in Deep learning - 21 Jan, 2017
by Art Murray - no comments
Building the deep learning enterprise

We are well into the knowledge transfer crisis, brought about by the wave of retirements and ensuing panic as organizations scramble to capture, retain and grow their institutional knowledge. As the urgency intensifies, organizations quickly realize it’s not as simple and straightforward as saying to their top experts: “Tell me everything you know about “X.”

Surface-level vs. deep structure

Because of the time urgency, experts are often forced to encapsulate their knowledge into bite-sized, memorable “nuggets.” For example, a long-time pumping station operator in a chemical processing plant might give an apprentice a set of catchy phrases like: “When the pressure’s high, bleed the tank dry; when the pressure’s low, increase the flow.”

The apprentice applies the rules and does rather well. That is, until the pressure starts fluctuating wildly. Being totally unprepared, the novice operator can’t decide what to do. This is because the underlying sense that created the rules in the first place is missing.

The expert, now long gone, possessed something innate that didn’t get imparted to the novice—the capacity to deal with novelty. The expert knows when to use the rules, and when to break them. But ask that same expert to explain how to tell the difference, the response usually goes something like: “I don’t really know why I know, or how I know … I just know.”

Oversimplifying knowledge about complex systems and processes fails to consider the stratified deep structure involved in human sense making. Breaking the surface-level learning barrier means transferring not only the rules but also the underlying processes generating the rules. That can only occur through repetitive cycles of observation, self-directed inquiry and self-discovery.

Transferring knowledge at a deep structure level

When an expert knows something but can’t explain it, it’s a strong indication of the presence of deep knowledge. In neuropsychology, deep knowledge consists of basic elements hidden in memory, at the engram level, along with rules for assembling those elements in order to: 1) assess a situation, and 2) determine what action(s) to take.

The military calls this the “OODA loop” (observe-orient-decide-act). The best leaders and strategists have mastered it, usually after years of war gaming, being mentored, and if the opportunity presents itself, direct experience in combat or crisis situations. The important thing to remember is that much of this deep knowledge can’t be transferred simply by memorizing rules and formulas.

Here are five simple steps you can take to apply deep structure learning methods to your knowledge transfer efforts.

1. Have the expert build a list, or better yet, a visual map, of topics relevant to the problem domain. Ideally, the topics should be arranged into a learning space, with an imaginary boundary separating the topics the apprentice knows from the topics the apprentice needs to learn.

2. Have the apprentice select and name a topic to be learned. In a subtle way, naming imparts a sense of ownership and unconsciously initiates the process of self-inquiry, a key component of deep learning.

3. Have the apprentice explain the selected topic in the form of a written narrative. Handwritten notes have proven to be far more effective at stimulating deep learning than typing, or worse yet, passive listening.

4. Have the apprentice write a completely different explanation/illustration of the same topic. This stretches the learner into viewing a situation from different perspectives. This process should continue until the expert is satisfied that the full dimensionality of the topic has been explored.

5. Have the apprentice write a summary narrative about the topic and its relationship to other topics in the learning space.

This process of guided self-discovery applies a principle borrowed from mathematics: knowledge about a topic is canonical if and only if it is orthogonal and complete. Like the pumping station rule mentioned earlier, information passed along in tweets, rapid-fire Q&A, e-mails and many other so-called attempts at knowledge transfer is anything but canonical. The five steps outlined above are designed to achieve the orthogonality and completeness found lacking in shallow approaches based on rote memorization.

Another important aspect is that deep learning creates long-term changes in the individual. Such changes are generally positive. Self-directed learning and self-discovery improve long-term retention. This allows the learner to grow more confident and enhance his or her self-image.

Think about the richness of this approach, especially when used along with more mainstream knowledge transfer tools such as rules, checklists and flow diagrams. Shifting responsibility for learning from the expert to the apprentice creates an awakened sense of ownership, along with the responsibility to demonstrate what has been learned. Applying these techniques will help break down organizational barriers and silos as people begin to understand at a deeper level the interrelationships among the many topics, perspectives and disciplines at work in their organizations.

A new vision for the enterprise

The goal of the deep learning enterprise is not only to transfer institutional knowledge. By making the above five steps an integral part of your organizational learning efforts, your topic space will continue to expand, along with your capacity to create new opportunities for innovation, as well as perceive and respond to risk. The increased capacity for innovation and learning will help your enterprise become more responsive to, and even lead, the changes in your market.

Updated from an article originally published in the KMWorld Magazine series “The Future of the Future,” February 2015.

To learn more about deep learning, visit adaptivedeeplearning.com.

Leave a Reply