The Democratic Backsliding Analysis Project

Project Goal and Strategy

This project was started on January 31, 2023.

The goal of this project is to very significantly help solve the global democratic backsliding problem. Nations are backsliding from democracy to authoritarinism, as see in the graph below in Problem Background.

The backsliding problem is arguably the most important problem in the world, since if it's not solved, nations will be unable to cooperatively solve any other large-scale global common good problem like climate change, war, or high inequality of wealth. This is because authortarians care only about solving their own problems and those of the small group of the rich and powerful supporting them, since the goal of an authoritarian state is to maximize the short-term benefits (profits) of the ruling elite. By contrast, democracies care about solving common good problems, since the goal of a democracy is to optimize the long-term good of all it citizens.

The strategy to achieve the goal is to:

1. Apply the System Improvement Process (SIP) to the problem and keep improving the process until it's mature enough to produce the results needed. This is process driven problem solving, the most powerful strategy known for solving insanely difficult problems.

2. Publish papers in peer reviewed journals on productive results. We choose papers rather than articles since this research is fairly academic, quite complex, and uses specialized tools like root cause analysis and system dynamics modeling. However, the papers are written in "activist style" so that they are as readable as possible by a wide audience.

3. Collaborate with other who are interested in a root cause analysis-based problem solving approach to difficult large-scale social problems, like backsliding and climate change.

4. Build a first-generation proof-of-concept model of the democratic backsliding problem. The model implements the root cause analysis and shows how the system will respond to pushing on the high leverage points.

5. Use all of the above to introduce the powerful tools of root cause analysis and feedback loop simulation modeling to social scientists, particularly political scientists.

Project Status

We are in the revise and resubmit stage of a backsliding paper to the Democratization journal, titled Managing the complexity of the democratic backsliding problem with root cause analysis and feedback loop modeling.

According to two referees, the main reasons for rejection of the first version were:

1. The paper covered too much, so it could not cover it well. It covered a root cause analysis (RCA), a simulation model, and the Truth Literacy Training experiment. The two research questions were "Why is democracy susceptible to backsliding? How can the backslide be slowed and reversed?" About this, one reviewer said: "Dealing with the very important two questions the authors mentioned in a single manuscript is almost impossible. I strongly suggest they separate them into two different papers."

This is an excellent suggestion. It was our mistake to put so much in one paper.

2. The paper "is full of sweeping generalizations that need to be better evidenced and justified."

These generalizations were not identified by the referees. In spite of that, we need to find and address them.

3. One reviewer rejects the contention that root cause analysis can work: "Something can be a root cause even if it does not come with a practical solution. Analyzing a problem based on a wish to find a solution, is perhaps practical, but also a flawed way of analyzing a problem as causes get prioritized (labeled ‘root’) on the basis of being actionable rather than on the basis of their actual causal impact on something. The analysis in this paper, along this line, does not aim to analyze causality of various factors."

This is just plain wrong. The reviewer doesn't understand what RCA actually is, so they make up an illogical, rationalized argument against it. The other reviewer had no trouble with the concept of RCA. We expect that if the RCA was more understandable and persuasive, rejection of the concept of RCA would disappear because the reader had just seen RCA work right in front of them.

4. The main root caused of democratic backsliding was hypothesized to be low political truth literacy (PTL). The referees found this unconvincing, since as the first reviewer said, "if political truth literacy was likely low for a long time, why do we see democratic backsliding intensifying in recent decades?"

This is excellent feedback. We said in the paper that explaining this was beyond the scope of the paper. The paper does not attempt to explain backsliding timing, such as the three waves of backsliding. It only attempts to explain current backsliding. However, we made an mistake and agree with this feedback. The analysis must explain why backsliding didn't occur long ago, since PTL has always been low. What are the factors that caused this and how do they behave?

The fourth reason for rejection is the crux. If we can perform a second pass RCA that persuasively explains why backsliding is occurring now, even though PTL has always been low, then we should be able to use that stunning finding (which is not in the literature) and the Truth Literacy Training experiment to create a strong, focused paper. To avoid covering too much in one paper, we would omit or only briefly refer to the experiment and simulation model.

We are thus currently performing a second pass of the RCA, which is the analysis step of SIP. To improve this step and make results easy to read, we will be using MECE Issue Trees for the first time to drive the analysis step of SIP. This will result in tree diagrams that are easy to grasp and easy to describe. Due to use of MECE, the diagrams should be persuasive.

We have engaged a freelance editor, Paulina Cossette, to review the paper and referee comments, and help us design and edit a paper rewrite. Paulina has a PhD in Political Science, has had her papers accepted, and does excellent work. Her feedback on the paper and reviewer comments have been extensive and helpful.

We will be using agent based modeling. Previously we used system dynamics modeling.

Problem Background

How complex and how serious the democratic backsliding problem is may be seen in the graph below. This is from a widely cited paper: A third wave of autocratization is here: what is new about it? The graph appears on page 1103. The paper is in Democratization, the same journal we submitted to. That link is to their Aims and Scope page. Notice how "authoritarianism" or "authoritarian" appears three times. That's how interested this journal is in the backsliding problem.

This is the same graph used by Wikipedia in its entry on democratic backslidiing. (On January 31, 2023. The graph will eventually change.) That entry is a good introduction to the problem.

Note the Wikipedia section on Causes and characteristics. Possible causes described are populism, economic inequality and social discontent, personalism, COVID-19, great power politics, authoritarian values, polarization, misinformation, incrementalism, and multi-factor explanations. What we don't see is any sign that these causes are anything more than educated guesses. RCA was not used. Once you become familiar with Thwink's social force diagrams, you will probably instantly realize that all these causes are superficial. All arise from deeper causes. For example, why does populism work to convince people to elect authoritarians? Why does economic inequality make backsliding more likely? Etc.

We encountered the same pattern in the backsliding literature. They too did not use any well-structured approach to identifying the root causes. The rejected paper discusses this point at length in the section on Backsliding theory and the two gaps to fill and says:

Waldner and Lust’s review of the democratic backsliding literature found six theory families. These “emphasize political agency, political culture, political institutions, political economy, social structure and political coalitions, and international actors.” Each offers a different set of loosely related factors that could logically contribute to backsliding. None offer a rigorous theory, forcing Waldner and Lust to conclude that “despite the existence of six well-populated theory families, we do not have an obvious theoretical framework for explaining backsliding.” This is the theory gap.

(Then going all the way to the end of the section, we conclude that:)

From the viewpoint of root cause analysis, Figure 1 allows us to see at a glance why the theory gap that Waldner and Lust found exists: we lack a theory that explains the root causes of backsliding and how they can be resolved. This is because backsliding researchers have committed what Jones13 found to be the most common reason for complex problem-solving failure: lack of a properly structured analysis. This is the method gap, which has caused the theory gap.

The high level problem to solve is democratic backsliding. But academics want to know what is the knowledge gap in the literature, which is the author's problem to solve. In this case there are two gaps: (1) We lack a theory explaining why backsliding occurs, and (2) We lack a method of finding that theory. Those are our problems to solve.

Reading the rejected paper, its appendix, and the related follow-up paper on Truth Literacy Training will bring you up to speed on more of the problem's background.

Managing Problem Complexity

This is done by using a process that fits the problem. For this class of problems Thwink developed the System Improvement Process (SIP). The process and its use are described at length in the book Cutting Through Complexity.

Risk Management

The main risks at this point appear to be:

1. The analysis does not address one referee's concern: "as political truth literacy was likely low for a long time, why do we see democratic backsliding intensifying in recent decades?" In other words, if political truth literacy (PTL) was always low, why didn't backsliding begin long ago?

Resolve by focusing on analyzing this behavior clearly and correctly from the start. This risk is the crux of the entire project. Get it right, and the rest follows easily and we have made a strong, crucial contribution.

2. Paper readers cannot understand the analysis.

Resolve by using MECE Issue Tree diagrams instead of feedback loops when possible, change the papers to cover much less per paper so we have more room to explain things clearly, use the simplest argument possible, etc.

3. Paper readers are not persuaded the analysis is correct.

Resolve by take an airtight approach to the analysis argument. Use of MECE Issue Trees should help greatly, since if applied correctly that is automatically airtight. If the model can mimic graph behavior, makes structural sense, and gives deep insights that fit the literature well, then the model should persuade readers the analysis is correct. To maximize correctness, use MECE, SFDs, system dynamics modeling, and agent based modeling as needed.

4. The analyst, Jack, is new to agent based modeling.

Resolved by Jack first used an existing ABM framework, Ventity, to develop a prototype model and learn the concepts of ABM modeling. Ventity was particularly good for this due to its state-of-the art design. Jack then proceeded to create an ABM framework in JavaScript, HTML, and CSS, and then use the framework to build the model.

The Agent-Based Simulation Model

The JavaScript agent-based model (ABM) is here. The status is the first version is complete. Model development began on February 20, 2023. The first verison was completed on June 27, 2023. All feedback loops in the design below have been implemented with realistic behavior, using a framework designed to be easy for others to reuse.

The goal of the model is to produce a simple JavaScript framework for running ABMs and a first-generation proof-of-concept model of the democratic backsliding problem. There are millions of JavaScript programmers, so this will open the door for political scientists to more easily apply root cause analysis and simulation modeling to the backsliding problem.

The model will eventually be rigorously described in a paper. The code is freely available to anyone and we encourage its reuse. Below is the model design diagram: