Decision Theory Concepts Useful in Life

Explore and exploit

Exploration is trying out different things to see what works for you. For example, you may try different notes apps like Google Keep, Notion, Apple Notes and Microsoft OneNote. Exploitation is when you’ve found one and you want to make the most of it, say learning all the features of Notion.

Both are necessary: without exploitation, you’re less productive and less skilled. For example, if you’ve been coding in Python for 3 years but using the same features with the same coding style as 3 years ago, you’ve stagnated and your career opportunities will be limited. Without exploration, you’re stuck using the best option from several years ago. This is why there’s a stereotype of people in their 40s and 50s as dinosaurs. They’ve stopped exploring and are using the best tools and technology as of 2006. In fact, I ran into one person who uses the clunky OneNote and hasn’t even heard of Notion.

Ask yourself the following questions:

1. Have I explored in the past 6 months?

2. Have I exploited in the past 6 months?

If the answer to either is no, work on it.

Sometimes people respond to this with, “No, I’m too busy” but you might be too busy working in an inefficient way or using the wrong tool for the job and exploring may be the thing that actually helps.

To hold myself accountable to follow my advice, in the past 6 months I explored the Raspberry Pi, and exploited powerful features in Google Docs.

Eliminate irrelevant factors

Often, problems come to us with relevant and irrelevant factors all mixed together. Sometimes, it helps to identify these, remove the irrelevant factors, and then solve the simplified problem.

Say someone asks you a question about investing: they’re investing ₹45,000 every month. What kind of returns should they expect from an FD vs an equity mutual fund?

The percentage return doesn’t change depending on the amount invested, whether it’s ₹1, ₹1K or ₹1 lakh every month. So I do the analysis by assuming an investment of ₹1 every month. This way, the calculation becomes simple, and I can even do it mentally: If you invested ₹1 every month, how much have you invested over the past year? And if you invested ₹45,000 every month, how much have you invested over the past year? In the first case, you can do the math mentally and get more clarity, while in the second case, you can get bogged down in the irrelevant.

Simplify the problem and then solve the simplified problem.

Eliminate relevant factors

Sometimes, when you examine a decision and break it down into its constituent factors, you may find that all of them are relevant, so the aforementioned technique of eliminating irrelevant factors doesn’t work. What do you do then? Remove one of the relevant factors, solve the simplified problem, and then add back the removed factor and asked what you’d change. In other words, make the decision in 2 steps, rather than a single complex step that bogs you down.

For example, when I bought my iPhone 15 last year, I wondered if I should buy the 14 instead to save money. I thought I should buy the 14, but wasn’t able to convince myself. A part of me kept wondering whether I was being swayed too much by the lower price of the 14. To get clarity, I asked myself, If both phones were priced the same, which one would you buy? The answer was the 15. That gave me the clarity to reject the 14.

Another question arose: should I buy the 15 Pro? I asked myself, If both the 15 and 15 Pro were priced the same, how valuable is each phone to you? The answer was that both are equally valuable, since the 15 Pro had nothing I cared about that the 15 didn’t. Then I added back the factor I’d removed (price), and asked myself, Given two phones that are equally good, would you buy the cheaper or the costlier one? The answer was obvious, so I bought the 15. This two-step decision-making process gave me clarity.

Flip the status quo

I was deciding whether to get my broken oven repaired or buy a new one. I flipped the status quo, asking myself, If I didn’t have this oven, would I buy it, or a different one? The answer was a different one, because the existing one wasn’t right for me, so I gave it away and bought the one I wanted.

This technique is important because of the status quo bias, where we tend to accept the flaws of the status quo while holding a novel option to an impossible standard. To overcome this bias, flip the status quo. Sometimes, telling yourself, “I must not suffer from this bias!” works less well than accepting the bias and solving the problem tangentially.

Amortised cost

A startup was working with bad code that was slowing them down, so they were paying more in lost productivity than it takes to fix it. So we decided to start improving it. The first task was a feature estimated to take a week. I asked the engineer to fix whatever code he touches while he’s at it. This fixing would take 2 weeks, and then implementing the actual feature would take only half a week (since the code was better). The founder objected to my decision saying that spending 2 weeks to save 1/2 week isn’t the right tradeoff. What he missed is that once the code is better, it will speed up multiple tasks in the future. You incur the cost once, but the benefits multiple times. This is called amortised cost, and makes many things worthwhile that wouldn’t otherwise be.

Risk aversion

People always prefer a safe choice to a risky one. So, if one has to be persuaded to take a risky choice, one has to be compensated for this risk. What differs from person to person is how much compensation they need. For example, if a risk-free FD pays 6%, and an equity mutual fund declines significantly in value from time to time, it has to pay more to encourage people to pick it. If it paid the same 6%, no rational person would choose it. What differs from person to person is how much the equity mutual fund must pay for it to be viable. I’m fine with 12%, but someone may demand 20%¹.

At some times in our life, our risk aversion can change. When I left Google, I was keen to invest all my money in my startup. Today, I’m not. That’s okay. Just be aware of what’s right for you now, and whether it’s indeed right for you now.

Optionality

A few years back, a friend of mine suggested I specialise in Blockchain / Crypto. I refused saying that I want to keep my options open as a generalist advisor. And looking at what happened with Blockchain, I was proven right. Unless you’re sure, don’t reduce your available options. Think thrice before making any decision today that narrows the range of options tomorrow.

Degrees of freedom

This determines how many levers you have under your control. For example, project management has three degrees of freedom, where you can trade off scope for time, quality or manpower:

In order to make effective decisions, you need to understand how many degrees of freedom you have. Otherwise, you may not make use of all the levers you have under your control. For example, if you’re asked to launch quicker, you should make a counter-proposal that you’ll do so in exchange for lower scope. You can’t do this if you don’t understand the degrees of freedom you have. The opposite problem can also happen where you keep banging your head against the wall even after using all the degrees of freedom you have, rather than realising that you’ve already achieved everything that’s possible and so need to accept what you have.

Downstream and upstream

If you’re by a river and find toxic chemicals in it that you want to eliminate, you need to go upstream to the factory that’s dumping them and fix the problem there. You can’t eliminate the problem downstream.

Similarly, some decisions are downstream of others. For example, founders often complain that development is slow. But this is downstream — the symptom. We can’t fix it. We need to identify what’s upstream by asking ourselves why. After some discussion, I may find that the engineers are being interrupted all the time which, contrary to what non-technical managers believe, reduces productivity. So now we’ve moved upstream. But the story doesn’t end there. When I asked the founder, “Why do you do that?” he replied, “We have a small team, so we need to get a lot of work out of each engineer.” Now we’re two levels upstream. But why do they have a small team? It may be because the engineering budget is low. Now we’re three levels upstream. Why is it low? Because the startup has insufficient funding. Now, four levels upstream, we’ve hit the root cause, also called the leverage point: if you were to create a list of all the changes you could make along with the impact of that make, and sort them in decreasing order of impact, the one at the top is the leverage point.

Asking the founder or yourself “why?” moves you upstream. Each time you ask this question, you’re moving one level upstream. This is why you ask why 5 times to travel all the way up the river.

I don’t take up projects where the founder cuts off the 5 whys discussion and just tells me what to do. Working downstream is a poor use of my time: it produces second-rate results. They may create a dispute and demand their money back. The project is likely to be canceled mid-way. I’ll have negative rather than positive word of mouth. Nor will I have the satisfaction of a job done as well as I can do it.

How much upstream you can go depends on how experienced and mature you are. But whatever you’re capable of, don’t take up any gig that doesn’t let you go as upstream as you’re capable of.

Reference point

When picking between multiple options, you can first pick one as a reference point and then compare the others with that, to simplify the process. For example, if you’re buying a car, you could pick the Wagon R as a reference point, compare other cars to it and ask, “Is this car better than the Wagon R?” That’s an easier question to answer than “Is this car good?”, which is abstract.

Pre-commitment

… is a concept from decision theory where you make a decision ahead of time. That way, when you’re in the heat of the moment, the decision becomes easier because there are fewer possibilities to consider. You’re more confident. Which, in turn, makes you sound more impressive to others. For example, I work as an advisor — I don’t work with business people who try to lecture me how to do my work, since they’re not knowledgeable enough to do so. That’s a pre-commitment, a decision I made ahead of time. Now, when I’m in a call with a prospective client, and they try to tell me how to do my work, I tell them that I either engage as an advisor or I don’t engage at all. There’s no confusion in my mind, so I’m able to quickly resolve things one way or the other.

Subtractive decision-making

In this style of decision-making, you decide not to do something, like not getting a loan to buy a car. Then other things will fall into place. Cars you can’t afford to buy outright are automatically eliminated from consideration, making the decision easier.

You could look at subtractive decision-making as the converse of pre-commitment. Instead of pre-deciding to do something, you’re pre-deciding not to do it.

Sequential decision-making

When you have multiple decisions to make, you make one, and then the other decisions become clear. When I vacationed in New Zealand, I booked the ticket for each city only when I was clear I was done with the previous city. If, on the other hand, I’d pre-booked the entire trip, I may have to remain in a city after I’m done (which is boring), or leave and skip fascinating things to see and experience. Sequential decision-making applies when things happen one by one.

Jeff Bezos’s regret reduction framework

If you have to choose between two options, choose the one you’ll regret less. For example, I was thinking about whether I should charge $175 or 275 per hour. If I quote $275 and the client chooses not to work with me, I’ll regret it. On the other hand, if I quote $175 and the client chooses to work with me, I might have left $100 on the table. In both cases I may regret something, but I’ll regret the latter scenario less. So I went with $175.

Opportunity cost

Say a startup wants all my time for the next year. That will prevent me from being able to take on any more clients. This is the opportunity cost. I should charge enough that I don’t regret the opportunity cost.

On the other hand, if a startup wants only 3 days of my time in a month, then I can be more flexible on pricing, since it doesn’t prevent me from working with other clients.

Bounded loss decisions

A startup approached me to fix their systems, for a fee of $30K. Should I say yes? Achieving the goal may require me to rebuild their entire tech, which can be a hundred thousand dollars worth of work. If I’m charging only $30K, I’ve lost $70K. If it turns out to cost 200K, I’ve lost $130K. In this situation, there’s no limit to how much I can lose. It’s an unbounded loss, which I won’t agree to. The problem here isn’t that there’s risk — there’s a certain amount of risk everywhere, and obsessing about avoiding any risk at all holds you back. It’s okay to take a risk, as long as

1. You know the worst-case scenario and

2. Are okay with it.

Don’t take on unbounded risks.

EV

When a situation can play out in multiple ways, each with a different probability and a different outcome, Expected Value boils this down to one number.

When you’re offered a deal, calculate its expected value. If it’s not positive, you probably shouldn’t take it.

Suppose I offer you a deal: pay me ₹10. Then we’ll roll a die. If 6 comes up, I pay you ₹50. Should you take this deal? To calculate this, consider both scenarios: you win ₹50 - 10 = ₹40, and you win -₹10. The former has a 1/6 chance of occuring and the latter, 5/6. So the EV is ₹40 * 1/6 - 10*5/6 which is -2. Since this isn’t positive, you shouldn’t take the deal I’ve offered you.

Let’s apply this to a real-world scenario: car insurance. There are two types of car insurance:

1. Third-party insurance: If you get into an accident and you’re found to be at fault, your insurance company will pay the other party. This is legally required.

2. Comprehensive insurance also covers damage to your car. This costs more, and isn’t required.

You should prefer third-party insurance over comprehensive. Why? Because insurance has a negative expected EV. In other words, if the insurance payout is ₹10 lakh and the chance of a payout is 1%, then the insurance company will charge you more than 10 lakh x 1% which is 10K. They might charge you ₹13K. The ₹3K goes to covering their overheads and making a profit. Self-insure. Pretend you’re your own insurance company. Pay yourself 13K. Of this, set aside 10K to cover the loss. Then you get to keep the 3K for yourself. So, insurance always has a negative EV. You shouldn’t take it if you can afford to buy another car out of pocket.

When you’re offered a deal, calculate its expected value. If it’s not positive, you probably shouldn’t take it.

Diversification

You should invest in a mutual fund, which invests in multiple companies. That way, if one company, or one sector (like energy), or one country delivers poor returns, you’re protected.

This concept applies to many things in life: as a consultant, I have conversations in progress with multiple leads, increasing the chance one of them converts. This applies to networking too: I know three lawyers, so when the first two were unresponsive, I contacted a third, who took up the gig. A third situation diversification applies is upskilling: I read multiple books and some help and others don’t. Diversification applies broadly in life.

Utility function

I’m looking to buy a car, and I love the Creta. The Kiger is pretty good. The Wagon R is okay, and the Alto is terrible. How do we quantify these instead of using subjective terms like “pretty good”?

We give numerical scores: the Creta gets 10, the Kiger gets 9, the Wagon R gets 6, and the Alto gets 0. Each of scores is called a utility, and quantifies how good a product or service is for you.

A utility function takes a product as an input and returns its utility as the output.

The utility function is non-linear: Notice that I scored the Wagon R 6. A car priced half as much as the Wagon R doesn’t earn half the score — 3 — but 0, since I don’t want it at all. I probably can’t even fit in it! Similarly, considering that the Creta earned a 10, a car that costs twice as much may earn a score of 12 rather than 20, because of diminishing returns. Utility functions quantify these, bringing you clarity.

Avoid sunk cost fallacy

The sunk cost fallacy is when you invest time, money or your reputation in a project, which doesn’t work out. So you invest more, hoping to declare victory, but finally it fails, costing you more time, money and reputation than if you’d given up when it was clear that it’s a failure.

For example, a client refused or delayed my fee for the previous month’s work for no valid reason, and demanded I do another month’s work after which he’ll pay me. He may be lying, and I may end up losing two months’ fee. Not to mention that negotiating with him when two months’ fee is due puts me in a weaker negotiating position than having one month’s fee due. So I told him that unless I receive the fee in two days, I’ll terminate the engagement. I’m willing to take a smaller loss to avoid a bigger loss.

Limited commitment

One good practice for founders starting a startup is to set aside some amount of their funds and when that’s done, if there’s no money available from other sources, to shut down. This is an example of a limited commitment, and is a way of overcoming the sunk cost fallacy (see previous section).

Pareto improvement

… is a change you make to a system that makes one party better off without making anyone else worse.

For example, my advisory contract had terms that let me sue my client if they refused to pay me for work already done. A founder friend pointed out that since I’m not going to sue them anyway, I should remove these terms from the contract: I’m no worse off (since I wasn’t going to sue them). The prospective client is better off (the contract is less threatening to them), so there’s a better chance of them signing the contract.

Work backward from the desired outcome

In chess, you don’t want your king to be mated, which means that the opponent’s pieces shouldn’t come close. This, in turn, means that you need to flank with your king with your other pieces. And you want to forward-deploy yet other pieces for defense in depth. In this process of reasoning, we started with D (avoiding checkmate). Then, we worked backward to C (keep the opponent’s pieces far): C→ D. Then we worked backward to B, ending up with B → C → D. Then A → B → C → D.

You can apply this concept from decision theory in your life. For example, in my advisory, I want a certain minimum monthly income to take care of expenses. I worked backward from that desired outcome and structured my contracts to have that amount as the minimum monthly fee irrespective of the amount of work done or the time spent that month.

Err in the direction of feedback

If one of the two options you have in front of you will give you feedback, prefer that. For example, if you’ve always hired people who have at least 1 year experience, hire one fresher and see how it goes. If it doesn’t work, you’ll get feedback that helps you improve. If you don’t experiment, you may continue with your wrong decision. It’s okay to make a mistake if it’s self-correcting. Not being open to such mistakes results in ossification and is a bigger mistake.

Occasionally try something you disagree with

In my startup Futurecam, I used to review everyone’s code. After a while, I asked the engineers to start reviewing each other’s code. But many commits went unreviewed. So I reminded my team. Things didn’t change after a few weeks, so I reminded them again, explaining why code reviews are important, and asking if they have any objections. They said no objections, but after a few more weeks, there was no change. So I reminded them again in any angry tone. Again, nothing happened. So I decided to let it be, and when things go wrong, point it out to them so that they can see for themselves why code reviews are important. To my surprise, nothing went wrong. So the next time someone insists on doing something different from what you believe in, try it out. We’re all wrong from time to time, and the only way to correct is to occasionally try something we disagree with.

Don’t aim to always be right

As we get more skilled at our work, we’re right more and more often. But the goal shouldn’t be 100% right. That means you’re never trying a new way of doing the same thing. Or new things. Over-conservatism also produces worse results 90% of the time. I also know people who are afraid to answer a question because may be wrong 5% of the time. This deprives others of their insights, and themselves of opportunities, because they appear less insightful. Insisting on always being right is a mistake.

Zero-based decision making

This is how the iPhone’s control center looks:

And this how is how my iPhone’s looks:

I removed all the controls and continued using the phone. When I needed a control, I tried to find a different way to do it. For example, when I needed to adjust the volume, I used the physical volume controls. To use the torch or camera, instead of swiping down from the right to open Control Center, I swipe down from the left to open the lock screen:

… which has the torch and camera controls.

When faced with an inconvenience, I try finding alternatives or living with it. I don’t solve inconveniences immediately — only when they come up again and again, at which point I add the control back to the Control Center. In the process, I explored different forms controls come in. For example, the rotation lock control comes in two forms:

and

I find the latter clearer, since it’s labeled. I don’t have to stop and wonder, “What does this do?”, considering that there are multiple controls that have a glyph within a circle.

At the end of this process, I ended up with a Control Center that doesn’t have controls I don’t use often, and the ones I do use are labeled, so it’s quicker to use.

Zero-based decision-making is sometimes better.

Companies also use this in the form of zero-based budgeting. With traditional budgeting, if department A gets a budget of $2m and B gets $4m this year, when budgeting for next year, these budgets are taken as a baseline. Only if a department wants a higher budget do they need to justify why, not why their present budget should continue. By contrast, in zero-based budgeting, when budgeting for next year, each department starts at $0 and has to justify every expense.

Avoid hyperbolic discounting

If I offered you a choice between ₹10 today or a year later, you’d pick today, because the same money a year later is worth less. Let’s say ₹10 a year later is worth ₹9 today. You’ve applied a discount of ₹1. Since that’s less than ₹10 today, you’d take ₹10 today. That’s a rational decision.

However, many people discount the future too much. If I were to offer a choice between ₹10 today and ₹12 a year later, many people would pick ₹10 today. Actually, they should pick ₹12 a year later, because that amounts to a 20% annual return committed in advance, which is a deal you can’t get anywhere else. If you invest in an equity mutual fund, the long-term returns have been 18% and you don’t know ahead of time what your return will be. If you want a predictable return, you can invest in an FD for 6%. Neither of these match up to the 20% assured return I’m offering.

Another, simpler example of hyperbolic discounting is buying a car on a loan without considering the interest. A rational person would ask if he could buy a cheaper car, a used car, or save up for a few years and then buy without a loan.

Choice architecture

Some level of perturbation is good

1. Which is higher than what it actually pays, so these people avoid an equity MF.