Systems Thinking for Knowledge Management

Using Causal Loop Diagrams to Grow Your Organization’s Collective Intelligence

Figure 1: A causal loop diagram for my writing motivation includes a reinforcing loop (R) and a balancing loop (B)—image by the author.

This blog covers how we can use Systems Thinking to understand the dynamics of building your organization’s collective intelligence. We will review the key Causal Loop Diagrams (CLDs) as models we use to understand organizational knowledge capture and retention. Then we describe how we can use these models and apply leverage in strategic places to increase our organization’s collective intelligence cost-effectively.

We create CLDs because we need accurate models of the world around us. We need to understand and share these models with our strategic planning teams, understand the leverage points we have, and how to change an organization’s behavior cost-effectively. Causal modeling is sometimes considered difficult since causation is profoundly different from simply documenting simple data flows between systems. My hope is by starting small, you can get more familiar with this powerful tool.

This post goes back to the roots of my passion for helping organizations manage knowledge. Many readers think of me as “the knowledge graph guy.” And that has been a focus of my work since 2017. However, this work builds on over 25 years of work on general knowledge management solution architecture, organizational psychology, content publishing, search, databases, and ontology design.

Knowledge management causal models are an important archetype in many organizations with many knowledge workers. They are an essential component to any course on System Thinking in what Peter Senge called a “learning organization.”

The Brain Drain Crisis

Figure 2: High turnover from high-demand skills. Image generated by Stable Diffusion and edited by the author.

Today, many companies are having difficulty managing their internal knowledge. Internal human resource policies for large companies often limit raises to values substantially below inflation. Outside companies often provide salary increases substantially above inflation to attract critical knowledge workers. As a result, high turnover in many areas causes knowledge to “walk out the door,” causing many project setbacks and failures. Turnover in many high-tech areas, such as software development, machine learning, data science, and knowledge graphs, can typically be over 30% per year. As staff picks up these valuable skills, they see their peers making more money by switching companies.

Replacing workers with new employees can be very expensive. Typical costs to bring new employees up to speed typically run from $50K to $100K. In large companies, it sometimes takes years to know who to contact about various problems.

So how can we use Systems Thinking to retain more of this valuable knowledge? Let’s build some simple causal models! We start with a simple model based on a blogger or someone who writes documentation for internal or open-source projects. We will then generalize these models to organizational knowledge management and discuss the critical leverage points we can apply for positive change.

Intranet Black Holes

Figure 3: The intranet back hole: where authors put in content but get zero feedback. Image by Stable Diffusion and edited by the author.

Let us start with a brief discussion of the psychology of why people contribute to the knowledge of their community. And to make it very concrete, let me use myself as an example. I am fortunate to be in a community of readers that appreciate my writing. And my readers have rewarded me with lots of feedback. They “like” my articles on LinkedIn, they give me “claps” on Medium.com, and they “star” my GitHub repos. They can add comments, and if they like what I work on, they do pull requests with new content for my STEM lesson plans and my online microsites and books. They share their insights on the topics I cover. I love this feedback, which inspires (i.e., causes) me to write even more.

Writing and teaching also compel me to think clearly and organize my thoughts. Writing takes time and requires uninterrupted focus. This is a rare skill in today’s Tick-Tok-driven world with 15-second attention spans.

However, many good writers don’t do writing for their company’s internal intranet sites. One of the reasons is the lack of feedback intranet sites provide. I was once asked to write an article for an internal intranet site that discussed issues around AI, machine learning, and knowledge graphs. But after I viewed the website, I realized that there was no feedback for the authors:

1. No indicators counted impressions, views, or reading time of my articles.
2. I could not tell who was reading my content.
3. There was no way for readers to add comments to the article’s website.
4. There was no way to tell what topics were the most popular or what other authors were popular.
5. There was no way to “like” an article.
6. I could not go to my manager and say, “Look! I have influenced hundreds of my peers” in ways that align with my department’s goals.
7. There was no way for my manager to give me measurable objectives as part of my annual performance reviews that reflected my influence on the organization.
8. I couldn’t challenge my peers to see who could write an article that got more views.

The writing process was painful. I had to submit articles in MS Word, and a non-technical communications person might convert it to HTML after a few weeks. Just putting images in was painful.

Worst of all, with content management systems constantly changing, I had no guarantee I could send people a link to my article a few weeks after I wrote it. Broken links were everywhere! I couldn’t even put a link between two articles and be assured the links would work in a week. There was no organizational commitment to knowledge as a permanent link. There were no Permalinks or PURLs.

In summary, writing on intranets can be like throwing your content into a big black hole. Lots of work and information goes in, but no information comes out. It is a real negative experience.

We can describe this feedback system using a causal loop diagram or CDL at the top of this post in Figure 1. These diagrams have feedback loops with arrows representing the causal relationships between things we can measure. The diagram at the top has a “stock” that has my writing quantity in the center.

Reinforcing Loops

Figure 4: A positive feedback loop reinforces my desire to write blogs. The “R” is for Reinforcement, and the plus sign shows an addition to the number of articles I write—image by the author.

The loop on the right is connected to positive feedback; the feedback I get causes me to write more. Positive feedback causes more writing. The letter “R” is in the center of this circle and stands for a “Reinforcing” causality loop. The plus sign near the My Writing shows that the loop positively increases the “store” of articles. This is the model that happens for me on Medium, LinkedIn, Facebook, Twitter, and GitHub.

Balancing Loops

Figure 5: A Balancing loop where we get no feedback from our readers. Image by the author.

The loop on the left side of Figure 1 is when I work hard to write an insightful article and get no feedback. This is an energy drain on me and my career. I feel I get punished for the time I spend away from my normal workload; even though I know that educating my peers is a crucial goal for my company, I have no evidence my writing has been read.

Negative Feedback

Do we get negative feedback on our writing? Yes, sometimes we do. But it is pretty rare. Most social media sites tie your comments back to your account, and there are ways of removing negative people from your peer groups. Negative feedback tends to discourage content creators, so sites like YouTube no longer display the “Thumbs Down” count to anyone but the author. If you don’t agree with someone, sometimes the best method is to ignore them.

As a result of this rarity, we don’t model negative feedback in our CDLs, and I tell other authors to try to listen to their critics and have a thick skin.

Generalizing Blogging Causal Models

Now let’s build a more complex causal model that includes some of the dynamics of what happens when people with essential knowledge leave your organization.

Before we show you these models, we must emphasize that not all knowledge is equally valuable to an organization. Let’s describe the types of knowledge that have high value to an organization.

Explicit Codifiable Knowledge vs. Tacit Knowledge

Figure 7: We need models that help us manage high-value knowledge. A summary of the two types of knowledge. Image by the author.

Before we build a more detailed model for knowledge management stores, we need to talk about the different types of knowledge and how they are stored in documents and code.

Some type of knowledge is easy to express as simple IF/THEN/ELSE rules that work on a short list of clearly defined input properties. We call this explicit knowledge or codifiable knowledge. We have a high explainability of explicit knowledge. Once we write declarative business rules that capture this knowledge, we can quickly explain why we made specific decisions.

Codifiable knowledge is often low-cost to create and maintain. You can purchase general business rules from commercial sources and quickly integrate them into your business processes. AI and machine learning continue to lower the cost of incorporating codifiable practices into information systems and standardize data mapping into sources needed by these rules. Simple rules don’t give an organization a long-term competitive advantage in the marketplace.

In contrast, some knowledge is not so easy to codify. We often talk about “locked up” knowledge in our employees' brains. We call this tacit knowledge. Tacit knowledge is difficult to express or extract and thus more challenging to transfer to others by writing it down in documents, workflows, or code or even verbalizing. The process of putting your knowledge in a blog post or a software algorithm is called codification.

Tacit knowledge requires us to understand the details of a situation's context, which can include carefully weighing thousands of data points. You gain tacit knowledge in a field by having many years of real-world experience, and you rely on your intuition and deep understanding of the relationships and complexities of the world. It is the knowledge we seek to find and retain when we want an organization to be competitive. The people who hold this knowledge are called Subject Matter Experts or SMEs.

A Causal Model for Knowlege Management

Figure 8: What forces impact the store of organizational knowledge? Here we add the impact of staff turnover and internal content—image by the author.

Figure 8 is a slightly larger causal loop model for knowledge management with all organizational knowledge at the center. On the right, we replace the “My Writing” store with a more general “Internal Content,” which can be the collective writing and code shared by your employees. We still have Positive Authoring Feedback that reinforces the store of Internal Content, which in turn reinforces the total store of Organizational Knowledge.

On the left, we now add the causal impact of high turnover. As SMEs with much tacit knowledge leave your organization, the most valuable knowledge decreases (note the negative sign and the “B” in the loop). What is one of the causes of high turnover? One is low raises compared to what a new job offer might be. Meager raises are a positively reinforcing factor that causes high turnover.

Low raises are just one of the many factors that cause high turnover. Meager raises must be compared to the supply and demand for these skills in any industry. If people must be at a specific location, the salaries must be geography adjusted. Human resources departments must continually monitor the short-term supply and demand of skills and predict turnover for subgroups within an organization. Many sophisticated human resources departments can monitor job boards and salary surveys and adjust their new employee offers if their job offers are not getting accepted.

Increasing the Precision of Your Knowledge Management Models

Figure 9: Including knowledge decay, incentives, quality, and profitability into your model. Image by the author.

Now that we know the basics of Knowledge Management CLDs, we can grow our simple model to include other factors to make our model more precise. Model complexity has tradeoffs. Simple models are accessible even to managers without training in Systems Thinking or CLDs. But simple models often don’t capture the actual dynamics of the natural world and don’t reveal the critical leverage points we seek in our analysis. Never assume a manager can read and understand a CLD.

For this exercise, we will keep the Organizational Knowledge store (using a database icon) at the center of our causal model. On the left, you can model all the factors making knowledge stores shrink and become out-of-date. On the right, we will draw causes that increase organizational knowledge and learning. You can then ask questions about the causes of knowledge flowing into and out of your organization.

Knowledge Decay — The Half-Life of A Document’s Value

Figure 10: Documents become less valuable over time. Search tools can lower the relevancy of documents that are old and out-of-date—image by Stable Diffusion.

When we model knowledge stores, we want to model how valuable or actionable stored knowledge is. That PowerPoint slide deck on the bugs in Windows 10 from 2015 — how valuable is it today? Not very useful since we upgraded to Windows 11. But that class on Systems Thinking, which we taught five years ago— is still valid! Systems thinking concepts have not changed much in the last 30 years. You get the idea.

Many documents lose half their value after about a year and another half in the next year. Obsolete documents should be archived or given a lower relevancy ranking in search results. Your knowledge model must include the temporal nature of knowledge and incentive programs for SMEs to purge obsolete content from the System.

Nothing drives people away from an internal intranet more than having the search results filled with outdated documents and broken links. Can you use machine learning and NLP to predict how relevant documents are and change the relevancy ranking based on the half-life of a topic? Can you gain organizational commitments to permanent URLs?

Now that we have a more detailed model of how knowledge lives in the social context of an organization, we can model how we can improve the reinforcing loops and lower the impact of the balancing loops. But not all changes will be equally cost-effective. We need to start to itemize the potential changes we could make to the model and begin to rank them by both impact and cost-effectiveness. The least costly will be our leverage points.

Building Your Own Library of Knowledge Management Microstrategies

Figure 11: Building a library of small strategies. Image by Stable Diffusion.

At the end of this analysis, we will gather a set of small initiatives called our Knowledge Management Microstrategies that we can use to increase our collective organizational knowledge. Our goal is to find cost-effective leverage points! Giving your crucial knowledge creators and maintainers large bonuses will encourage more knowledge sharing, but it might not be the best use of your knowledge management budgets.

Here is just a sample of the most cost-effective leverage points you might consider

1. Encourage everyone to write and present at conferences within and outside your organization.
2. Place a premium on high-quality intranet search.
3. Make knowledge-sharing part of all MBOs.
4. Give authors of internal articles extensive feedback, including how many people viewed their articles.
5. Allow readers to comment on articles and add links to related information.
6. Allow readers the ability to like articles and easily forward them to others.
7. Allow popular articles to get a higher search ranking on your intranet search.
8. Make it easy to author articles and add simple formatting, links, and images.
9. Make an organizational commitment to pertinent and deep linking. Bookmarks and links to sections of a document should never break.
10. Support the ability to link to organizational terminology and concepts quickly.
11. Add support for human-readable permalinks and PURLs.
12. Encourage business terms, concepts, and even subject matter experts to be assigned links and contact information that does not change.
13. Allow readers to view how many top-level executives actually got through their report and read the conclusion at the end of page 10.

Conclusion

I hope that this article was a valuable introduction to knowledge management modeling. We used Systems Thinking and causal loop diagrams to help us understand how organizations can use low-cost leverage points to encourage employees to create and maintain helpful knowledge. These microstrategies will not turn your knowledge management system into a self-aware Skynet overnight. They are tools to find the small steps to help you move in the right direction.

Keep an eye out for my book on Graph Systems Thinking for future archetypes. And please, give me feedback on this article or any other topics you would like to see me discuss.