Mental Models: Upgrade your thinking

May 31, 2020 · Better Living Thinking

It might seem obvious that making good decisions is important, yet we spend very little time evaluating the quality of our past decisions and working on upgrading the frameworks to make better ones. Improving this toolkit massively increases your leverage on future returns. If you can increase the probability of making the right decision from a modest 60% to 70%, that figure rapidly compounds over years of decisions and provides over a thousand-fold worth of outsized returns in the long run.

Why mental models?

The decisions we make are based on our understanding and mental models we have of the world. They are a combination of heuristics, concepts, and shorthand abstractions we’ve internalized over the course of life. We view reality and make decisions through the prism of our mental models.

Imagine having a cognitive swiss army knife of relevant concepts to deploy whenever you’re faced with a tough problem. If you want to upgrade the quality of your thinking and make better decisions, you need a strong framework of models to actively utilize.

What exactly is a mental model?

Mental models are tools for making decisions that maximize the probability of the best outcome.

I’ve picked these mental models from a large range of disciplines, spanning from cognitive psychology, mathematics, to complex systems theory. This is by no means an exhaustive list, just ones that I’ve come across over the years that I find useful.

How to understand this list

I haven’t counted the exact amount, but I would guess that there are over 100 models listed here (which will likely continue to grow).

Here is a quick tour of the categories:
1) General thinking principles
2) Cognitive biases
3) Statistics
4) Decision making (choice architecture, processes, outcomes)
5) Game theory
6) Systems thinking
7) Evolutionary psychology
8) Physics
9) Biology
10) Neuroscience
11) Computational models (algorithms and optimization)
12) Economics and finance
13) Business and marketing

Okay firstly, let’s acknowledge that this is an insanely long and overwhelming list. It’s going to take you a long time to even go through each concept, much less retain and utilize them. Either way, it’s a starting point to deep dive into some of these concepts.

1. General thinking principles

Here are 10 key mental models, some of them are singular and straightforward, and some of them are more complex, representing a collection of models or an overall approach. To me, they act as fundamental lenses to interpret the world, surfacing different facets of the same, richly layered and interconnected reality.

1. Bayesian reasoning and thinking in probabilities: How do we know what is true? Our ability to discern what is real should form the very bedrock for all our beliefs. Statistical reasoning lies at the core of epistemology.

Every belief you have should be held probabilistically. It means that you can never be 100% certain of anything. All you should possess is the likelihood that your belief in something is true. Without going into the math, there are two main components: what the prior probability is of whether something is true (the prior), and the likelihood of getting some information given that that something is true (likelihood). Baye’s theorem tells us the probability that is true is the prior x the likelihood.

In other words, when you’re trying to figure out whether something is true, you need to focus not just on the new information you’ve received, but your prior probability that something would happen.

Update your opinions based on the latest evidence. Changing your position represents intellectual honesty. Consistency is just ideological dogmatism.

2. Signal vs Noise: The environment overloads us with information, but almost all of it is irrelevant. Our brains are marvelously adapted to be pattern-seeking machines. We never perceive reality in its entirety - we only create approximations. This is a necessary biological constraint given that the enormous sensory input far outstrips our processing ability. But sometimes we do it too well. We start to see patterns in the noise when nothing is there (think horoscopes, superstitions, super-natural coincidences). The world is generating more data every day, but with very little information. Learning how to distinguish between relevant information (signal) versus irrelevant (noise) is essential. Applying good Bayesian thinking can help, but so can reducing the noise you take in.

3. Thinking from first principles: Deconstruct a difficult or complex issue by breaking the problem down into it’s most basic assumptions. Elon Musk famously had this approach to SpaceX. He asked the question, why is it so freaking expensive to send rockets to space? He then decided to chase down the sequence of why’s and reverse engineer every component down to its costs. The hard part was then figuring out a way to build everything up from those core components, but at least he was free from orthodoxy and bias. When you face a problem or even an argument as presented by someone, identify the key premises that a person is making. Deconstruct the premises and ask why each one is true. Go as deep as you can down the rabbit hole until you reach axiomatic bedrock, then try to reframe the entire proposition.

4. Simplicity
Occams razor: The simplest solution is usually the correct one. Pretty self-explanatory, but is true surprisingly often. (Technically speaking it’s phrased as “Entities should not be multiplied unnecessarily”, but mine is easier to remember)
Hanlon’s razor: Never attribute to malice that which is adequately explained by stupidity. Basically saying not everyone is out to get you, so chill out. If someone does something that affects you negatively, it’s highly likely that they are just incompetent. There’s no need to immediately assume some vindictive plot against you.

5. Understanding distributions, power laws, statistics: The nature of all events occur in distributions (height, wealth, people, etc). It’s important to understand what is the underlying distribution structure (are events independent or linked?). Normal distributions and concepts like mean and standard deviation help to frame most of nature when events are independent and uncorrelated. Power law distributions help to understand remaining phenomena where distributions of events are interdependent (sales, networks).
6. Decision-making framework: Think of decisions as placing bets. What is the underlying probability distribution of outcomes? What are the risks, payoff structure, and resulting expected value? The main concept is to measure your expected utility (or value) of the decision: Payoff (sum of probability payoff x amount) - costs - risks (probability of downside x amount). Consider nth order effects > is payoff exponential and uncapped, is risk capped? For example, investing in startup = low probability of payoff x enormous payoff amount - fixed risk (seed investment). Even if the expected utility is positive, is this a local vs global optimum?

7. Game theory: Once you are in a scenario that involves cooperation or conflict, basically any situation with multiple agents and incentives, game theory dominates. It’s important to understand what the different incentives are and what the Nash equilibrium is (the state where if everyone acted with their self-interest and cannot further move to get an advantage) because it allows you to predict default states. Does it create a system with a multipolar trap? See more here: Game theory.

8. Complex systems: When you have many units (people, ants, neurons, nodes, companies), interacting with non-linear dynamics to produce complex emergent behavior. Results in unintended consequences, chaotic systems (sensitivity to initial conditions making prediction impossible), emergent phenomena (properties that cannot be reduced to its constituent parts). See more here: Complex systems.

9. Computational complexity and optimization: Every process faces the same fundamental constraints of time and space. Some problems or processes can take exponential or polynomial time, learning how to reduce complexity to linear processing time is one key problem. Because of the constraints, if the problem cannot be reformulated and simplified, there will be necessary trade-offs. Common trade-offs are, explore/exploit (Given limited time, how much resource is spent exploring versus exploiting to get the optimal outcome), search/sort, memory caching. Whether you realize it or not, everything is computation, so the constraints and solutions can be approximated universally.

10. Evolutionary psychology: This is the binding principle behind all of life and human behavior. We are all survival-replication machines shaped by evolutionary pressures, so every desire, inclination, fear, motivation, and problem solving ability can be explained from an evolutionary lens. Sexual selection can account for a tremendous amount of human behaviour (costly signaling, vanity, desire for status/wealth etc). See more here: Evolutionary psychology

2. Cognitive biases

Most of us would consider ourselves rational. What you see is what you get, right? We think of our eyes as cameras that peer outward, capturing reality in its full glory, and that our brains render this data into ideas with precision and perfect fidelity. The reality is that most of what we perceive and believe is an active fabrication process that is governed by top-down processes. We possess virtual models of the outside world within our brains. Whether it’s looking at an angry face or staring into an open vista, most of this is internally generated. The incoming sensory information merely augments the model that exists.

The result is that our minds are utterly subject to a whole host of cognitive biases. There are too many to list here, but I’ll cover the ones I encounter most often in my life.

1) Anchoring effect
When faced with uncertainty, our brains grasp at straws, looking for a reference point to base our decisions. Because of this, the first piece of information we receive tends to subconsciously bias all our subsequent perceptions and decisions. This effect is so strong that you can be told completely unrelated information, but your brain goes into pattern recognition overdrive, using that information to impact later judgments. The best part? You have zero awareness of this even happening.

In the famous Daniel Kahneman study asking people to estimate how many African countries were part of the UN, they first gave people a wheel to spin from 0-100, but had it rigged to land on either 10 or 65. When they were later asked to guess the number of countries, those that landed on 10 guessed 25%, while those that landed on 65 said 45% on average. It’s important to emphasize that the wheel spinning obviously had zero logical connection to the answer, yet it still had a subconscious effect.

2) Framing: Loss aversion
Your mind does not accept facts in a blank state. The presentation of the facts and existing beliefs will affect your perception of new information.

Here’s a quick example
Scenario 1: A treatment for a fatal disease is known to work, but it ends up killing 5% of the people who receive it.
Scenario 2: A treatment is known to save up to 95% of people who receive it.

Should we be administering this treatment to people? These statements declare the exact same thing, but people are naturally averse to losses. More people will likely approve the treatment in the second scenario rather than the first even though they are mathematically equivalent.

3) Availability bias: You have the tendency to pick facts and filter opinions based on what is most recently available in your mind, whether it corresponds to reality or not.

4) Fundamental attribution error: We tend to explain other people’s behavior on internal factors (poor personality etc) but not external (situational reasons like someone having a bad day). We do this in reverse for ourselves.

5) Cognitive dissonance: If there is a gap between your belief about something and your behavior or another belief, it creates a tremendous discomfort psychologically. There will be an urge to relieve this resulting tension. The tension can be so strong that the reconciliation happens almost automatically or unconsciously.

Example: You can care a lot about your health and you know smoking is unhealthy, yet you lack the discipline to stop. The urge to then reconcile these facts can lead you to irrational beliefs that the statistics behind the dangers of smoking are exaggerated, or that you’re inherently lucky and immune yourself.

6) Affective forecasting: It’s hard to know events will truly impact your emotional state in the future. People are generally horrible in forecasting the impact of events in their lives.

7) Motivated reasoning: We are inherently biased; we look for information that already confirms our existing beliefs. This is especially true if the belief is tied to our identity. Confronting new information forces us to not only readjust our beliefs, but also creates immense tension from cognitive dissonance and attacks our identity. For example, if you sell some health food and find out one day that the evidence for the active ingredients is weak, you will (usually subconsciously) choose to ignore it and find supporting evidence to the contrary. This happens much more often than people realize.

8) Self-esteem and identity preservation: We have a powerful drive to protect our identity and self-esteem. Any information we receive that threatens our self-esteem is usually rationalized away.

9) Dunning-Kruger effect: In order to understand the limits of your own intellectual ability, you need to have the self-awareness or metacognitive ability to evaluate it. Basically, some people can be too incompetent to realize their own incompetence.

3. Statistics

Thinking probabilistically, not succumbing to narratives. Our brains evolved to solve problems that are immediate to our environment. Numbers and statistics are abstractions of reality; we are not well equipped to understand them intuitively without explicit training.

Statistical significance: Understanding the basics of hypothesis testing and A/B testing. Type 1/2 errors: How do we know what is true? Figuring out whether an intervention has made a real impact (a function of change in mean + standard deviation).

Base rates: If I tell you that Steven is an American who happens to be an avid reader and loves organizing, would you say he is more likely to be a farmer or librarian? If you guessed librarian, it’s because you’ve ignored the base rate. There are 100 times more farmers than librarians, so even with the new piece of information, it is still far more likely that he is a farmer. Never be skewed by new information and always consider the base rate probabilites. See more here.

Normal distributions and standard deviations: Most natural phenomena that have independent elements are normally distributed. Human heights, blood pressures, IQ scores. One person’s height doesn’t affect another. This leads to a central (mean) number, followed by decreasing probability towards the sides.

- Power laws, zipf’s law
- Law of large numbers
- Correlation vs causation

4. Decision making

Deconstructing problems: Always break down a big problem into smaller problems that are easier to solve. Do not add small problems into a big problem that becomes impossible to solve.

80/20 principle: 80% of the results come from 20% of the causes. Identify the high leverage actions where results are coming from. Consider if your choice of action fits this rule.

Opportunity costs: Every decision comes at a cost of not doing something else. There is no free lunch. Making a choice adds a cost of the next best alternative.

Activation energy: Term borrowed from physics. There’s a barrier of effort required to do something. If the effort is too high, it becomes unlikely that behavior will happen.

Forcing functions: Setting up steps to make something happen. Making everyone stand up during meetings to keep meetings short. Booking an early gym class to make yourself wake up early. Sometimes it’s easier to control the environment than rely on willpower. Also known as commitment devices.

Understanding risk and options: Each decision doesn’t just come with cost and benefits, they also come with a set of risks. Minimize the total risks and buy long options for safety. Always create margin of safety because no forecast is ever accurate.

Decision processes

Expected utility, or thinking in bets: Every decision is a bet on your time, effort, or money, with an expected payoff. Whether it’s deciding to change your job, making a financial investment, or even choosing where to eat, you are making a bet about your expected utility in the future. For each decision, you need to consider the costs (and risks) and the expected value of each outcome. The expected value is the confidence of attaining the outcome times the actual payoff.

Global vs local optimization: Imagine a situation where you are trying to find the tallest part of a hilly landscape. The problem is that you can only see 1 meter ahead of you. The natural instinct is to continuously move toward a higher direction until you hit a peak. The problem is if you started off near the base of a smaller hill, you might have found a “local” peak, but there might be a higher peak in the “global” landscape that you’ve missed by following the constrained optimization pattern. Sometimes a decision might seem great within a set of constraints, but you might be missing a larger, global optimum because it requires you to reset your perspective by climbing downhill.

Over-optimization vs redundancy

Outcomes of decisions
- Non-linearity > nth order effects

Law of unintended consequences: There’s a famous story about how Australia faced a massive rat problem in the early 1900s. In order to combat this, they decided to award people for catching rats and bringing them in to receive their reward. What they didn’t predict was that people began breeding rats to claim the reward. People operate by incentives, but sometimes it’s difficult to consider all the unintended consequences when incentives are easily distorted.
- Distorted incentives
- sunk costs

5. Game theory

When there’s more than one person (or agent), the dynamics of cooperation and conflict lead to strange outcomes. What happens when you have multiple agents with different incentives interacting? Often times, it leads to suboptimal outcomes when agents are motivated by their proximate goals. Compounded by the nature of multipolar regimes, preference falsification, and coordination problems, it can lead to large-scale self-sustaining systems with poorer outcomes without any central cause or agent.

Prisoners dilemma

Coordination problem/ multipolar regimes: Dictatorless dystopias, red queen effect, capitalism. See here

6. Complex systems

Everything interesting in life is mostly a result of complex systems. From ants to consciousness, immune systems to social networks, the principles of complex systems are unavoidable. What constitutes a complex system? Simple units, locally interacting with other units through non-linear dynamics, resulting in emergent properties and behaviors that cannot be reduced to its individual parts.

7. Evolutionary psychology

We were designed through natural adaptive pressures that led us to develop specific reward functions and problem-solving modules that let to higher fitness functions.

8. Physics

- Path of least resistance: The formula that describes the behavior of all particles and forces in the universe.
- Entropy: Things naturally break down towards disorder. The inescapable law of the universe. Applies to systems and things in general - without a counteracting force, things naturally degrade.
- Activation energy

9. Biology

10. Neuroscience

1. The brain is a team of rivals
2. Most of your thinking happens unconsciously
2. Bayesian thinking machine

11. Computational models and general optimization

- Optimal stopping rule
- Search/sort tradeoff
- Memory caching
- Explore/exploit tradeoff

12. Economics/ Finance

- Law of diminishing returns
- Compounding effects
- Homo economicus

13. Business, Marketing, Startups

Product market fit