Tuesday, August 05, 2008

The effect of measurement

A friend of mine bought a hybrid car—one of those cars that runs on both gas and a battery engine—and when he gave me a ride, I was struck by two things. First, the car is amazingly quiet when it is running on battery power; in fact, I wouldn’t have even known that the engine was running except we were driving down the street. This, I thought, would be perfect for sneaking up on pedestrians (but that’s a topic for another blog post). Second, my friend, who is otherwise quite sensible, spoke at great length about the gas mileage that he gets with the car, and how it varies by driving patterns and terrain. Apparently braking slowly is good (or bad) because it does (or doesn’t) charge the battery. (You can tell I wasn’t paying too much attention).

Now, I thought my friend was unusual in his fascination with miles-per-gallon until I read this Washington Post article about hybrid owners. It tells of various owners who seem willing to do anything for that extra mile-per-gallon. One driver changed his route to work, just to avoid a big hill that drops his mileage to below 20 mpg. Another is so keen on keeping his mpg high that he won’t let his wife—who apparently just drives normally—drive his hybrid.

The key feature of hybrids that makes this mpg obsession possible is a dashboard mileage monitor that indicates how many miles-per-gallon the car is getting at that exact moment. This feedback appears to change drivers’ behavior.

Sociologists have long understood what is called an observer effect (also called the Hawthorn Effect). The idea here is that people change their behavior when they know they are being observed. This is why sociologists will sometimes use covert observation to study a situation, so as not to change it unnecessarily. It’s also why experimenters will often deceive their subjects about the true purpose of the study.

Well, related to the observer effect might be something that we can a measurement effect. Just the act of measuring a behavior changes it to be (usually) more in-line with our preferences and goals.

This principle applies to much more than driving hybrid cars. In fact, when people want to change their own behavior or that of others, one of the first things they’ll do is start measuring that behavior. It’s remarkable in how many areas of life we use this measure-to-change-it approach to behavior.

Weight Watchers is one of the best known weight loss programs. What’s one of the first things that a person does at the Weight Watcher’s meeting? Step on a scale, and have someone write down how much you weigh. This measurement brings your attention to what you’re trying to do, and it indicates how well you’ve done it in the previous week.

Most money management programs operate on the same principle. They have you keep track of all your expenses (i.e., measure them), and then see how they change over time. (I tried a program, called Money Counts, and I realized that I’m better at sociology than managing money.)

Want to live a more holy life? Start confessing. The Catholic Church encourages its members to periodically tally up their sins for a priest who then (hopefully) absolves them of these sins. The awareness of sin that the periodic confessions encourage should help move the individual away from behaviors they don’t desire.

Fundraising programs not only ask for money, but they also let their target audience know how much they’ve already raised. This is why every summer we see signs with thermometers painted on them. The more money given, the higher the red-mark on the thermometer.

I suppose that even classroom grading works this way. Students who know their grades throughout their semester probably study harder and are more engaged in the tests than those who are not told their grades. (This is why professors always tell students their grades.)

This principle has implications for social research. Just the act of measuring someone’s behavior, e.g., as is done in a survey, can change that person’s behavior by making them more aware of what they are doing. This may not matter in a cross-sectional survey, done only once, but with longitudinal research, it may alter the data. That is, if we measure a person’s behavior a second time, we may observe something different than if we hadn’t measured it a first time.

What’s the bottom line here? Well, when we measure anything, whether in professional research or everyday-life, realize that we’re probably changing some aspect of it. If we want to change something, probably the first thing to do is to start measuring it.

No comments: