The Committee of Public Safety

Zenpundit has a fortuitously timed post On “Learning Organizations”. It covers many of the same themes from the bottom up as Adam Elkus’s post Can Strategy Be Crowdsourced? (which I linked to earlier) covers from the top down. Zen asks:

Why would the “tactical” level have achieved “learning organization” status and not the “operational” and “strategic” levels of military command?

He offers some suggestions:

• The social networks within the official hierarchical org at the tactical level can effectively leverage both weak and strong ties
• Greater degree of shared purpose and sense of mission
• The tactical level, being a “smaller world” in systems terms than the operational or strategic levels, has a much better “signal to noise” ratio.
• The social networks within the hierarchical org at the tactical level create an environment of greater transparency -discussion may be squelched but situational awareness can’t be.
• Encouragement of critical discussion and incentives for problem-solving.
• Greater tolerance for uncertainty and ambiguity.
• Tacit knowledge is likely to quickly become explicit organizational, knowledge through “shop talk”, the grapevine, de-briefing and formal “lessons learned” dissemination procedures.
• The stress and danger of the tactical environment itself is an incentive to adapt and learn – “Depend upon it, sir, when a man knows he is to be hanged in a fortnight, it concentrates his mind wonderfully.” Dr. Samuel Johnson.

The underlying themes between each of these points which I’d sum up as:

• Tactical OODA loops operate over shorter time-frames and smaller space-frames and involve fewer people per individual loop and more people per group of tactical loops.
• Tactical loops directly interact with external forces and are therefore in a better position (and have a more pressing need) to observe and adapt to “unfolding interaction” with the environment.
• There are generally many tactical loops.

These factors make the tactical level more likely to be iterative learning machines. More tactical loops mean more hypotheses are proposed and tested, This will populate the solution space with more potential answers through more rapid tactical iterations.

Zen then asks:

Can the operational and strategic levels of the military ( or any organization with a bureaucratic structure – schools, corporations, government agencies etc.) become a “learning organization” despite greater scale, distance from events, degrees of abstraction and other obstacles?

This is a similar question to what Adam Elkus raised in his post. Elkus argues that the strengths attributed to crowdsourcing and other “wisdom of the crowd” phenomena are potentially strong on the tactical level but fail to scale to higher levels on the CPSOT stack. The higher layers on the stack, when compared to tactics:

• Have OODA loops that operate over longer time-frames, larger space-frames, and involve fewer people per loop and fewer people overall across all groups of loops.
• Have loops that are, to varying degrees, more removed from direct interaction with external forces. This places them in a worse position and under less pressure to observe and adapt to “unfolding interaction” with the environment since that interaction has to be transmitted through tactics and other intermediate layers (producing a higher “signal-to-noise” ratio, as Zen points out).
• Have (increasingly) fewer loops.

Concepts from Howard Bloom’s Global Brain are useful here to demonstrate the elements the higher layers of the CPSOT stack need to align in order to steal some of that vim and vigor from tactics:

1. diversity generators: agents that generate new hypotheses.
2. conformity enforcers: agents that ensure that agents have enough in common to exchange hypotheses.
3. inner-judges: systems that decide if an individual hypothesis is true (it other words, that it survives) or if something is untrue (if it fails).
4. intergroup tournaments: competitions between groups that test the hypotheses produced by one group against hypotheses produced by other groups. Victory produces truth and defeat is the father of lies.
5. resource shifters: systems that do the dirty work for inner judges and intergroup tournaments. They heavily reward winning hypotheses and heavily punish hypotheses that perform poorly. Bloom summarizes this by quoting the Lord: For he that hath, to him shall be given: and he that hath not, from him shall be taken even that which he hath.

Tactics have more of each of these elements than higher layers on the stack which involve fewer and fewer people. This means fewer hypotheses to loop through and more conformity enforcing gatekeepers. A great deal of what the higher layers do is filter out the multiple hypotheses produced by tactics and its direct interaction with the unfolding environment and use conformity enforcement to pare down the number of hypotheses and make them more consistent. This is similar to how Jeff Hawkins hypothesizes that the human brain works in his memory-prediction framework:

The central concept of the memory-prediction framework is that bottom-up inputs are matched in a hierarchy of recognition, and evoke a series of top-down expectations encoded as potentiations. These expectations interact with the bottom-up signals to both analyse those inputs and generate predictions of subsequent expected inputs. Each hierarchy level remembers frequently observed temporal sequences of input patterns and generates labels or ‘names’ for these sequences. When an input sequence matches a memorized sequence at a given layer of the hierarchy, a label or ‘name’ is propagated up the hierarchy – thus eliminating details at higher levels and enabling them to learn higher-order sequences. This process produces increased invariance at higher levels. Higher levels predict future input by matching partial sequences and projecting their expectations to the lower levels. However, when a mismatch between input and memorized/predicted sequences occurs, a more complete representation propagates upwards. This causes alternative ‘interpretations’ to be activated at higher levels, which in turn generates other predictions at lower levels.

Consider, for example, the process of vision. Bottom-up information starts as low-level retinal signals (indicating the presence of simple visual elements and contrasts). At higher levels of the hierarchy, increasingly meaningful information is extracted, regarding the presence of lines, regions, motions, etc. Even further up the hierarchy, activity corresponds to the presence of specific objects – and then to behaviours of these objects. Top-down information fills in details about the recognized objects, and also about their expected behaviour as time progresses.

Following this framework, you can call your OODA loop an OOPA loop (Observe-Orient-Predict-Act) (or, alternatively, a OOMA loop (Observe-Orient-Moralize-Act)).

Zen concludes by providing requirements for better hierarchies:

[Whether the operational and strategic levels of the military (or any organization with a bureaucratic structure - schools, corporations, government agencies etc.) can become a “learning organization”] depends greatly on two things – creating a “tighter” network with a high velocity of meaningful communication and a new kind of leadership committed to the hard work of re-engineering the organizational culture around adaptive “fitness” and learning.

Makes you wonder how you do Scrum on a mass scale.