Thinking about thinking (Part 2)
From ESG systems to distrust in society

In the first part of this series on five types of thinking for data scientists [1], I looked at systems thinking at a high-level. In this article, I will take you through the journey of how to use systems thinking in understanding everyday problems we face. A good understanding of the underlying causal structure of problems can help illuminate potential solutions. As there are some excellent sources – books [5,16] and blogs [2–5, 13–15] – on the steps for systems mapping and the tools for systems mapping and thinking, I will focus more on the application of these tools and highlight the steps and tools as part of this process. The interested reader should explore the references in this article to get a more in depth view of the subject.
Environmental, Social, Governance (ESG)
The history of ESG goes back to 2006 when the United Nations-supported Principles for Responsible Investment (PRI) mentioned issues related to ESG and urged companies to adopt a broader set of stakeholder (as opposed to shareholder) capitalism and strive towards sustainable development for the benefit of all. The fledgling ESG movement, which started with 63 investment companies and a total of $6.5 trillion in Assets Under Management in 2006, has now grown to $40.5 trillion globally in 2020 [18].
The reason I have chosen ESG to illustrate systems thinking is that it embodies a number of characteristics that make it ideal for systems mapping. First, ESG is a complex concept that embodies many of the global problems that we face today – climate change, inequality, distrust, etc. Second, the causes and the impact of ESG is global and involves a number of individuals and institutions – literally all of the human population on earth, all of the species on earth, all companies, and political institutions. Third, the timescale of the problem and the impact lasts for decades, if not centuries. An inevitable part of this timescale is delayed effects. Fourth, is that concepts such as trust, transparency, prosperity are qualitative aspects that should be measured and modeled. Also, there are very few other approaches to get a better understanding of such a complex system. Now let’s start analyzing ESG from a systems perspective.
What is a ‘system’?
In my earlier article I reviewed the definition of ‘systems thinking’ [1]. Dynamic, system-as-cause, closed-loop, operational, and forest thinking are some of the key aspects of systems thinking. But what exactly is a ‘system’? What are the essential characteristics of a system? How does one define a system?
A system is something that can be decomposed into simpler parts that are interrelated, whose behavior is constantly changing or dynamic, and the whole is greater than the sum of its parts. The first aspect of the system is its structure and the second is its behavior. The third captures the essential element of a system – remove a component of the system and it no longer behaves like a system. Another way of saying this is a famous quote attributed to Albert Einstein
"Everything should be made as simple as possible, but not simpler."
Capturing this aspect that the ‘whole is greater than the sum of its parts’, Draper Kauffman defines a system as follows
"A system is a collection of parts which interact with each other to function as a whole".
He also provocatively claims that
"dividing a cow in half does not give you two smaller cows."
Let’s get back to ESG. How do we define an ESG system? What are the components of such a system, what behaviors do we see and what makes the system greater than the sum of its parts?
The figure below takes the perspective of a ESG-conscious company and highlights entities in the four major components (4Ps) or pillars of ESG [9]:
- People: In its broadest definition this includes everyone living on the planet today and arguably also addresses the future generations.
- Planet: This includes the entire planet with all its rich natural resources, biodiversity, and the changing planet and ecology.
- Prosperity: This includes not only the material, emotional, and spiritual prosperity of everyone on this planet, but also how the people and planet live in harmony in a sustainable manner.
- Principles of Governance: This includes principles of agency, accountability, and stewardship that ensures sustainable growth for the welfare of all people and the entire planet.
Almost anything that you read today in newspapers, magazines, or your favorite podcast or social media touches upon one or more of these pillars. A few examples of some of these behaviors include [17]:
- Increasing income and access to opportunity: Those with access to opportunity can create better circumstances for learning. As a result, oftentimes, this can lead to better education and future careers. This behavior can be observed within each country.
- Polarizing social media content: People like to read about things that interest them, and social media algorithms deliver relevant content to readers. If the algorithm provides more content that interests a reader, often the result is an imbalance of the alternative point of view across the socio-economic-cultural spectrum. This inevitably leads to polarization of views.
- Decreasing trust: Increasing polarization is causing people to distrust other people and institutions that are not consistent with their views. This distrust can prevent people and institutions from coming together to solve common problems, which in turn, can increase the distrust between the groups.
- Climate change: Growth and prosperity demands more resources and energy consumption. This results in short-term exploitation of natural resources causing greater pollution and CO2 emissions. This along with other factors is causing the warming of the planet and a demand for slowing the growth. Slower global growth disproportionately disadvantages the less developed nations widening the gap between developed and developing nations.
These are by no means the only behaviors we are seeing from the ESG system. There are a number of others and each has a strong inter-relationship between the 4Ps of ESG. Take any one of the pillars out and we would be challenged to understand the other pillars making this a self-contained, albeit a very complex system.
The video below explains the process of system framing. The three key steps of (a) identifying the topic and purpose; (b) identifying the metrics that matter; and (c) identifying key stakeholders are described using ESG as an example.
Icebergs, Telescopes, and Microscopes
One frequently used analogy when describing systems is the iceberg view. What we typically see and also react to is the event that manifests itself or the tip of the iceberg above the water. System thinking is about going below the water level to see the patterns of behavior, the systems structure and the underlying mental model.
Another pair of analogies that come in handy is the telescopic and microscopic viewpoints [3]. When you look at a system we can take a bird’s eye-view or a telescopic view of the problem so we don’t miss out on the big picture. Yet, to be actionable we need to take a microscopic view of the problem and probe the problem at a microscopic level. The art of systems thinking is to frequently shift between these two views to understand the problem, define the boundary of the system under consideration with a clear purpose of what we are trying to solve. As Jay Forrester states
"Formulating a model of a system should start from the question where is the boundary, that encompasses the smallest number of components, within which the dynamic behavior under study is generated".
From the telescopic view of ESG let me take you to the more microscopic view of some key aspects of ESG. Let’s focus on the notion of decreasing trust that may impact the social side of ESG. The tip of the iceberg or the events that we read every day is about the loss of trust in the variety of information sources. The Edelman Trust Barometer has been tracking the trust of individuals across 28 countries and in the past decade, the trust index of traditional media has dropped from 62 in 2012 to 53 in 2021; 62 to 56 for search engines; 45 to 41 for owned media and 42 to 35 for social media [10].
But we need to understand the underlying patterns of behavior to venture below the water level of the iceberg. The behavior pattern of distrust in society is summed up by David Brooks [6,7] and Uri Friedman [8]. Describing the distrust of marginalized groups, David Brooks makes the following comment:
"Most of the time distrust is earned distrust. Trust levels in any society tend to be reasonably accurate representations of how trustworthy that society has been. Trust is the ratio of the times someone has shown up for you versus the times somebody has betrayed you. Marginalized groups tend to be the most distrustful, for good reasons – they’ve been betrayed."
The lack of trust with information sources, with government institutions and employers, and finally with other fellow humans are all closely interrelated. The drop in trust of all groups and especially the marginalized groups also closely mirrors the way they feel they have been treated.
This growing distrust in society over the past couple of decades is not a point in time occurrence but is driven by the system structure. If a group distrusts another group they are less likely to reach out and work with them. As a result both these groups become more insular, communicating and working with others in their own group which builds suspicions and mistrust with other groups leading to a worsening feedback loop which David Brooks calls the distrust doom loop [6].
"The other thing to say is that once it is established, distrust tends to accelerate. If you distrust the people around you because you think they have bad values or are out to hurt you, then you are going to be slow to reach out to solve common problems. Your problems will have a tendency to get worse, which seems to justify and then magnify your distrust. You have entered a distrust doom loop."
Finally, trust itself is a deep rooted mental model – one of the 339 mental models that falls under ‘Models in human nature and judgement’ [12]. Trust involves two aspects – dispositional trust or the trust propensity and situational trust or learned trust [19]. Dispositional trust or trust propensity is the willingness to trust someone else with no prior information or interaction. Situational trust emerges from interaction and is either enhanced or diminished based on reciprocation of trust.
What we have done here is taken the big universe of ESG at the macro-level or with a telescopic lens and then zoomed into the microscopic level of the notion of trust. In the process, I have illustrated how we can go from the tip of the iceberg i.e., the events, to dive deeper into the patterns of behavior, system structure and mental model of trust.
While we can dissect the concept of trust, the question that arises in one’s mind is whether there are observable patterns of behavior that we can reuse from all the mental models that have been studied and whether we have a formal approach and language to capture the system and its behavior. Luckily there are plenty of frameworks and tools that we can use to do this.
Tools for systems thinking: Behavior Over Time, Causal Loop Diagrams and System Archetypes
There are a number of tools for systems thinking [1–5] and I explore three of these tools – behavior over time (BOT), causal loop diagrams (CLD) and system archetypes (SA) – to illustrate the concept of trust that we have been analyzing. All of these are tools for dynamic thinking [4,5].
Behavior Over Time
Behavior Over Time (BOT) in its simplest form is a graph with its x-axis being time and the y-axis being the behavior that is being studied. Don’t be fooled by the simplicity of the graph. This simple behavior over time can capture a variety of behaviors – steady state, random, constant increase, constant decrease, exponential increase, exponential decrease, step function increase or decrease, hockey stick increase, S-curve growth, increasing oscillations, decaying oscillations, and constant-amplitude oscillations. As a data scientist one needs to know the functional form of these behaviors over time. A domain expert on the other hand needs to have observed the behavior in the past or have an hypothesis as to how the behavior will change in the future. The behavior here is of a single variable. While there can be complex inter-relationships between multiple variables over time, the purpose of BOT is to start by capturing the behavior of a single variable. In my future articles I will explore some of these behaviors over time and how they shape our mental models. The video below illustrates a few sample BOTs.
Causal Loop Diagram
In my description above I mention variables but have not clarified what I mean by a variable. Let’s address this as we describe causal loop diagrams. What are causal loop diagrams? John Sterman captures the essence of causal loop diagrams and why they are useful [13,14].
"Causal loop diagrams are an important tool for representing the feedback structure of systems. They are excellent for (a) quickly capturing your hypotheses about the causes of dynamics; (b) eliciting and capturing mental models of individuals and teams; and (c) communicating the important feedback processes you believe are responsible for a problem."
Causal loop diagrams consist of three key elements – variables, links, and loops. Variables represent things or state of being [15]. For example, quantifiable things like cost, revenue, profit, inventory, customers, companies, etc., can all be treated as variables. In systems thinking we don’t have to restrict our attention to just the quantifiable things, we can also have variables for qualitative states of being. For example, qualitative states like satisfaction, happiness, trust, love, etc., can be specified as variables. Just like quantifiable things can be measured e.g., costs going up or profits coming down, the qualitative state of being can also be measured e.g., satisfaction improved or trust decreased. This ability to represent non-physical things or state of being is what gives systems thinking and causal loop diagrams the ability to understand and model behaviors across human psychology, sociology, philosophy, and economics. Variables represent ‘nouns’ from our everyday conversation.
Links have three elements – an arrow that connects one variable (or source variable) to another (or the destination variable), a ‘+’ or ‘-‘ sign, and optionally a delay denoted by a ‘||’ on the link. Links represent ‘verbs’ or actions from our everyday conversation. Nouns exist at a point in time and verbs exist over time. The link below shows two variables – population and births with a ‘+’ sign. This can be read as "An increase in births causes an immediate increase in population". Similarly, there is a link from the variable deaths to population with a ‘-‘ sign. This should be read as "An increase in deaths causes an immediate decrease in population". Given that a causal loop diagram has a number of variables and a number of links connecting them we usually add the phrase "… all else being equal" at the end of each of these sentences I.e., "An increase in deaths causes an immediate decrease in population all else being equal". We can also draw a link from population to births with a delay. This can be read as "An increase in population causes an increase in births with a time delay". Similarly, we can have a similar link with a delay from population to deaths.
A loop is a chain of links that start and end at the same variable. A loop has two elements – the chain of links that creates a feedback loop and the type of feedback loop. Two types of feedback loops are distinguished in CLD’s – reinforcing loops (denoted by a loop symbol and R) and balancing loops (denoted by a loop symbol and B). The diagram below shows that "An increase in population causes an increase in births with a time delay AND an increase in births causes an increase in population". This is a reinforcing loop as this feedback process causes the population and the births to increase – each reinforcing the other. If unchecked this will lead to an overpopulation. The feedback loop that balances this, called the balancing loop is shown on the right side of the diagram. Here the diagram shows that "An increase in population causes an increase in deaths with a time delay AND an increase in deaths causes a decrease in population" thereby balancing the unbridled population explosion. A quick rule of thumb to determine if a feedback loop is reinforcing or balancing is to count the number of ‘-‘ in the loop; if the number is even the feedback loop is reinforcing and if it is odd the feedback is balancing. The video below illustrates the basics of a causal loop diagram with an example.
System Archetypes
System archetypes are common patterns that occur repeatedly across a number of different problem areas or fields that it is worth naming them and studying them. These patterns include various combinations of reinforcing and balancing feedback loops that have been studied extensively. Analyzing our problem and comparing it to these system archetypes can accelerate the understanding of our problem and also seek solutions or policies that have worked in the past. "Drifting goals", "escalation", "fixes that fail", "growth and underinvestment", "limits to success", "shifting the burden/addiction", "success to the successful" and "tragedy of the commons" are eight system archetypes elaborated by Daniel Kim [5]. Let me review two of them called "success to the successful" and "escalation" that will prove useful as we come back and explore the "distrust doom loop".
In the "success to the successful" archetype, there are two groups – Group A and Group B. The more resources that are allocated to Group A the more likelihood of Group A being successful (assuming both groups are equally capable). The initial success of Group A justifies the more allocation of resources to Group A at the expense of Group B (assuming fixed resources that need to be allocated). This in turn reduces resources to Group B, reducing its success and hence decreasing its resource allocation.
In the ‘escalation’ archetype you typically have two groups. Group A takes actions to protect itself that are perceived as threatening to Group B. Group B takes actions to protect itself that is perceived as threatening to Group A. This in turn increases the actions of Group A and the cycle of escalation continues unabated. Some examples of this escalation archetype is when the Group A-B combination is unions and employers; protesters and police; government and rebels, etc.
Wealth Disparity Loop & Distrust Doom Loop
The "success to the successful" system archetype can be used to understand why the rich get richer and the poor get poorer, increasing the income and wealth inequality that some of the ESG efforts are aiming to address. We chose five variables to illustrate the wealth inequality feedback loop – wealth of rich relative to poor, education level of rich, success of rich, education level of poor, and success of poor. The first reinforcing loop shows the ‘rich getting richer’ pattern – "Increase in wealth of rich relative to poor causes an increase in the education level of rich; An increase in the education level of rich causes an increase in the success of the rich; An increase in the success of the rich increases the wealth of the rich relative to the poor". In contrast, we also have the reinforcing loop of ‘poor getting poorer’ – "Increase in wealth of rich relative to poor causes a decrease in the education level of poor; this in turn causes a decrease in the success of the poor which in turn increases the wealth of the rich relative to the poor. It is not difficult to extrapolate from here the growing disparity between the rich and the poor. The figure below shows the wealth disparity loop.

The "escalation" system archetype in its simplest form can be represented using three variables and two reinforcing loops. The three variables are distrust of A towards B, distrust of B towards A, and the collaboration between A and B. An increase in distrust of A towards B causes a decrease in collaboration between A and B; a decrease in collaboration between A and B causes an increase in distrust of A towards B. With two negative links we have a reinforcing feedback loop of distrust of A towards B. The same logic works with B as well. A decrease in collaboration between A and B causes an increase in distrust of B towards A, which in turn results in a decrease in collaboration between A and B. These two reinforcing loops builds accelerates the distrust between the two groups leading to the distrust doom loop that David Brooks talks about [6–7]. The video below illustrates the distrust doom loop.
A simple example of this escalation can be seen in the polarization of the media. If one group (i.e., A) believes that the media channel (i.e., B) is biased, they ignore and don’t read articles published by B. As the audience from Group A decreases, media channel B publishes less articles appealing to A and this feedback loop goes on to create polarization of media.
I will return to the potential fixes or policy interventions for such situations as well as how to formalize these into computation models in future articles.
References
[1] Anand Rao. Five types of thinking for a high-performing data scientist. Towards Data Science, April 25, 2021
[2] Leyla Acaroglu. Tools of a systems thinker. Disruptive Design, Sep 7, 2017.
[3] Leyla Acaroglu. Tools for systems thinkers: Getting into system dynamics and bathtubs. Disruptive Design, Sep 13, 2017.
[4] Daniel Kim. Palette of system thinking tools. Systems Thinker
[5] Daniel Kim. System thinking tools: A user’s reference guide. Pegasus Communications, 1994.
[6] David Brooks. Our pathetic herd immunity failure. The New York Times, May 6, 2021
[7] David Brooks. America is having a moral convulsion. The Atlantic, October 5, 2020.
[8] Uri Friedman. Trust is collapsing in America. The Atlantic, January 21, 2021.
[9] Measuring Stakeholder Capitalism: Towards common metrics and consistent reporting for sustainable value creation. World Economic Forum, White Paper, September 2020.
[10] Edelmen Trust Barometer 2021
[12] Understanding the world with mental models: 339 models explained to carry around in your head.
[13] John Sterman. Fine-tuning your causal loop diagrams – Part I. Systems Thinker.
[14] John Sterman. Fine-tuning your causal loop diagrams – Part II. Systems Thinker.
[15] Colleen Lannon. Causal loop construction: The Basics. Systems Thinker.
[16] Barry Richmond. An introduction to systems thinking. High Performance Systems. 1985.
[17] Blair Sheppard. Ten Years to Midnight: Four Urgent Global Crises and Their Strategic Solutions. Berrett Koehler Publishers, Inc. , 2020.
[18] Robert Eccles and Svetlana Klimenko. The Investor Revolution: Shareholders are getting serious about sustainability. Harvard Business Review, May-June 2019.
[19] Michael G. Collins, Ion Juvina and Kevin A. Gluck. Cognitive Model of Trust Dynamics Predicts Human Behavior within and between Two Games of Strategic Interaction with Computerized Confederate Agents. Frontiers in Psychology, February 2016.