Keep Two Thoughts

Personal essays

Competing Errors - Essay from Newsletter 53

Drawing the line

Dad’s cooking

In the old days I’d be cooking in the kitchen not seeing that it was getting a little smokey in there and Kim would come rushing in to open the window.

Too late - the smoke detector would start beeping and I’d have to stop cooking and stand on a chair and take the battery out of the detector until the smoke cleared.

One of the kids would yell from upstairs, “mom, what’s that noise.”

“It’s ok,” Kim would call back, “your father is cooking.”

There’s no fire, I’d think to myself, I don’t know why the alarm went off.

But what’s the alternative?

I’d rather the alarm goes off when there isn’t a fire than it doesn’t go off when there is one.

Where’s the line

The folks who design and make the smoke detector have to draw the line somewhere and they would rather we be inconvenienced every now and then by an alarm going off than it underreport a real fire.

But where do you draw that line?

You don’t want to have the alarm go off every time you make a cup of coffee.

And so the line is drawn so that there is some balance between going off to often and not going off when it should. Some acceptable level of failure is determined.

In statistics this is the notion of Type I and Type II errors.

You make a hypothesis which we call the null hypothesis. For example, supposedly in the American justice system the null hypothesis is that a defendant is innocent. Unfortunately, in reality this depends on who the defendant is - but that’s an issue for another day.

A Type I error is where we reject the null hypothesis based on an experiment even though the null hypothesis is true.

In our courtroom example, that means we find the defendant to be guilty even though they are really innocent.

A Type II error is where we don’t reject the null hypothesis even though it is false.

Again, in our courtroom example, this means we find the defendant to be not guilty even though they are guilty.

The other two possibilities are not errors - a guilty person is found guilty, an innocent person is found to be not guilty.

We want to reduce the errors if possible, but often reducing the possibility of a Type I error increases the possibility of a Type II error and vice versa.

So what do we do?

We make a judgement that wrongly convicting innocent people is bad and so we accept that unfortunately this means that we may increase the number of guilty people we fail to convict.


And then politics gets involved.

Because every once in a while one of those guilty people who have been let free go on to commit other crimes.

And the next thing you know we have Willy Horton commercials splashed all over the air that if it weren’t for this politician being soft on crime, those nice people - people who could have been you - would still be alive today.

Maybe the judicial system is too charged a topic - what about something like voting.

We want everyone who is eligible to vote to exercise this right while at the same time we want to prevent anyone who is not eligible from voting.

It turns out that that’s not quite true.

Not everyone wants everyone who is eligible to vote to exercise this right.

Before the Georgia election I signed up for some texting through Stacey Abrams’ Fair Fight. The training was clear, we were to encourage everyone to register and vote. It didn’t matter who they wanted to vote for - our goal was to help them vote.

That was such an easy cause for me to get behind. The goal was to make it easier for everyone who should vote to vote. The risk was that some people who shouldn’t vote would be able to.

And yet - looking back at the November election and the January special election, there were almost no cases of voter fraud.

You’d think this would be a win, and yet

Not everyone wants everyone who is eligible to vote to exercise this right.

Null Hypothesis

The Type I and Type II errors depend on what your underlying hypothesis is.

In the case of the COVID relief bill, the democrats said that they’d rather make the mistake of giving some undeserving people help they didn’t need rather than miss out on helping people who needed assistance.

Not one GOP senator or representative voted for the bill. They didn’t recognize that some people needed and deserved help from the government and they were awfully concerned about people who make too much going off to cash their government checks.

So many of us live in bubble of like-minded people. My friends can’t understand how you can make it illegal to give food or water to someone standing in line to vote in a line that is this long because of arbitrary policies that kept those in line from voting early, by mail, or at a more convenient location that was closed down.

The bubble we don’t see is worried about making it easier for felons, dead people, and folks who live in the city to vote. You know “those people.”

I don’t know how to bridge this gap. For so long I thought we were arguing about Type I vs Type II errors but now I understand we’ve got completely different null hypotheses.

The people who were so worried about four dead in Benghazi that we needed to have endless investigations think that we don’t need any investigations for gun violence, insurrection, or half a million unnecessary COVID deaths in this country alone.

In our bubbles we don’t stand on opposite sides of an argument, each side is arguing with their own straw man and wondering why nothing is changing.

Essay from Dim Sum Thinking Newsletter 53. Read the rest of the Newsletter or subscribe

See also Dim Sum Thinking — Theme by @mattgraham — Subscribe with RSS