this could be taken by the VP as the president no longer investing in their working relationship. He came up with multiple other solutions, none if which produced the effect he was after. He did not, in fact, have the necessary skill set for double loop learning; the ability to understand which subconscious assumptions he was basing his current decisions on. Simply recognizing the problem and understanding the concept of the solution wasn’t enough.¹
The issues of successfully utilizing double loop learning are deeply imbedded in the psyche of most businesses and the people working there. People fear reporting on corporate inefficiency, problems with products, or other systemic issues to their superiors. If they do, those superiors are often afraid to report on them to their superiors. No one wants to be the bearer of bad news, as they are afraid it will reflect badly on them. This mindset leads to a multitude of problems; departments that are underperforming will blame other departments, ones who are over budget will invent reasons why it was unavoidable. Because the cultural norms of most workplaces do not allow for questioning the basic assumptions of the company, no one will do it, and the company as a whole will suffer. When problems never get reported, or get buried along the way, The people who are making the highest level decisions aren’t doing so with the full breadth of knowledge available to them.
Many companies do this to themselves because the emphasis from day one is on suppressing failures, or avoiding the appearance of failure, and on maintaining full control. When a company realizes a product is doing poorly, tries to figure out why, and attempts to fix the issues, that is single loop learning. When a company is able to question if the product should be being made at all before it fails so catastrophically that they have no other choice, that is double loop learning. A company whose executives never know that the product is underperforming will never be able to make that decision- and a company who hasn’t built a culture of the openness to do so, from the top down, will never have that kind of information reach them at all.
This kind of culture is difficult to establish in already established companies, but not impossible. It will often require an outside expert come in and help leadership understand and implement the changes from the top down. It is a slow process, but in the end will be better for the company as a whole, and for every individual working there at every level. In a new company, it is easier, but not easy. There is obvious benefit to structuring a company this way from the first day, but only if the founders have the willingness to honestly examine their own assumptions at every step, and allow others to do the same. We see this in the way a few modern tech companies conduct business, though not as much as many of them would have you believe. At an early stage startup, you will undoubtedly spend a lot of time talking about iterations, and working towards a Minimum Viable Product. These are functions of single loop learning. At certain iterations, you may also discuss a pivot. A pivot, when done honestly and correctly, is an example of double loop learning.
So, pay attention to how your company runs. Do employees, at any level, have the ability to report on the problems they see? And when they do, are those issues not only seriously considered, but measured against the basic assumptions the company made which resulted in the current situation?
If not, don’t be surprised- your company is in the vast majority that is not operating as well as it could be.
¹More on this case study can be found in the paper Argyris published in the September 1997 issue of the Harvard Business Review, Double Loop Learning in Organizations.
