"Empirical" Agile?

While reviewing a colleague's discussion of high-speed collaboration in a planning-centric culture (Discovery/Exploration vs. Execution/Transaction), I noticed that he used the term "empirical" to describe the collaboration activity and cited an Agile Development book. Below is an edited reply of my comments to him.

Your use of "empirical" caused me some confusion.....which made me think about why.....here's a few snap reactions:

1. It implies a taxonomy of "empirical" and "non-empirical." At the very least, planning-oriented folks will resist being characterized as "non-empirical".....and, the smarter & more aggressive types may directly attack the taxonomy & its implications...... "if anyone's empirical, we are."

2. I suspect that Agile folks use the term because they're often seen as using an ad-hoc process with few metrics. So, a term that emphasizes their ongoing use of data to continually course-correct is understandable.

3. Planning-oriented approaches are also empirical. The difference is that they assume that many (most? almost all?) significant aspects of a decision context can be represented by a relatively static model. They then create organizations, roles/responsibilities, processes, and metrics to allow them to act consistently and coherently across many instances of that decision context. To the degree that the model fits a specific instance, planning works. The "empirical" parts include (a) model-building, and (b) repeatable & measurable processes.

4. Probing-oriented approaches are (in some sense) less empirical in that they don't have static models of a decision context. Instead, they use Snowden's "narrative fragments", Klein's "frames", and high-speed collaboration to constantly manage the creation/testing/modifying/deletion of multiple working hypotheses.....framing, modifying, abandoning them as needed to best anticipate future needs. They are more empirical in the sense that they have a much broader perspective of the decision context and are much more likely to pick up weak signals that would be filtered out by standardized processes....in other words, their claim to "empiricism" is based primarily in a real-world flow of data that richly informs their activities, while planning-oriented claims to "empricism" are based primarily in a "validated" model with pre-defined metrics (that, by implication, assert that "if it's not measured, it's not relevant").

5. It's easier for me to think about the differences using Cynefin's known/knowable (or simple/complicated) "ordered domains" vs. the complex/chaotic "unordered" domains. Specifically, if it's complex, then you can't model it, you're limited in your ability to plan, and you have to be able to smartly probe for patterns using attractors and limiters (instead of standardized processes and/or expert analysis).

6. However, in the almost 3 years I've been using Cynefin, I've found it difficult to explain....most people don't seem to have good narrative fragments (or frames/models) associated with consciously addressing complex contexts. You have to first make them aware of the complex vs. simple/complicated contrast, then discuss the need for different organizational structures, roles/responsibilities, and processes to address complex contexts. Unless they have a pressing complex context need, they're unlikely to invest the time needed to grasp the value of Cynefin. And, often, they simply impose Analysis & Best Practices on a complex area, leaving those on the edge to bend "the system" to adapt it to each specific complex instance.