Questions, Proxies, and Health

When faced with a difficult question, we often substitute a simpler question and answer that instead. Here are three examples:

• What’s the crime rate in your neighborhood? – We probably don’t keep close tabs on the crime rate. We’re not naturally good at statistics either so it’s difficult to develop a reasonable estimate. So we substitute a different question: How easy is it for me to remember a crime committed in my neighborhood? If it’s easy to remember, we guess that the neighborhood has a high crime rate. If it’s hard to remember, we guess a lower rate. (I’ve adapted this example from Daniel Kahneman’s book, Thinking Fast and Slow).

• How’s your car running? – It’s hard to know how well a car is running in this age of sophisticated electronics. So we answer a different question: How does the car sound? If it’s not making strange noises – knocks and pings – it must be running well and performing optimally.

• How effective is your shampoo? – I suppose we could study our hair’s health every day but most of us don’t. So we answer a simpler question: How much lather does your shampoo produce? If we get a lot of lather, the shampoo must be effective.

How many do I need to get 10 calories of energy?

In each case, we substitute a proxy for the original question. We assume that the proxy measures the same thing that the original question aimed to measure. Sometimes we’re right; sometimes we’re wrong. Most often, we don’t think about the fact that we’re using a proxy. System 1 does the thinking for us. But we can, in fact, bring the proxy to System 2 and evaluate whether it’s effective or not. If we think about it, we can use System 2 to spot errors in System 1. But we have to think about it.

As it happens, System 1 uses proxies in some situations that we might never think about. Here’s an example: How much food should you eat?

(The following is based on a study from the University of Sydney. The research article is here. Less technical summaries are here and here).

We tend to think of food in terms of quantity. System 1 also considers food as a source of energy. System 1 is trying to answer two questions: 1) How much energy does my body need? 2) How much food does that translate to?

Our bodies have learned that sweet food delivers more energy than non-sweet food and can use this to translate from energy needs to food requirements. Let’s say that the equation looks something like this:

1 calorie* of energy is generated by 10 grams of sweet food

Let’s also assume that our body has determined that we need 10 calories of energy. A simple calculation indicates that we need to eat 100 grams of sweet food. Once we’ve eaten 100 grams, System 1 can issue a directive to stop eating.

Now let’s change the scenario by introducing artificial sweeteners that add sweetness without adding many calories. The new translation table might look like this:

1 calorie of energy is generated by 30 grams of artificially sweetened food

If we still need 10 calories of energy, we will need to eat 300 grams of artificially sweetened food. System 1 issues a directive to stop only after we’ve eaten the requisite amount.

System 1 can’t tell the difference between artificially and naturally sweetened foods. It has only one translation table. If we eat a lot of artificially sweetened food, System 1 will learn the new translation table. If we then switch back to naturally sweetened foods, System 1 will still use the new translation table. It will still tell us to eat 300 grams of food to get 10 calories of energy.

We would never know that our brain makes energy/quantity assumptions if not for studies like this one. It’s not intuitively obvious that we need to invoke System 2 to examine the relationship between artificial sweeteners and food intake. But like crime rates or cars or shampoos, we often answer different questions than we think we’re answering. To think more clearly, we need to examine our proxies more carefully.

*It’s actually a kilocalorie of energy but we Americans refer to it as a calorie.