#33: Preference uncertainty

Apples and pears—strategy and decision-making when you don't know what you want.

Greetings to readers old and new. This is yet another of Vaughn Tan’s weekly attempts to make sense of the state of not-knowing. The first issue explains the project; you can see all the issues here.


Strategy is driven by preferences and is easy when preferences are clear—and we haven’t given enough consideration to the circumstances under which preferences can become uncertain. Warning: half-baked.


Strategy is about choosing actions to achieve desirable outcomes.

It’s lunchtime: I am offered an apple or a pear to complete my MealDeal™. I don’t especially like pears and love apples, so my choice is clear.

The correct strategy in both cases is clear because my preferences in each situation are clear: get the apple because I know it’s more delicious, and I know that I prefer more delicious things.

But preferences often become uncertain as soon as we give them serious consideration.


What does it mean to have a “preference” and how does it affect strategy and decision-making?

When talking about my preference for apples over pears, I mean that I prefer the [outcome of the action of choosing and eating the apple] to the [outcome of choosing and eating the apple]. This is how preferences drive decision-making and strategy: they provide a basis for choosing between actions based on what outcomes those actions result in.

More specifically, making a Decision (now) to take an Action requires a hypothesis about the value of the Outcome that will result (after):

Decision (now) → Action → expected Outcome (after)

A concrete version of this D-A-O chain for the apples vs pears situation is:

I should choose and eat the apple → I choose and eat the apple → This apple that I will choose and eat will be more delicious than a pear

In other words, the decision to take any action is based on how much we expect to value the outcome we expect from that action. That last bit is important.


To examine how preferences become uncertain when we take them seriously, it’ll help to simplify the thought experiment and assume that everything else is equal (or ceteris paribus). Let’s assume apples and pears cost the same because they’re part of the MealDeal™, etc. (Famous economists frequently do this to highlight important dynamics when analyzing and making policy decisions that affect billions of people.)

Even in the case of fruit, uncertainty easily arises in how much we should expect to value the outcome (eating the chosen fruit). Because the rest of the situation is now cartoonishly simplified, we can make the fruit bit more realistic.

Here are just two scenarios in which fruit preferences become uncertain:

  1. Just one type of apple and one type of pear are offered, but the pears are certified biodynamically grown, while the apples are grown liberally bathed in pesticides and fungicides. Would I get more value from eating a delicious apple which is probably toxic or from eating a meh pear which is healthful?

  2. Many different types of apples and pears are offered, each one with a different texture and flavor—I’ve tried some of them before but there are other types unknown to me. Would I get more value from eating a type of apple with a known deliciousness or from eating an unfamiliar type of apple of unknown deliciousness?

In the first scenario, the set of possible outcomes is clear (eating a toxic delicious apple vs a healthful meh pear) but each fruit combines qualities that aren’t interchangeable (deliciousness and toxicity). This makes it hard to tell which outcome is more relatively valuable, and creates uncertain preference ordering in the set of possible outcomes. These kinds of incommensurable qualities co-occur frequently in outcomes.

In the second scenario, the composition of the set of possible outcomes is itself uncertain because some of unfamiliar types of pears may be wildly delicious and some of the unfamiliar types of apples may be absurdly untasty. I don’t know how much I should expect to value eating this unfamiliar type of apple relative to that unfamiliar type of pear. Because the composition of the set of possible outcomes is uncertain, my preference ordering of that set of possible outcomes is uncertain too. Again, incomplete knowledge of the range of outcomes is a common occurrence.

In both scenarios, my preference for an apple vs a pear has become uncertain, and the two causes of this preference uncertainty (incommensurable qualities for outcomes and incomplete knowledge of outcomes) are both common.


Preference uncertainty isn’t restricted to inconsequential fruit dilemmas. It’s a real problem for businesses too. For instance,

  1. Uncertain preference set ordering: A company that has a choice between [making more profitable widgets in a way that endangers workers] or [making less profitable widgets in a way that keeps workers safe] faces uncertain preference set ordering if it doesn’t have an explicit understanding of how much it is willing to trade worker safety for profitability. This tradeoff is particularly hard—perhaps impossible—to make explicitly because worker safety and profitability aren’t (and shouldn’t be) truly interchangeable. (But workers and businesses still make their own implicit tradeoff decisions all the time. We see this in the profusion of workplace injury lawsuits, the existence of hazard pay, and occupational safety regulations.)

  2. Uncertain preference set composition: A company that has the option of [investing in new product development] alongside [investing more in making and selling existing products] faces this problem. Innovation definitionally means not knowing for sure what to do or what the outcome will be—this makes the composition of the set of outcomes uncertain, so preferences among those outcomes become uncertain too. Not knowing for sure the value of innovation outcomes makes it hard to know how much to invest in innovation work. (This is a common problem for companies and one explanation for why incumbents often get overtaken by more innovative new entrants.)

Many business decisions are made in relation to outcomes which cannot really be compared with each other, or which aren’t fully understood. Both situations create preference uncertainty—this makes it hard for the company and its leaders to decide what actions to take.


When deciding how to act, we choose from among available D-A-O chains: decisions to act are based on our preferences for the outcomes we expect from taking action.

When preferences are certain, it is clear which outcome to aim for and so which action to decide to take. Preferences can become uncertain when 1) the set of possible outcomes is clear but relative preferences between them is not, or 2) when the set of possible outcomes is itself unclear. When preferences are uncertain, decision-making becomes hard.

The diagram below is my attempt to graphically illustrate how decision-making is screwed up by these two causes of preference uncertainty (B and C), and to contrast those situations with decision-making when preferences are certain (A):

Regardless of the cause, decision-making becomes difficult when there is preference uncertainty—and strategy becomes difficult too.


More time with my back to the wall.


Share

Find me on the web at www.vaughntan.org, on Twitter @vaughn_tan, on Instagram @vaughn.tan, or by email at <uncertaintymindset@vaughntan.org>.