What Should we Eat?

My go to examples are related to food or transportation because I figure most people have a wide variety of experience in those scenarios. Intuition about reasonable choices and priorities (I think) helps make my examples make more sense. But what if we were discussing dinner options with an alien? And what if we take it one step further: On the alien’s planet they don’t need to eat, so the idea of choosing something to eat is totally new to the alien. How should we describe the problem of choosing a meal for someone who has no context of what we mean?

First, an important clarification: There are many kinds of problems. Math problems are one kind where all the information is laid out in front of you. Complaints could be interpreted as a different kind of problem. Maybe there are others. But if you want to do something practical as a result of your analysis of a problem, we will call that a “decision problem.” Any time I use the word problem (at least in this post), I am referring to the category of situations where you are considering making a choice. Why are you making a choice? I believe the key reason to make a choice is that you expect your decision to have an impact on outcomes.

Without formally writing down a decision problem, how would you know that you are in that kind of situation? To answer, I will use one more term in a special way: Status Quo. If you plan to keep the status quo, you do not have to consider what decision you have to make. If you plan to do something to change that status quo, you have to decide what it is that you will be changing. So, to recognize a decision problem you start by saying “I think I want to change something in the real world through my action, as compared to my inaction.”

Back to our alien, what status quo are we considering changing when we talk about “choosing what to have for dinner”? When we are considering the decision of what to eat, I think inaction / the status quo would be eating nothing. We can then ask about actions (not the status quo) of things we could do as an alternative to eating nothing. This is a different way to generalize a problem statement, similar to last post’s idea of a “problem topic.” Also, I say “I think” eating nothing would be the status quo because problem framing is a tricky thing… it’s about what makes sense to an individual or group as a way to discuss the issue at hand.

Taking the above a step further, we now have a new framing from the title of this post and how people normally talk about deciding what to eat. Instead of saying the problem is what to eat for dinner, we have a more general question of “What decision would we like to make to change our status quo of eating nothing?”

Let’s pause here for a moment. When you frame a problem specifically as a decision problem to change the status quo, the natural next step is to ask “in what ways are we able to change the status quo? That is, what decisions could we make, and what would change about the world as a result of a chosen action?” As you may notice, this way of framing problems only makes sense when we are talking about a decision problem.

As an opposing perspective: what does it means to make a decision without considering the status quo or how our decisions would change the future. Imagine you simply asked the alien what to have for dinner from a list of 4 options. Would an alien who has never eaten have any basis to make that choice? How would they be able to choose between a salad, a burger, and eating nothing if they don’t know about hunger, nutrition, cost, or taste?

My point in this post is that decision problems specifically make sense in terms of understanding the status quo, and what would be different about the world if you make a choice to change that status quo. A decision problem that doesn’t at some level compare outcomes is like when your family member walks over to you and says “Tell Jim you agree with me.” You have no context of what it means to agree or disagree. Sure, you could make a decision, but will it be the right one?

p.s. This way of framing problems is the heart of prescriptive analytics. In prescriptive analytics you give a computer a set of decisions it could make and tell it how to connect those decisions to outcomes. There is usually no special “status quo” decision in those models, but comparing alternatives is built in the code.