What is optimization?

Optimization has appeared as a key topic of discussion across this blog for good reason. It’s really difficult to talk about the behavior of systems and the ways they can go awry without invoking or at least alluding to optimization.

Defining optimization is challenging due to its abstract nature and varied applications in fields like mathematics, engineering, and science. Furthermore, in each domain, there are nuances that distinguish each of these uses of the term from one another. However, in this post I’m going to do my best to provide an ontology or “basic shape” of optimization. Afterward, we’ll talk about how this is relevant for building our conception of Misaligned Markets.

Defining optimization

In the broadest possible sense, optimization occurs whenever an aspect of the world is modified in order to better suit some set of criteria. In this way, many processes and technologies within human society can be thought of as optimization processes. This could include:

• The development of irrigation systems to increase agricultural yields.
• The deployment of vaccines and/or antibiotics to reduce the occurrence of disease.
• Introducing exogenous insulin as a treatment to lower blood sugar levels in a diabetic patient. Endogenous insulin functions this way within non-diabetic individuals.

Optimization processes can be conceptualized as systems that consist of an optimization “target,” as well as any criteria by which the target is judged against.

Using some of the above illustrations as examples:

• Irrigation systems
  • Target: Increased crop yield per unit area.
  • Criteria: Water-use efficiency or maximizing the amount of crop yield per unit of water used.
• Vaccines
  • Target: Minimize disease incidence and/or disease transmission within a population.
  • Criteria: Vaccine coverage across a population and/or vaccine efficacy against a specific disease.

Optimization targets may be deliberately set by a planner, but natural systems can often revolve around implicit targets without intentional design. For example, many systems in the human body exist to maintain homeostasis or an optimal operating condition.

Distinguishing between “Simple” and “Complex”

When we talk about optimization processes, we can generally distinguish between optimization that occurs via “simple” or “complex” systems. Complex systems, like markets, have more components than just an optimization target and optimization criteria, but the language of targets and criteria give us a basic starting point to talk about many common types of optimization processes.

Simple systems tend to be “closed-loop” systems. This means that they rely on negative feedback to achieve a desired target. Negative feedback involves feeding the current output of a system back into itself, which is used to reduce the difference between the desired target and the actual state of the system. Once the optimization target is reached, feedback is applied so that the system maintains a steady state.

The behavior of a thermostat provides a good example of these principles in action. At a high level, thermostats:

1. Take in a user’s desired temperature as an input, which might be represented as a voltage in digital thermostats or the expansion of material in mechanical ones.
2. This input is checked against the ambient temperature, thus incorporating the change in temperature into the system. The ambient temperature serves as feedback to inform the temperature controller whether it should apply heat or cold to reach the desired temperature.
3. Once the thermostat reaches the desired temperature, heat or cold is applied as needed to maintain it, ensuring the system remains at steady state indefinitely.

These types of systems tend to be studied in fields like control theory, which is relevant to physics and engineering.

Complex systems, on the other hand, often involve multiple interacting components and may exhibit emergent behaviors, making their optimization processes less predictable. Given the dynamic nature of complex systems, they’re much harder to generalize. Typically, though complex systems tend to exhibit some of the following behaviors:

1. Complex systems are dynamic and adaptive, meaning they respond not only to initial inputs but also to changes in any constituent components or the surrounding environment.

• This is because they consist of interdependent components whose behaviors may be subject to feedback loops and can be modeled independently yet interact in ways that create effects across multiple system layers. This interactivity regularly leads to multiscale structures, where not just individual components but relationships between different layers within the system are critical.

2. Complex systems, and the individual components of these systems, can shape and be shaped by their environments in a recursive feedback loop.

• This interaction between a complex system and its environment can result in unpredictable or emergent behaviors, producing outcomes that cannot be easily understood by examining individual components alone, making the system "greater than the sum of its parts."

3. Unlike with simple systems, iteration in complex systems does not necessarily produce a steady state or stable equilibrium; instead, these systems typically remain in flux or evolve continuously over time.

• It’s important to note that the evolution of a complex system is highly path-dependent, meaning its current state and behavior are influenced by its prior states, adding to its unpredictability and uniqueness.

These types of systems are studied in fields like systems dynamics or systems theory, as well as cybernetics.

It’s really critical to note that while we’ve distinguished between “simple” and “complex” systems, measuring system complexity can be hard because it’s not binary. A system can grow to be more complex over time, for example. Similarly, some types of systems may sit between these different descriptions. It’s best to think of system complexity as existing on a spectrum, going from least complex to most complex.

Some practitioners across fields like management, programming and engineering subscribe to a sliding scale defined by: simple vs complicated vs complex vs chaotic systems.

The human body as an example of a complex system

When we think of complex systems that diverge from straightforward, steady-state optimization targets, like thermostats, the human body stands out as a sophisticated example. Biological organisms, including humans, aim to achieve balance or homeostasis within dynamic and ever-changing internal and external environments.

The Human Body as a Dynamic and Adaptive System

• The human body is constantly adjusting to internal and external changes, making it a highly dynamic and adaptive system. For example, in response to environmental changes like temperature shifts, the body adjusts blood flow, shivers, or sweats to maintain a stable internal temperature. This adaptability relies on the body's multiscale structure, with complex feedback loops connecting systems from cellular signaling to organ systems.

The Human Body as a System in Constant Interaction with Its Environment

• Human physiology doesn’t operate in isolation; it’s continually interacting with and adapting to the surrounding environment. This interdependence creates a recursive feedback loop, where both the body and its environment shape each other. The immune system illustrates this recursive interaction. It constantly monitors the body’s internal environment and responds to foreign pathogens. When it encounters a pathogen, the immune system not only fights the infection but also “learns” and adapts by producing specific antibodies, changing its future response to that pathogen. This creates an environment that is less hospitable to similar pathogens in the future, shaping the overall system’s health. Such adaptations can result in emergent behaviors like immunity or resistance, which are outcomes that can’t be easily predicted by analyzing any single component of the system.

Iteration in the Human Body Often Avoids Steady State and Remains in Flux

• While the human body does aim for a steady state, it rarely maintains a true steady state because it’s continuously adapting to internal and external stimuli. This ongoing process of adaptation and evolution is highly path-dependent, meaning that past states influence how the body responds to new changes. The cardiovascular system, for example, is sensitive to life-long factors like stress, exercise, diet, and age. When exercising, the heart rate increases to supply more oxygen to muscles, and over time, regular exercise induces long-term changes, like improved heart efficiency and vascular health. These changes accumulate, meaning that an individual’s cardiovascular response to stimuli today is shaped by their history, diet, and lifestyle.

Optimization and systems that revolve around optimization targets are a complex topic (no pun intended). This post serves as a cursory introduction to the subject. In future posts, we’ll analyze markets through a complex systems framework, allowing us to better understand the optimization processes and failures occurring in modern markets. We’ll also look at other ontological questions surrounding the nature of optimization.

Summary

• In the broadest possible sense, optimization occurs whenever some aspect of the world is modified in order to better suit some set of criteria. In this way many processes and technologies within human society can be thought of as optimization processes.
  • Optimization processes can be conceptualized as systems that consist of an optimization “target” as well as any criteria by which the target is judged against.
  • Optimization targets may be deliberately set by a planner, but natural systems can often revolve around implicit targets without intentional design.
• When we talk about optimization processes, we can generally distinguish between optimization that occurs via “simple” or “complex” systems.
  • Simple systems tend to be “closed-loop” systems. This means that they rely on negative feedback to achieve a desired target.
    • Negative feedback involves feeding the current output of a system back into itself, which is used to reduce the difference between the desired target and the actual state of the system.
    • Once the optimization target is reached, feedback is applied so that the system maintains a steady-state.
  • Complex systems, like markets, often involve multiple interacting components and may exhibit emergent behaviors, making their optimization processes less predictable. Given the dynamic nature of complex systems, they’re much harder to generalize.
    • Complex systems are dynamic and adaptive, meaning they respond not only to initial inputs but also to changes in any constituent components or the surrounding environment.
    • Complex systems, and the individual components of these systems, can shape and be shaped by their environments in a recursive feedback loop.
    • Unlike with simple systems, iteration in complex systems does not necessarily produce a steady state or stable equilibrium; instead, these systems typically remain in flux or evolve continuously over time.