What Story Does Your Model Tell?

This article will delve into how economic models interact with society, with a special focus on artificial intelligence (AI) implementation. It will take a look at the “puzzles” prescribed in the book “Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity,” by 2024 Nobel laureates in Economic Sciences Daron Acemoglu and Simon Johnson. The book describes the struggle between capital and labor and who receives the productivity and prosperity gains provided by technology. In short, labor is the workers, and capital is the elite at the top who control and own the technology.¹

Framing the Narrative

Why do I believe this article is important for actuaries? Because it is essential to know that there are many different approaches and theories to understanding the economy. As the economist Nicholas Gruen explains, economics does not come in just one flavor. There is a “Baskin Robbins” of economic choices, such as Neoclassical, Complexity, Evolutionary, Quantum, Ecological and others. We should not look at any flavor as dogma. We should consider them as different modeling toolsets to help us understand problems from many perspectives. These toolsets help us create better narratives.

At the 2024 Society of Actuaries ImpACT conference, I attended “The Power of Story to Turn Data into Insights” workshop. It got me to contemplate how theories transform models into stories. I thought about behavioral economics’ fundamental framing concept. Framing occurs when the phrasing of a question unintentionally changes the perceived meaning. ² In short, the narration of the problem will dictate the solutions used to solve it. Therefore, framing gives way to our accepted narrative to explain the situation and solutions.

Ponder this: whoever controls the narrative controls the perception of the outcomes and their success. The problem is that the widespread economic theory does not stay theory. Assumptions underly models that support theories. Narratives explain model results. The narratives bleed into the culture, which drives politics and develops into policy. Policy funds research, leading to new assumptions and theories. Therefore, as I see it, the governed believes and votes based on this culture because they are the foundation of accepted beliefs.

Credit: Bryon Robidoux

The Neoclassical Flavor

Neoclassical economics is woven deeply into actuarial science and risk metrics, albeit as the generic option. I am not too fond of it because it is the only tool people seem to use and, in my experience, is overoptimistic with its foundational assumptions, which economist and author Steve Keen explains:

• Any movements away from equilibrium are due to external factors or shocks.
• There is only negative feedback, which means that no matter how big the shock is, all results will revert to the current level and come to equilibrium.
• Connections do not matter because all products and services are 100% fungible.
• The system is the sum of just its parts and ignores its interactions.
• Lastly, people have perfect knowledge of all possible outcomes.³

Given these assumptions, Neoclassical economics theorizes people can:

• Make more predictable and rational decisions,
• Accurately predict the likelihood of various events,
• Can optimize their profit, and
• Adjust their behavior to move toward equilibrium.

“I don’t like the word ‘rationality’ because it’s been so tainted within economics,” economist Steve Levitt said in a podcast with Neil deGrasse Tyson. “What rationality has come to mean in economics is a heartless, emotionless calculation that often leads to absurd conclusions that are very much at odds with how society works (because relationships and trust are not entirely fungible). So ‘rational,’ in an economic sense, is an extreme. There’s no room for anything else. And I’m opposed to that.”

Complexity Flavor

Complexity economics derives from complex systems science, which studies how many system components interact. As a massive hierarchy of interconnected systems working in a parallel distributed fashion with no central control, the economy qualifies as a complex adaptative system.

The foundation of complexity science:

• Both positive and negative feedback can occur.
  • Positive feedback is when events can reinforce each other, such as exponential growth due to nonlinear interaction effects.
  • Negative feedback occurs when an excited system comes to equilibrium.
• Interactions and relationship strength become critical.
• The system is the sum of its parts and the interactions between them, so internal and external factors can emerge unexpectedly.
• The future is uncertain.
• The universe is continuously computing its future states. Hence, new events will occur that have not happened in the past.
• We use heuristics and stories to cope with our limitations. (In economics, heuristics are mental shortcuts or rules of thumb that people use to make decisions when there isn’t enough information to make a perfect decision.)

The complexity economics foundation is a theory based on “as-is” assumptions instead of “as-if” assumptions of Neoclassical economics. Here’s one way to describe the Neoclassical vs. Complexity economics view of Value at Risk measurement required by Basel II for banks during the 2008 Financial Crisis. In Neoclassical, correlation measures how much two companies move together due to external factors. The correlation goes from [-1,1]. -1,0,1 implies they move perfectly in opposite, independent and same directions. Correlation completely ignores the movement due to connections or internal factors, meaning companies only move to outside factors. Hence, each bank should be able to implement Value at Risk without impacting the others. It is a positively perceived practice for the banking industry that everyone should do so. The sum of the outcomes is just the sum of the outcomes from each bank. What could go wrong?

However, from a complexity perspective, the internal factors or interactions outweighed the parts. One bank implementing Basel II Value at Risk was a fantastic idea in 2008. But, all banks implementing Value at Risk was not such a good thing. As the market tanked, Basel II required all banks to behave similarly. Rather than Basel II reducing risk, it compounded risk. Similar behaviors caused a positive feedback loop as the market started to drop, which made the crisis more severe.⁴ Therefore, not accounting for internal interactions and relationships ultimately caused regulator’s model, apparently, to overlook a fundamental market dynamic when creating Basell II. Could insurance and actuaries fall into this same trap in the future?

Friedman Narrative

Milton Friedman, a Nobel-winning economist at the Chicago School of Economic Thought, promoted the Friedman doctrine. Chicago economics is the stronghold for the Neoclassical flavor of economics, a foundation of actuarial science. In the 1970s, three concepts influenced today’s accepted economic narrative.

1. The Friedman doctrine, also called shareholder theory, says that a business’s entire social responsibility and goal is to maximize shareholder value by increasing profits. This shareholder primacy approach views shareholders as the economic engine of the organization and the only group to which the firm is socially responsible.¹ Furthermore, economist Michael Jensen suggested that corporate managers should have compensation tied to their stock price to engrain this responsibility deep in the company culture.¹
2. In “Capitalism and Freedom,” Friedman explains the importance of capitalism to an individual’s freedom, which could be a person or corporation.⁵ This perspective was echoed by the Supreme Court in Citizens United v. Federal Election Commission in 2010, establishing corporate personhood.
3. From the Neoclassical assumptions, we get the “invisible hand” theorem, which says that if everyone is out for their self-interest, the markets will self-correct, and the economy will achieve better outcomes.⁶

Narrative Outcome

“Power and Progress” makes it clear that if you look at history from the Industrial Revolution until now, labor rarely receives productivity gains. It happened post-World War II due to all significant economies being destroyed. (Represented in the graph below.) Capital was desperate for labor during this period, so labor had the upper hand. As the world rebuilt after WWII, the money flowed, raising all tides and increasing prosperity for everyone. When the gains plateaued and inflation skyrocketed in the late 1970s, Friedman’s doctrine was promoted.¹

Source: Economic Policy Institute; analysis of BLS and Bureau of Economic Analysis data.

President Franklin D. Roosevelt helped swing the narrative in favor of supporting labor during the Great Depression.¹ In the 1970s, the heavy flow of gains changed to a slow, insufficient trickle because the narrative returned to its norm due to the rise of the Friedman doctrine and supporting narratives, such as Reagan’s trickle-down economics. Trickle-down economics is the concept that government economic policies that disproportionately favor the upper tier of the economic spectrum (wealthy individuals and large corporations) eventually benefit the economy as a whole.

The invisible hand theorem justified abolishing the Rooseveltian regulations that favored labor because it promulgated the belief that the market would automatically fix any injustices. Labor lost its advantage and lost its productivity gains. Stagnant wage growth, atrophying unions and increasing CEO compensation are the evidence of this outcome.¹

Understanding AI

“The father of AI was Alan Turing,” Palm Computing co-founder Jeff Hawkins proclaimed in “On Intelligence.” “He (Turing) wanted to define a test to determine whether a machine is intelligent, called the Turing Test. The Turing test: if a computer can fool a human interrogator into thinking that it, too, is a person, then the computer must be intelligent. The central dogma of AI: the brain is just another computer. It doesn’t matter how you design an artificially intelligent system; it must produce humanlike behavior. The AI proponents saw parallels between computation and thinking because humans and computers manipulate abstract symbols based on well-defined rules to create decisions.”⁷

Let’s take a look at two types of AI: machine intelligence (MI) and machine usefulness (MU). Machine intelligence describes AI that tries to automate tasks that humans usually do, i.e., replace humans with machines. Machine usefulness encourages AI to supplement humans by making them more productive and increasing their capabilities.¹ A takeaway from understanding this brief history of AI and the Turing Test is that AI’s original definition is machine intelligence; hence, most AI projects are machine-intelligence-focused.¹ Furthermore, machine intelligence fits squarely within the Neoclassical storyline because it proposes reducing costs and potentially increasing shareholder value by reducing labor costs. Machine intelligence neglects the computational power of the human community. Given current trends, capital will likely receive the gains from AI projects, leaving labor further behind due to diminishing productivity gains.¹

Changing the AI Narrative

In “Range: Why Generalists Triumph in a Specialized World,” journalist David Epstein asserts that AI is great at finding patterns and doing intense calculations with a fixed rule set, but it is horrible when tasks require creativity, adaptability or innovation. Hence, computers can easily beat chess champions, making their expertise worthless. These games have an immense but fixed, finite event space, which is easy for computers. These traditional games live inside a Neoclassical economic world of known outcomes. However, as openness and uncertainty increase, and rules dynamically change, the computers do dismally in isolation. However, when paired with generalist, non-expert humans, the human-computer duo was on par with or surpassed the experts.⁸

“Power and Progress” explains that machine intelligence projects used to replace humans have dismal returns, except when feeding users their confirmation bias to amass ad revenue. Machine usefulness projects increase people’s skills, making the returns significantly better. The average person becomes more productive. With machine usefulness, labor usually keeps most of the productivity gains. This situation better distributes the productivity gains from labor into the economy.¹ People feel more prosperous and happier.

Therefore, I believe we need to switch the focus of AI from machine intelligence to machine usefulness to make society better able to distribute productivity gains. Human interaction and relationships are critical for a well-functioning economy in our complex world, as explained by complexity economics. The Neoclassical mantra, which prioritizes shareholder gains, minimizes the importance of other parties vital to the economy’s function. To make this change, we should consider changing the narrative. To change the narrative, we should move beyond the Neoclassical point of view and incorporate other economic theories into our discipline. We should move beyond a narrative that focuses on personal freedom and praises narcissism. We should balance the needs of the collective with those of individual freedoms to raise up society.

Bottom line

There is a phrase, “Pull yourself up by your bootstraps,” which means that as an individual, you can lift yourself on your own, which is absurd. No one can accomplish anything independently, especially in a business setting. Americans vastly overstate their upward socioeconomic mobility because it’s part of the American Dream, the great promise of the invisible hand theory, and the Friedman Doctrine—that if the market goes unchecked, you can reap all the gains, just like capital. Aggrandizing shareholders, personhood and self-interest leading to better outcomes do not account for the extreme cooperation required for an economy to work. Therefore, the Neoclassical concepts are only plausible by ignoring relationships and social interactions between parties.

We cannot look down on people because of their station in life, and no one can choose their parents. We all, it appears to me, learn to navigate the world from our family and peers. We should treat people equally and balance personhood with collectivehood, which I believe should be a goal of diversity, equity and inclusion. We should understand that the collective prospers when all its members prosper, so shared gains should start at the bottom and work their way up. I’ll call this “trickle-up” economics. As the economic model includes interactions and social relationships, it is much easier to see how we all impact one another, prosperity can compound and make us all better off.

The economics theories chosen for this article frame the problems to be solved and their potential solutions. In my opinion, actuaries would benefit from understanding the narrative that their models promote because the narrative frames the solution. We should want technology to make us all more productive so we all prosper and share in the gains. Therefore, we should not look to AI to lower costs and replace workers but to machine usefulness to help us flourish and create better futures for all.

Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.

Copyright © 2025 by the Society of Actuaries, Chicago, Illinois.