Heuristics vs. Best Practices
In this post, I’d like to explore an important anti-pattern that I’ve seen emerging more and more recently, using Dave Snowden’s classic piece:
The anti-pattern that I’ve observed is fairly straight forward: in a genuine attempt to foster learning and help people avoid past mistakes in dealing with a complex challenge practices rather than principles are being ascribed, leading to sub-optimal outcomes.
Let’s start with a more concrete definition of the two terms in the title of this post:
- Best Practice — a procedure that is prescribed as being correct or most effective
- Heuristic — a loosely defined rule (“rule of thumb”)
As well as Snowden’s definition of three types of problems (there’s also a fourth that is less relevant to this post):
- Simple problems — clear cause-and-effect relationships are evident to everyone; a right answer exists; known knowns; Example: solving a simple arithmetic problem
- Complicated problems — expert diagnosis required; cause-and-effect relationships are discoverable but not immediately apparent to anyone; more than one right answer exists; known unknowns; Example: diagnosing a car malfunction
- Complex problems — no right answers; emergent instructive patterns; a need for creative an innovative approaches; unknown unknowns; Example: forecasting the weather
As Snowden points out, (best) practices make sense only in the first two types of problems, the simple and complicated. They differ from one another by the level of expertise required to uncover the relationship between cause-and-effect, but once you have, following the best practice algorithm/checklist will lead you to the best possible outcome.
In complex problems, best practices don’t work. Solving complex problems is all about the unique context that you’re currently in, and what was the best solution to this problem in one context/situation may not be the best way to solve it in another. This is the realm of heuristics and principles, pointing you in the general direction of ways to classify the problem, and general approaches that seem to work for a specific pattern. But they cannot ascribe a specific solution that’s guaranteed to work.
Almost any problem that involves humans tends to be a complex problem. As multiple research papers seems to suggest, when it comes to human behavior, the relationship between cause-and-effect often times are murky at best, putting it squarely in the “complex” category.
And this is what I believe most people seem to (implicitly) fail to consider, when they ascribe a “best practice” solution to a human-oriented problem…