As part of my ongoing attempt to formalize, characterize and apprehend the decisions that rule our lives, I had to consider different approaches to tease out a framework in which decisions could be scrutinized. My goal is to explain why decisions that we experience and witness constantly are so impractical. The way we make decisions is extensively studied in plenty of disciplines such as philosophy, neurobiology, maths, economics, game theory, or even language theory, and today I want to lay out a simplified view to outline why our decisions are much less perfect than they seem. Yes, any decision: what to have for dinner, which job to accept, where to live or go on holiday, the name of your children, or when to stop editing your article. Before reading further, think about a decision you would qualify as perfect (if you’re already hesitant to view your choices as perfect, semantically speaking, then consider a “pretty much perfect” decision; that will do the trick).
When we think about what it would take to construct a perfect decision, the first striking argument is to start by asking, what is a perfect decision? This simple question already produces multiple branches and entangled layers to examine. It presupposes that “decisions” are objects we can compare. But what do we really want to balance? Let us do a bit of semantic: are we referring to the outcome or the process itself associated with the decision? And how do we get that ability to compare different decisions, a power that is needed to claim that “decision A” was superior to “decision B”?
In practice, we think of a “perfect decision” as a judgment for which we picked the most rewarding option, the one that yielded the best outcome. That’s perfect, and this is precisely where I was going. Not only is this wrong, as this is solely an effect called “resulting” on which authors like Annie Duke wrote extensively, but this implies we can decompose a decision into three blocks. Indeed, in its simplest form, a decision combines these three components:
- A set of available options to choose from.
- The “core” of the decision, the process itself, the act of deciding.
- Something that occurs from that process: the outcome.
Amongst all the more or less scientific methods to define a decision, this time model in which the information would be flowing from the past to the future might well be the simplest and safest representation.
We already touched upon the middle block: it’s messy, the boundaries are blurred, and the blueprints of this mystical engine that would rule our decisions are out of reach. How do we know that this engine operates as expected and doesn’t require maintenance? How do you know you’re not affected by a syndrome that alters your judgment (drugs, stress, fatigue, autistic syndrome, tumor, to name a few)? This argument alone, dear reader, should be sufficient to persuade everyone it’s seemingly wiser to remain humble before declaring our decision was a perfect one. Substitute “perfect” by “good,” and I think it would still be overstated.
Next, what do we truly know about the options that serve as fuel for our engine? My opinion on this, as you probably guessed, is “very little.” First, there are plenty of decisions for which we cannot analyze the whole set of possible options. When we need to decide which candidate to hire or where to live, it’s not comparable to playing chess, where both players share a complete knowledge of the board and available pieces. Even deciding between a croissant and chocolate bread at your favorite bakery could induce an overload of details to compare, should you have enough time to dedicate to the task. But our time is limited, so we are doomed to deal with greatly deficient approximations of the available options in front of us: we won’t have access to all the relevant details a candidate has to offer; we won’t get to know what it’s like to live in Paris or in Tokyo before effectively residing there. Even then, we can still argue that the experience will vary from one street or district to the next. When comparing houses, the most relevant judgment we can make would be something like “they’re comparable,” when they effectively have a similar number of bedrooms, bathrooms, etc. Yet, we’re subject to miss a plethora of less prominent features such as roof and wall insulation, the quality of the double-glazed windows, the quality of the schools (or bakeries!) nearby, and so forth.
My tailor is rich. Yours might be richer.
Thus far, imperfect inputs processed by a messy decision engine seem to be a more accurate depiction of real-world situations. Maybe we’re better off rolling dice instead and claim we’ve made a genius decision if it has led to an outstanding outcome. Well, most of the decisions aren’t just a lottery, or again, a chess game. In games, a set of rules define the conditions for wins and losses, and the game’s creator assumes you’ll be looking for the big prize or the victory. At the end of each round, you either win, lose, or draw. This situation allows a unique property: you get access to high-quality feedback and learn from there. One could argue that some real-world decisions, primarily based on numbers, follow this paradigm: when we decide to buy (or to not buy) a stock, we eventually get to know if that was a correct judgment, based on the price evolution. Even there, the time window used to analyze the decision will influence how we perceive the outcome: the same decision could be deemed “outstanding,” “catastrophic,” or anything in between. If you purchase 10,000 shares of a company at $7.00 in 2021 and decide to sell them at $12.00 per share in 2022, it’s reasonable to think you completed a fruitful transaction. But what if you realize you could have sold them at $21.00 a few months later? Those “what if” analyses make the learning process incredibly challenging: what if you had joined this other company instead, or if you had looked more closely at the other properties on your list before relocating to Paris? This counterfactual reasoning is getting even worse when we’re blind and refuse to accept reality, thinking we effectively made the best choice. I’ve got another bad news. This buy/sell decision is highly driven and measurable by an objective metric: the money you made or lost. But when it comes to purchasing a new home or choosing your next job, we can’t rely on a singular figure. Symmetrically with what we considered regarding the options, the set of possible attributes to examine is infinite; at best, we would compile that long list with something along the lines of “quality of life.” To summarize, not only do we have access to a minimal set of possible outcomes, but we can hardly compare the outcomes between them.
Is this only a game?
A proper introspection of the different outcomes should enable high-quality learning, which is to be utilized later to make us smarter. Said differently, our output — past outcome — has become an input, or a modifier of the newly available options, for a subsequent choice, nurturing the cycle of decisions that we hope will only get better as we accumulate insights. However, we saw that we have to rely on flawed inputs, unclear decision rules, and hardly actionable learnings. In fact, even in the context of games with perfect information like chess, analyzing your performance demands some effort, and it’s probable you won a match based upon incorrect decisions and vice-versa. This brings me to the closing words of this brief article: there is no such thing as a perfect decision, and the best decision in your life is probably not the last one you made, but the next one you have yet to make.
About the Author
I manage engineering teams to build impactful products, and I’m reflecting on the past decisions I struggled with to prepare a book on what I call “Unpractical Decisions.” I support the Alliance for Decision Education.