THE GOAL OF THIS ARTICLE is to teach you the basics of how to read and create causal diagrams. It’s a powerful tool, so powerful that hundreds of thousands of business reports, articles, presentations, books, and academic papers use causal diagrams because they reveal so much and are so persuasive. Wouldn’t it be nice if we started seeing causal diagrams in Kos articles?
Like a good picture, a good diagram is worth a thousand words. Or more.
This is the second article in a series dealing with the topic: Is it possible to take a more effective problem-solving approach to achieving progressive goals? The first article is here.
Chain-style causal diagrams
A causal diagram shows the cause-and-effect structure of interest of a dynamic system. This is incredibly useful in analyzing complex problems. With the right approach to drawing causal diagrams, one can reduce confounding complexity to insightful simplicity. This leads to solution designs that can be predicted to work, because the behavior of the system is understood and therefore predictable, within a broad range.
Above is the simplest possible causal diagram. It has two nodes, cause and effect. The arrow means this causes that.
The above diagram also represents the pattern of the simplest possible problem. For example:
Here racial prejudice causes discrimination. One cannot say racial prejudice is the root cause, since a root cause analysis has not been performed. With causal diagrams like this, one can only say these are the possible causes and effects.
What does discrimination appear to lead to? There are many things, but inequality of income and higher rates of crime are two important effects. That would look like this:
What are some of the important causes of racial prejudice? Family culture, community culture, TV news sources, and social media. Adding those causes (also called factors) to the diagram would give this one:
Next, let’s use our diagram to answer a question. What does social media cause? One effect is inequality of income. The path from social media to inequality of income is known as a causal chain, as shown in the blue boxes below:
The above diagram has a total of eight causal chains that are four nodes long. It can explain eight cause-and-effect relationships that are four nodes long. It can also explain causal chains that are three and two nodes long. For example: What are the causes of discrimination? That can be explained with four causal chains three nodes long. Already the diagram exhibits some useful explanatory power. However, we must remember it was built intuitively, so it is not reliable. It’s over simplified and does not contain root causes and high leverage points that have been validated.
Still, the diagram is much better than no diagram at all, because it can be used to support complex explanations in a clear, consistent manner. In other words, it can be used to build a theory of behavior, which is a cause-and-effect explanation of a behavior of interest.
Theories of behavior are required to solve difficult social problems. The more reliable and complete the theory is, the easier it will be to solve the problem. From the viewpoint of root cause analysis, a reliable and complete theory of behavior is one where all the main root causes and high leverage points are identified and have been validated with measurement and experimentation. We won’t explain how to build such a theory in this article, but in later articles. This article merely introduces the concept of what causal diagrams are and how they are built.
The above causal diagram is a bit loose since it was quickly drawn. Let’s think it over and improve it. Actually, inequality of income causes higher rates of crime. While we’re at it, let’s make our node labels more articulate. This is very important, since a good diagram is easy to read and unambiguous. The result is:
Our causal diagram now says a lot and says it well. The four factors on the left increase racial prejudice. More racial prejudice leads to more racial discrimination. This increases poverty in discriminated groups, which in turn leads to higher rates of crime in discriminated groups, in order to supplement their low income. Note how easy a good diagram is to read. A good diagram is like a good book!
Causal loop diagrams
Now we’re going to introduce a HUGE concept. So far, we’ve discussed only diagrams with causal chains. But that’s not sufficient to model how the real-world works, because all behavior that changes over time is the result of feedback loops. Change of some type occurs in all but the most trivial social problems. Population grows and slows down as it reaches limits, or crashes if it overshoots its limits. Political party membership goes up and down. Political opinion, as measured by polls, voting, and donations, varies constantly. National GDP grows, falls, and occasionally plummets into a recession. Inter-country conflict goes up and down, and sometimes explodes into war. And so on.
A feedback loop is system structure that causes output from one node to eventually influence that same node. The value of one node “feeds back” to itself, as system behavior plays out over time.
Our model of the racial discrimination problem is not yet realistic, because it has no feedback loops. Let’s fix that by asking: What does higher rates of crime in discriminated groups to supplement low income cause? Many things. But let’s pick an important one that relates to the rest of the diagram. It increases “proof” those groups are loaded with worthless criminals in the minds of some people. That in turn increases racial prejudice, and poof, we have a feedback loop. If we add the “proof” node the revised model looks like this:
Follow the loop around and see if it makes sense and tells the correct story about how the system behaves.
I realize “worthless criminals” is derogatory, but that’s exactly the kind of ad hominem attack language that racists use.
When you are building models like this, take your time and go for quality. Give everything a good self-descriptive name, not a short cryptic one. Arrange your models carefully, for readability and to emphasize feedback loop structure. In particular, be VERY careful what feedback loops you decide to model, because each one drives the behavior of the problem.
Up to now I’ve been using draw.io for creating the diagrams. It’s a free very high-quality tool for business diagrams of all types. But it’s not optimal for causal diagrams, so let’s switch to a free tool that is: Vensim. This has the advantage that not only can it create chain-style causal diagrams and causal loop diagrams, it can also create stock-and-flow diagrams and simulate them. The free version is Vensim PLE. It runs on Windows and OSX. Using it for non-simulation diagrams is easy to learn but takes practice. Learning simulation takes some effort and practice. (Please PM me if you need help on learning this tool.) Below is the same diagram drawn in Vensim:
There are two kinds of feedback loops: reinforcing and balancing. Above is a reinforcing loop. In a reinforcing loop, the change that one node causes on itself, by way of the other nodes in the loop, is in the same direction. In a balancing loop, the change is the opposite direction. Reinforcing loops cause growth or decline. Balancing loops “balance” a system by causing growth or decline to approach a goal of some type, such as the way the human body has an internal temperature goal of about 98.6 degrees Fahrenheit. Causal diagrams with loops are called causal loop diagrams.
Adding a balancing feedback loop
All realistic models of real-world systems have at least two feedback loops, one reinforcing and one balancing. Otherwise, they would grow or decline forever (Due to only reinforcing loops.) or would soon grind to a halt (Due to only balancing loops and no reinforcing loops to drive change.). This indicates our model it not yet realistic. Let’s examine it and add a balancing loop.
(By the way, the reason the social force diagrams described in the first article lack feedback loops is they are a simplified, high-level diagram of the analysis. If it’s a difficult problem, one or more causal loop diagrams or feedback loop simulation models are required to penetrate the fundamental layer to find the elusive root causes. These are hidden by the complexity that feedback loops add to a problem. Modeling those loops cuts through the complexity and reveals the root causes.)
If you follow the reinforcing loop in the above diagram around, you will see that currently each of its nodes would grow forever. This is unrealistic. Where would it be easiest but realistic to add a balancing loop?
That would be the racial prejudice node since it has four inputs that we could work with to add the balancing loop. Working on this, I found that four nodes need to be added to create a reasonably correct balancing loop. The result is below:
The family culture and community culture nodes were moved to the upper left. Together these create unbreakable deep social norms, which are essential to good social system behavior. The larger the social system, the more important these norms are, due to the complex cooperation large systems require.
Examples demonstrating the critical importance of social norms are:
- The Golden Rule: “Do unto others as you would have them do unto you.”
- The common criticism: “Have you no shame?” Shame means realizing you have broken unbreakable deep social norms.
- See this paper on Shame in Self and Society, where Thomas Sheff “proposes that shame is the master emotion of everyday life but is usually invisible in modern societies because of taboo.” In other words, “I must obey social norms so as to not feel shame” is the most important social rule humans have.
- A famous quote during the 1954 McCarthy hearings: “Let us not assassinate this lad further, Senator. You've done enough. Have you no sense of decency?” (Here decency means shame.) This was such an incisive statement that it instantly demolished McCarthy’s anti-communist demagogic witch hunt, which quicky unraveled.
One particular social norm can be called maximum racial prejudice allowed in a particular culture. When a person is evaluating their own behavior to determine their own sense of shame, they use their current racial prejudice and maximum racial prejudice allowed to calculate their excessive racial prejudice. This is the amount of racial prejudice that exceeds the maximum allowed.
Here's how to read the balancing loop: When racial prejudice rises enough to exceed maximum racial prejudice allowed, excessive racial prejudice starts to rise above zero. This causes pressure to reduce racial prejudice to also rise above zero. As it increases, racial prejudice decreases, and the loop starts over again.
My apologies for the complexity added by the balancing loop, but that’s how the real world operates. Complex political behavior must be correctly and deeply understood or progressives will simply not be able to achieve their goals.
Now we have a realistic model. The reinforcing loop causes racial prejudice to rise, while the balancing loop keeps it from rising past a certain point. If there was no balancing loop, we would see high amounts of behavior like lynching, murder, and all sorts of extreme violence against people of color. Instead, acts like this are rare. They are, however, on the rise.
One reason they are on the rise is Donald Trump and his supporters are hammering away at the unbreakable deep social norms node. A CNN article on The dangerous consequences of Trump’s all-out assault on political correctness describes how, during his 2015 presidential campaign, Trump’s words had the effect of eroding social norms on things like hate speech, incitement of violence, anti-Semitism, and white supremacy. It’s not hard to see how this increased racial prejudice, because it weakened the balancing loop by eroding social norms.
The model allows us to understand why Trump and his supporters were able to so easily increase harmful social behaviors like political violence and racial prejudice. They did it by eroding social norms. The above article had this to say:
“We have to straighten out our country, we have to make our country great again, and we need energy and enthusiasm,” Trump said during an appearance on “Meet the Press” in August 2015. “And this political correctness is just absolutely killing us as a country. You can’t say anything. Anything you say today, they’ll find a reason why it’s not good.”
People responded – big time. The idea that liberals and/or the elites had made it so that no one could say what they thought, for fear of being labeled intolerant or un-enlightened, was a powerful one in the very communities that Trump was appealing to: Whites watching the society and culture they had grown up with change faster and in ways that, in some cases, made them deeply uncomfortable. (Exit polls in 2016 showed Trump got 57% of the white vote, 8% of the black vote and 28% of the Hispanic vote).
The problem with Trump’s assault on political correctness is that he took it so far that he clearly emboldened not only those lurking in the shadows to bring their hate speech into the light of day, but also lowered the overall bar for what is considered acceptable discourse among politicians and other leaders in the country.
Political correctness is code for unbreakable deep social norms. Trump was essentially saying it’s okay to be racist because you don’t have to fear being labeled as intolerant or un-enlightened. “Lowered the overall bar for what is considered acceptable discourse” is the same as lowering unbreakable deep social norms.
Trump also made many racist appeals related to higher rates of crime in discriminated groups. For example, Trump’s claim that “Inner-city crime is reaching record levels” is wildly untrue. The data shows that average violent crime in large cities has fallen by about half from 1995 to 2014. However, there was a small uptick in 2015. Small statistical variations like this are normal. In 2015: “the 10 cities with the biggest spikes tended to have higher poverty rates, higher African-American populations and smaller Hispanic populations.” This confirms the higher rates of crime in discriminated groups to supplement low income node.
Adding two nodes to the model to capture Trump’s political incorrectness and racist appeals gives the model below:
The model tells an incredible story. It tells how Trump won the 2016 election and is on course for potentially winning the 2024 election. Trump’s claim that political correctness is okay has caused the maximum allowed for many unbreakable deep social norms to go up considerably, such as for hate speech, support of violence, and racism. This increases maximum racial prejudice allowed. When that’s combined with rising racial prejudice due to the reinforcing loop, it takes longer for racial prejudice to rise above the maximum allowed. That reduces pressure to reduce racial prejudice. As pressure falls, racial prejudice goes up. That’s the key behavior Trump and his supporters want, because it’s their biggest motivator. More than anything else, MAGA Republicans want white supremacy. Make America Great Again is code for Make America White Again (source).
The model also shows how Trump is pumping up the strength of the reinforcing loop. It’s because of his implication that higher crime rates in discriminated groups proves those groups are criminal. That causes many citizens to accept higher rates of crime in discriminated groups to supplement low income as “proof” those groups are loaded with worthless criminals. The disastrous result is racial prejudice goes up and up and up.
This is an overly-simplified diagram, since it serves as a simple educational example of how causal diagrams work and it not the result of rigorous root cause analysis from the start. There are many other methods that politicians like Trump use to increase racial prejudice, not just the reinforcing loop shown.
Finding the main root cause
Now we get to something really useful. If the model lets us understand how Trump is winning, can it allow us to understand how to solve the racial discrimination problem?
Of course, because we can think in terms of root cause analysis. The model is not the result of root cause analysis, so it doesn’t say what the root causes and high leverage points are. But for the sake of this article, let’s ask an interesting question: Given this model of behavior, what is the main root cause and high leverage point? In other words, we are about to take a root cause analysis shortcut to illustrate how powerful thinking in terms of root causes and high leverage points can be.
Look the model over. There are three main ways to solve problems once you know the key feedback loops involved: strengthen a loop, weaken a loop, or add a loop.
Above all, Trump and similar authoritarians have figured out how to weaken the balancing loop. They spend much more time weakening the Effect of Social Norms loop than strengthening the Poverty and Crime in Discriminated Groups loop and similar loops. For them, there’s much higher leverage in eroding social norms than in increasing racial prejudice due to the “proof” shown and other ways. That’s a smart move. But we can be smarter because with the right analysis, we can see what they are doing, why it works, and how to best counter it.
How does one best search a model for root causes? For that we start with the definition of a root cause. The first article in this series said:
A root cause is the deepest cause in a causal chain (or the most basic cause in a feedback loop structure for more complex problems) that can be resolved by changing something in the cause, such as stopping it, increasing it, or fixing it.
Thus, our strategy is to search the model for nodes that can be changed in a relatively easy manner to have high favorable impact. Most nodes cannot be changed easily, like family culture, community culture, the TV news sources people decide to watch, the social media they decide to use, and the two nodes Trump uses, the claim and implication nodes. No one can easily change what Trump says. Even though these six nodes are not in feedback loops, they are hard to change. All the nodes in feedback loops are hard to change, because they are strongly controlled by their loops, except those nodes which are also affected by nodes outside a loop. By the process of elimination, this leaves unbreakable deep social norms, maximum racial prejudice allowed, and the “proof” node.
We can’t change maximum racial prejudice allowed, since that’s the norms in unbreakable deep social norms that apply to racial prejudice. But we can change unbreakable deep social norms if we can identify a new node that inputs to it strongly. Likewise, we can also change the “proof” node if we can identify a new node that inputs to it strongly.
This leaves us with two nodes that are good candidates to relate to root causes, but only if we can figure out a new node (or more than one node) that can strongly affect the two nodes. How can we do that? I don’t wish to be too dramatic, but there better be a good answer because the life of democracy is hanging on the answer.
The answer, based on two decades of Thwink.org research, arises from a Five Whys question that applies here: WHY are Trump’s claims and implications working so well?
Because of low political truth literacy. That appears to be the root cause. Because political truth literacy is low, people are easily fooled by the fallacious statements politicians like Trump make to win supporters. Remember now, Trump’s deepest strategy is not promoting white supremacy, identity politics, violence against non-supporters, or any other particular message. It’s lying. That’s why we see headlines like these:
The high leverage point for resolving the root cause is obvious: raise political truth literacy from low to high. Thwink.org has experimentally shown that political truth literacy is low in the US and that it can be raised from low to high with a small amount of Truth Literacy Training. In addition, other practical solution elements for raising political truth literacy have been designed, like the legal right to Freedom from Falsehood from politicians and Politician Truth Ratings. The high leverage point has never been pushed on before in a prolonged, large-scale manner, because the root cause has never been clearly known. This indicates considerable opportunity.
Adding the root cause and high leverage point gives this causal loop diagram:
This is our final diagram because it says everything we need to complete a quick form of root cause analysis and show how causal diagrams work in a fairly detailed manner.
When solution elements push on the high leverage point of raise political truth literacy from low to high, that increases political truth literacy. This in turn causes two things:
- The fallacious claim that political correctness is okay no longer works nearly as well. This prevents erosion of unbreakable deep social norms. Now the balancing loop performs as it should and prevents racial prejudice from growing to an excessively high level. People are now ashamed when they break social norms such as having high racial prejudice.
- The fallacious implication that higher crime rates in discriminated groups proves those groups are criminal also no longer works well. As political truth literacy rises, “proof” those groups are loaded with worthless criminals goes down, and so does racial prejudice. The reinforcing loop no longer causes racial prejudice to grow so strongly.
This completes the goal of this article, which was to teach you the basics of how to read and create causal diagrams. As a bonus, we also showed how to take a causal loop diagram that shows why a problem occurs and find its root causes and high leverage points. That’s a root cause analysis shortcut.
The benefits of using causal diagrams
To summarize, the benefits of using causal diagrams (as a companion tool with root cause analysis) to analyze difficult large-scale social problems are:
- A fundamental principle of systems thinking is: The behavior of a complex system arises from its causal structure. If you don’t understand a difficult problem’s structure, then you cannot analytically solve the problem. Knowing how to build causal diagrams fills this gap.
- Learning how to identify a problem’s causal structure allow you to go one step further and identify its root causes and high leverage points.
- By capturing the causal structure of a problem in a causal diagram, you have a clear explanation of problem behavior that can be shared with others. This opens the door to group analysis, critique and feedback from others, reuse on other problems, use of diagrams as supporting proof of a correct analysis, and so on.
- A good causal loop diagram can serve as the starting point for a feedback loop simulation model. This is required for the most difficult problems.
- Most importantly, you are no longer relying on intuition. Instead, like all true scientists, you are relying on structured analysis of cause and effect. Now your problem-solving stories and actions can be based on structured analysis instead of intuition.
The next article in the series is here.
