If you aren’t familiar with “Am I the Asshole?” you should check it out. It’s a subreddit where people go to, well, find out if they’re assholes or not. They post a description of their situation, and the rest of Reddit judges them. The posters can be anywhere from outrageous villains (“Am I the asshole for telling my adopted brother that he’s not entitled to my time because he’s not my real brother?”) to clear victims (“Am I the asshole for taking back the beer I bought for a party from which I was uninvited?”), and it’s up the readers to decide which is which. The most common judgements are YTA (“You’re the Asshole”) and NTA (“Not the Asshole”), with ESH (“Everyone Sucks Here”) and NAH (“No Assholes Here”) making rare appearances. Once judged, people will do anything from listening and trying to make amends to yelling at the people who called them assholes.

I love this subreddit. It gives you a tiny window into people’s lives and a chance to be judgemental in a harmless way. Sometimes if my friends and I have nothing to do, we’ll pull it up and just give our opinions. It’s a fun, practical look at morality.

The most fascinating posts though, are the ones where everyone is SURE they know the answer, and no one agrees. 

There are obviously plenty of posts that are borderline, where you read it and think “Yeah, they could have handled that better, but I don’t know if their behavior crossed the line.” I’m not talking about those. I’m talking about the feeling of seeing a post that says “Today I kicked a puppy for no reason and then punched a baby, am I the asshole?” and watching the overwhelming verdict be “NTA”. It is surreal to see dozens of people sympathizing with the obvious villain of the post.

The source of human ethical reasoning is complex. We’re social animals, so certain behaviors like gratitude and fair play are hardwired into us. Others we learn as we grow, observing our parents and friends and our media. And, of course, we reflect on what we’ve seen and decide for ourselves what is right and wrong. But given all of that, how can we sometimes get such different answers? After all, it’s not like half the judgements are being beamed down from a UFO by aliens that have their own ancient rules and customs. The people I disagree with could be living next door.

Even if you don’t frequent the subreddit, you might have experienced a similar confusion when it comes to politics. Too often I look at a political situation and think “how could someone possibly agree with / support that?” I think it’s a fairly similar issue: different groups in our country have different values. Am I the Asshole feels exactly like that on a smaller scale.

This fascinates me. Where do these different values come from? What are the exact differences? If it’s hard to see in politics, might it be easier to see somewhere we can directly gather data from?

Let’s switch gears for a second.

Principal Component Analysis

Principal component analysis (PCA) is a technique for reducing the dimensionality of datasets, increasing interpretability while at the same time minimizing information loss. It does so by creating new, uncorrelated variables that successively maximize variance.

Sorry, let me try that again. 

As human beings we love categories and spectrums. Are you a Ravenclaw or Hufflepuff? What’s your Myers Briggs? Are you an extrovert or an introvert?

It’s no wonder we’re fascinated by these categories; they’re extremely helpful, cognitively speaking. In computer science we call these compression algorithms. I take my friend, with all of their complexity and nuance that I cannot possibly describe to you, and say “they’re such a Ravenclaw.” From that you can infer a number of things about my friend.

A lot of the more obvious spectra we like to use are political. Who hasn’t seen a chart like this?

This square takes something large and complex and boils it down to just two dimensions: left / right and authoritarian / libertarian. 

Like I said, people do this kind of thing all the time. But we don’t usually know if our divisions are good ones. What if there was some political stance that was left off that graph, like globalist / isolationist? What if authoritarian / libertarian isn’t actually a very good divide? We don’t really know.

But we can use math to find out.

Principal component analysis is a mathematical technique that lets us take something really complicated (we would say high dimensional) and boil it down to something like that square. We use this because things that are low dimensional (like the square) are easier to reason about and process. But the interesting thing about PCA is that it always gives you the best components. The spectra that it makes the most sense on which to judge the data.

You are probably not as blown away by this as I am, but I think it’s awesome. I tend to think of every person as carrying around their own little set of numbers representing how much they care about certain things, and if I knew those numbers I could figure out how they’d react to any moral situation. What would those numbers be? Would they be “How much I value kindness” or “How much I respect authority?” Are there just a few fundamental things people care about when they judge others? Are there a lot? How similar are we to our friends, our neighbors, or our fellow citizens?

And, more to the point, can we use PCA on the Am I the Asshole subreddit to figure out who values what?

The Answer!

No. Not really. To use principal component analysis, you need lots and lots of questions answered by the same people. While Am I the Asshole is huge, not everyone comments on the same posts, which means we can’t really find out the principal components of anything. I did manage to post a survey designed to gather some of the data I wanted, but it only ended up with around a hundred responses; a far cry from the thousands I would need to do any real math. 

Fortunately for us, someone else already had a similar idea. A reporter from Vice.com discovered the subreddit, and convinced the mods to post a short survey about people’s values. This data gives us the perfect opportunity to try our trick: if you could only know a few things about a person when trying to guess their answers, what would they be?

I realize this next section can be a little confusing if you’re not used to thinking about vector spaces, so I’m going to try and make sure all the terms I’m going to use are clear. Well, one term. When I talk about a component, all I’m talking about is a way of describing a person with a number. Think “how much of an introvert / extrovert someone is.”

When I ran it over the data the journalist gathered, PCA told me there were two main components, two lines on which we can distribute people. PCA is just a pile of math, so it can’t really tell me what they are, but I can look at what the most important things you’d need to ask to determine where they are on a given spectrum. Think of this like an online quiz in reverse: we have the questions, but we need to figure out how the quiz is trying to sort us. 

Here is the questions for the first, most important component:

• It matters whether or not someone showed a lack of respect for authority
• I think it’s morally fine that rich children inherit a lot of money while poor children inherit nothing. 
• It matters whether or not someone violated standards of purity and decency
• I am proud of my country’s history.
• Men and women each have different roles to play in society.
• Respect for authority is something all children need to learn

The answers to all these questions are tightly bound together. Look familiar? It does to me. It reminds me very strongly of things that would be considered traditional values. So the number one most important thing you could know about someone to guess their survey answers is “how traditional are they.”

Ready for number two?

• It matters whether or not someone suffered emotionally
• It matters whether or not someone cared for someone weak or vulnerable
• Compassion for those who are suffering is the most crucial virtue
• I think it’s morally wrong that rich children inherit a lot of money while poor children inherit nothing
• It can never be right to kill a human being

This one was a little trickier for me, but I decided to call it “valuing compassion.”

So for this dataset, we have something like this:

People don’t clump into groups on these axis, in fact, they mostly stay around the middle, which makes sense to me, since people don’t tend to be that extreme:

Playing into my biases, caring about compassion tends to correlate strongly with responder gender:

Coloring it by the responder’s stated politics provides more interesting results:

So what does this tell us? Well, for one, that all of this analysis isn’t completely off base, because otherwise we wouldn’t see any patterns in these graphs, which is reassuring. More importantly though, we can see that people who identify as liberal are lower on the tradition and higher on the compassion axis. Conservatives, on the other hand, have high values for tradition and lower values for compassion. Which, given that I started this project trying to figure out if different cultural groups tended to have similar values, is a pretty satisfying result.

Final Thoughts

I started this project earnestly believing that deep down, everyone was trying to minimize overall suffering, and we just couldn’t agree how to do that. I had this idea that people had different values from me, but in my head that meant they just had different ideas of the best ways to minimize other people’s pain. I had expected that the political graph, my best indicator of what an actual group of people valued, would have a red top half and a blue lower half.

I didn’t really understand what having different values meant. I expected that there would be people who valued tradition, or thought that chastity was a virtue. But it didn’t really hit me until I started writing this conclusion that there would be people who might value those virtues over hurting other people. I expected that if I encountered someone making that trade off, I could simply say “well, look, that person’s lack of chastity isn’t hurting anyone,” and they would realize that being compassionate outweighs the less important virtue of being chaste. I’m starting to realize that’s simply not the case. And to people for whom chastity is a higher level virtue, they might be equally confused that I don’t understand why it is important to keep someone chaste even if it hurts them.

This was supposed to end with something about how we’re more the same than we are different, but people are just…pretty different. We all value different things, and it is hard to reconcile some of those choices and values. I still have this urge to say that all views are equally valid. For people who scored differently on these questions, I get that there’s nothing wrong with you, that you’re reasonable people who just happened to come up in a different culture from me. I could just as easily written this essay from your perspective. There’s nothing special or better or what I care about.

But it sure is hard not to feel like there is.

If you found this interesting, there’s another way you can help me do more of this kind of analysis. I’d love if you took a second and filled out this survey. If I get enough responses, I might be able to do some more interesting analysis. Also, it’s kind of fun, especially to do with a friend or partner.