Deconstructing decision trees. Human beings have a hard time thinking in probabilities. Weighing probabilities and using game theory don't come naturally to most of us but they are teachable subjects. Decision trees for project options are particularly valuable but discarded options may come into play later as "branch and sequel" plans if a selected option fails. This is why documenting rationales for decision tree options matter. The article "The Curious Case Of A Broken Crumb Trail" from KMWorld March 2013 describes the utility of creating an Option Outline documenting the reasons why various project options were considered and discarded. This allows project managers to revisit archived knowledge for tips on alternate ways forward if a main effort runs into trouble. I think this approach to documentation is especially valuable for documenting options in decision trees.
KM governance of DM rules. This is a discovery I stumbled onto last year after hearing experts describe the advantages of automated business rule management systems (BRMS) engines. Automating every routine thing is great, and AIs can help, but human decision makers must remain in the loop at the top. Regular manual updates to the KM collection guidance insure that managers capture results that inform strategic key performance indicators (KPIs). Top management must publish those KPIs in media where every subordinate manager can track them.
Business continuity planning. This should be well-established by now in large organizations but small and medium enterprises (SMEs) should also do it. This is more than buying insurance. A business continuity plan should start from the SWOT matrix's identified environmental threats and continue with a risk assessment of hazards graphed in a 2x2 matrix. That's right, folks, I'm talking about severity versus probability once again.
A culture of honesty. I thought about putting this one up front but I decided it would be more effective as my last point. Human beings typically don't use Bayes' Theorem to update their assessments of conditional probabilities; they instead overweight the most recently acquired information. I may have misstated that explanation in the past. Whatever; I have to include this to salve my own conscience. Highly ethical organizations cultivate trust vertically and horizontally, which makes sharing information easier. It also accelerates actions during a crisis because teams that are honest and trusting won't hesitate to execute good decisions. This is my understanding from personal experience with both trustworthy and untrustworthy people.
These principles are worth incorporating into a continuous improvement model. Most people won't do it; they'd rather just wing it through life with no data supporting analysis. That's totally understandable in light of human nature. My thinking is for the handful of people on this planet who truly think for themselves and care about things that matter.