
Mental models: changing the perspective
Seventeen mental models, how to combine them, and a simple decision framework for clearer thinking.
I’ve been getting more interested in mental models lately.
Not because I want to become someone who has an answer for everything.
Actually, quite the opposite.
The more I learn about them, the more I notice how often I don’t really understand something as well as I thought I did.
A mental model, for me, is simply a way of looking at a problem.
That’s it. A lens.
And sometimes changing the lens changes the problem completely.
What I find interesting is that the same blind spots tend to show up everywhere — in work decisions, in personal ones, in how I read a situation or how confident I feel about something I probably shouldn’t be that confident about. Not because I’m careless, but because thinking quickly is efficient, and efficiency comes from habit. The problem is that habits carry assumptions, and assumptions are often invisible until something goes wrong.
That’s where I’ve found these models useful. Not as a system. Not as a checklist. More like a set of questions I can reach for when something feels off, or when I want to slow down a decision that deserves more thought than I’m giving it.
In work, they’ve helped me see around corners I didn’t know were there — understanding what people actually want, not just what they said they want; catching failure modes before committing to a direction; recognising when I’m inside my circle of competence and when I’ve quietly drifted outside it.
In everyday life, the value is different but similar. They create a little distance between me and my own instincts. Enough space to ask whether I’m thinking clearly, or just rationalising something I’ve already decided.
I don’t think this ever becomes automatic. But having a small set of questions close at hand makes it more likely I’ll actually pause and ask one.
Here are seventeen I keep coming back to.
1. Inversion — What would make this fail?
Instead of only asking how to succeed, ask what would guarantee failure. Avoiding obvious failure modes is often easier than designing a perfect plan.
Example: Before launching a new team process, ask “what would make this process ignored by everyone?” You’ll find the real blockers faster.
2. Incentives — What does each person actually have an incentive to do?
Behaviour follows what is rewarded, not what a policy says it wants. Before blaming people, inspect what the environment is actually rewarding.
Example: If a team metric rewards speed but not quality, don’t be surprised when quality slips.
3. Hanlon’s Razor — Is this actually malice, or could it just be a mistake?
Before assuming bad intent, consider confusion, missing context, or ordinary error. Most frustrating outcomes have a much more ordinary explanation than the one that first comes to mind.
Example: A colleague gives confusing feedback on your work. It’s more likely they were rushed or unclear than that they were trying to undermine you.
4. Opportunity cost — What am I giving up by choosing this?
Every choice is also a rejection. The real cost of a decision includes what you could have done instead.
Example: Saying yes to a two-hour meeting costs two hours of focused work, not just two hours of time.
5. Circle of competence — What do I actually understand well?
The goal isn’t to have a large circle. It’s to know where the boundary is, and to be honest about what lies outside it.
Example: Confident advice in an area you don’t understand well is more dangerous than saying “I’m not sure.”
6. Thought experiments — What changes if I alter one assumption?
Imagine a situation under different conditions to test an idea without having to act on it. Changing one assumption at a time can expose what is actually driving the outcome.
Example: If cost were not a constraint, would you still choose the same option? If not, cost may be carrying more weight in the decision than you realised.
7. First Principles — What do I actually know to be true here?
Break a problem down to its foundations — facts you can verify — and reason upward from there. This is harder than it sounds, because a lot of what feels like knowledge is actually inherited assumption or convention.
Example: Instead of asking “how do others structure this?” ask “what problem are we actually trying to solve, and what do we know for certain about it?” The answer is often simpler and different.
8. Occam’s Razor — Is there a simpler explanation that fits just as well?
When several explanations fit the evidence, start with the one requiring the fewest unsupported assumptions. Simple doesn’t always mean correct, but unnecessary complexity creates more places to be wrong.
Example: A process is slowing down. Before assuming a systemic failure, check whether a single bottleneck explains everything. It usually does.
9. Margin of safety — What if my assumptions are wrong?
Estimates are wrong and surprises happen. Building in extra time, budget, or redundancy creates room for error without turning every mistake into a crisis.
Example: If you think a task takes a week, plan for two. You’re probably underestimating something.
10. Compounding — What small advantage can grow enormously over time?
Small, consistent improvements accumulate faster than most people expect. The flip side is also true: small, consistent neglect accumulates too.
Example: Improving by 1% each week for a year produces something much better than starting from scratch every few months.
11. Base rates — What usually happens in situations like this?
Before treating a situation as unique, ask what normally happens. The specific story matters, but it should update the general picture, not replace it.
Example: Most software projects take longer than estimated. Your project is probably not the exception.
12. Probabilistic thinking — How confident should I really be?
Most outcomes aren’t simply possible or impossible. Thinking in ranges and likelihoods makes uncertainty visible instead of hiding behind confident language.
Example: “This will probably work” is more honest and more useful than “this will definitely work.”
13. Second-order effects — And then what happens?
Ask what happens after the obvious consequence. A shortcut might save an hour today and create extra work every week. An incentive might improve one number while quietly damaging the system around it.
Example: Cutting corners on documentation speeds things up now. The next person working on it — possibly you — pays the cost later.
14. Confirmation bias — What evidence would prove me wrong?
We naturally notice evidence that supports what we already believe. Asking what would change your mind forces you to look for information that might actually do so.
Example: If you’re only reading reviews that confirm your decision, you’re not researching — you’re reassuring yourself.
15. Social proof — Am I believing this because everyone else does?
Other people’s behaviour is useful evidence when you lack information. But following the crowd can also amplify a shared mistake.
Example: “Everyone’s using this tool” is a reason to investigate it, not a reason to adopt it.
16. Loss aversion — Am I avoiding a loss even though the alternative is better?
Losing something tends to hurt more than gaining the same thing feels good. This can make you hold on to the status quo long after it stops serving you.
Example: Staying with a familiar tool that’s slowing you down because switching feels like giving something up.
17. Pale Blue Dot — How does this look from a wider perspective?
When fear or sadness fills the whole frame, zooming out can create some distance without denying what you feel. Carl Sagan’s view of Earth as a tiny point in a vast universe is a reminder that our worries are real, but they are not the whole picture.
Example: When a setback feels overwhelming, pause and watch Carl Sagan’s Pale Blue Dot. Then ask whether this moment will carry the same weight in a year, or from a wider view of your life.
The deeper idea: models work together
This is the part of Charlie Munger’s philosophy I find most valuable.
He talked about building a “latticework of mental models” — not a list of definitions you memorise, but a set of lenses you’ve internalised well enough that they start to interact. The goal is to see a situation from multiple angles at once, and to notice when different models are pointing in the same direction, or quietly pulling against each other.
Don’t treat them as seventeen separate definitions. Build clusters.
For career decisions: Opportunity cost + Compounding + Incentives + Circle of competence + Second-order effects
The question underneath all five: what am I trading, what grows over time, what does the environment actually reward, am I qualified to judge this, and what happens after the obvious outcome?
For learning something new: First Principles + Compounding + Inversion + Circle of competence + Probabilistic thinking
Strip away what you’re assuming. Find what actually builds over time. Ask what would stop you from improving. Be honest about where your boundary is. And stay calibrated about how much you actually know versus how much you think you know.
For relationships and conflict: Incentives + Hanlon’s Razor + Social proof + Loss aversion + Second-order effects
What does this person actually want? Is this malice, or just a mistake? Am I reacting to them or to what others think of them? Am I holding onto a dynamic because changing it feels like a loss? And what happens to the relationship after the obvious next move?
For financial and investment decisions: Circle of competence + Base rates + Margin of safety + Compounding + Probabilistic thinking
Stay inside what you understand. Anchor on what usually happens. Build in room to be wrong. Think about what grows quietly over time. And be honest about the range of outcomes, not just the one you’re hoping for.
For avoiding bad decisions: Inversion + Confirmation bias + Loss aversion + Social proof + Occam’s Razor
How could this go wrong? Am I only looking for evidence that agrees with me? Am I avoiding a change because it feels like giving something up? Am I following the crowd? Is there a simpler explanation I’m ignoring because the complicated one feels more satisfying?
The clusters aren’t fixed. What I find more useful is noticing which question is missing from how I’m currently thinking about something. Sometimes I’ve mapped the risk carefully but haven’t asked what I’m actually giving up. Sometimes I’ve thought through incentives but haven’t considered what the second-order consequences look like.
The models become more useful the more they stop feeling like a list and start feeling like questions you reach for without thinking.
Mental models don’t give you the answer. They’re questions, not conclusions.
The value isn’t collecting them. It’s having a few useful prompts you can reach for when you’re stuck — to notice the question you weren’t asking.
Further reading: Super Thinking by Gabriel Weinberg and Lauren McCann, and The Great Mental Models by Shane Parrish.