Editor’s note: The following is the prepared text of the address delivered by statistician and writer Nate Silver H’18 at Kenyon’s 190th Commencement on May 19, 2018. More coverage of the Class of 2018’s Commencement is available here.
Thank you to President Decatur and to the whole Kenyon College community for having me here today and for giving me this honorary degree and this opportunity to speak with you. It’s a really wonderful honor and something I’ll always cherish.
And more importantly … congratulations to the 420 students who are graduating today! This is your day, one of those rare days when you’ll wake up in one chapter of your life and go to bed in a whole new one, and I hope you’ll remember it fondly.
Before I get into the core of my talk, I want to start out with one fairly general piece of advice based on my own experience after graduating from the University of Chicago 18 years ago. As someone who deals in statistics, I’m wary of drawing conclusions from a sample size of one. Especially when the one person is yourself. But I just turned 40 this year, and that’s officially the age at which you’re allowed to pass off anecdotal evidence as “wisdom.”
The advice, however, is also drawn from the knowledge that this is going to be a fairly abrupt transition for most of you. Kenyon College is a small, tight-knit community — but, statistically, not very many of you are going to stay in Gambier after college. Many of you won’t even be staying here tonight. You’re going to make new friends. You’re going to lose touch with old friends. You may be starting a full-time job for the first time. You may be responsible for your financial situation in a way that you weren’t before.
At a time like that, it will be easy for you to feel untethered. It will be a lot to deal with — probably a harder transition than you realize. It may be difficult to find the right balance between experimenting with your personal and professional identify on the one hand, and staying true to yourself on the other hand.
So the advice is: don’t let yourself get stuck. Don’t let yourself get stuck in a career that’s neither (a) something you really want to be doing right now, or (b) something that positions you to do something you really want to do down the road.
Most of you will be in category (b) — you won’t get your dream job right out of college, and instead it will take a lot of hard work to arrive at it. But if you don’t think you’re on the right path, take advantage of the fact that you’re untethered and be willing to switch gears pretty quickly.
I can speak from experience that the downside to having a job you don’t like is really down, and the upside to having one you do like is really far up. I started out doing something called transfer pricing consulting, which is about as exciting as it sounds. I quit that job to — of all things — play poker, and to develop a statistical system to forecast how Major League Baseball players would perform. (I can imagine some of your parents are terrified at this advice right now.) Anyway, one thing led to another — the poker route dried up, and I kept pivoting. In 2008, I founded FiveThirtyEight in what I thought would be a side project that would take up a few hours a week. The site took off and was later licensed by the New York Times in 2010 and 2012, and then later sold to ESPN in 2013 and more recently to ABC News. It’s gone from a one-person blog to a 35-person news organization.
If I were graduating from college right now instead of in 2000, then frankly my path might be a little more straightforward. There are lots and lots of careers now in “data science” — a term that hardly even existed until a few years ago. There are also careers in data journalism, which is the name we use for what FiveThirtyEight does. There are even several hundred people around the country who make their careers performing statistical analysis for professional sports teams, which is not that many but is way more than when I graduated, when there were like … two.
So back when I graduated, people using statistics in fields like sports and politics and journalism were really on the outside looking in. There was a rebellious spirit. We questioned authority. We wanted to shake things up. We were critical of what we saw as lazy, complacent, self-serving decision-making by established companies and institutions. We thought these organizations didn’t want to employ statistical analysis — or any other sort of rigorous analysis — because it might reveal that the conventional wisdom was wrong and that they didn’t know what they were talking about.
Nowadays, everything is a whole lot different. Everyone is using data. Three out of the four remaining teams in the NBA playoffs — the Rockets, Celtics and Warriors — are extremely proficient at statistical analysis. (The fourth one is the Cavaliers — so I guess you can say that to win in today’s NBA you either have to have a great analytics department … or LeBron James.)
Of course, there are lots of questions about how people are using data — we’ll get to that in a moment. But for better or worse, it’s no longer really acceptable to claim you don’t care about being data-driven. Even President Trump claims that he couches his decisions in data: “I call my own shots, largely based on an accumulation of data, and everyone knows it,” he tweeted last year.
The flip-side to this is that the the “nerds” are no longer on the outside looking in. Instead, they’re probably running the company. Power has shifted toward people and companies with a lot of proficiency in data science.
I obviously don’t think that’s entirely a bad thing. But it’s by no means entirely a good thing, either. You should still inherently harbor some suspicion of big, powerful institutions and their potentially self-serving and short-sighted motivations. Companies and governments that are capable of using data in powerful ways are also capable of abusing it.
What worries me the most, especially at companies like Facebook and at other Silicon Valley behemoths, is the idea that using data science allows one to remove human judgment from the equation. For instance, in announcing a recent change to Facebook’s News Feed algorithm, Mark Zuckerberg claimed that Facebook was not “comfortable” trying to come up with a way to determine which news organizations were most trustworthy; rather, the “most objective” solution was to have readers vote on trustworthiness instead. Maybe this is a good idea and maybe it isn’t — but what bothered me was in the notion that Facebook could avoid responsibility for its algorithm by outsourcing the judgment to its readers.
I also worry about this attitude when I hear people use terms such as “artificial intelligence” and “machine learning” (instead of simpler terms like “computer program”). Phrases like “machine learning” appeal to people’s notion of a push-button solution — meaning, push a button, and the computer does all your thinking for you, no human judgment required.
But the reality is that working with data requires lots of judgment. First, it requires critical judgment — and experience — when drawing inferences from data. And second, it requires moral judgment in deciding what your goals are and in establishing boundaries for your work.
Let’s talk about that first type of judgment — critical judgment. The more experience you have in working with different data sets, the more you’ll realize that the correct interpretation of the data is rarely obvious, and that the obvious-seeming interpretation isn’t always correct. Sometimes changing a single assumption or a single line of code can radically change your conclusion. In the 2016 U.S. presidential election, for instance, there were a series of models that all used almost exactly the same inputs — but they ranged in giving Trump as high as roughly a one-in-three chance of winning the presidency (that was FiveThirtyEight’s model) to as low as one chance in 100, based on fairly subtle aspects of how each algorithm was designed.
And in psychology and other fields, there’s what’s known as the replication crisis, which refers to the fact that something like 60 percent of findings published in academic journals can’t be adequately duplicated by other researchers. Change just one small thing in the experiment — or even run the exact same experiment again — and what’s claimed to be a highly statistically significant result can disappear.
Before I go any further, I want to be clear about what I’m not saying. I’m not saying that there’s no such thing as an objective world. I’m not saying that truth is merely socially constructed, etc. I’m not a postmodernist. Nor am I saying you can’t tell a good statistical method from a bad one. But the goal of a statistical model should be to describe the world as it is; since the world is complicated, the modeling can be complicated, too.
Data science also requires moral judgment. Once you gather a certain type of information, it’s extremely hard to contain it — it can be hacked, sold for profit or commandeered by the government. Or a company that once had benign intentions could decide that maybe it does want to be evil, after all.
These moral issues can be even more profound when they involve questions of how to treat different groups of people. For instance, a lot of supposedly “objective,” data-driven algorithms to determine criminal sentences can encode and reinforce racial bias based on how variables are chosen.
There’s also judgment required in knowing what problem you’re trying to solve. To turn back to Facebook, one reason they’ve had so much trouble figuring out what news content to show to people is that they have conflicting goals — at various times, they’ve talked about wanting to prioritize news that’s (i) trustworthy, (ii) that’s “relevant,” (iii) that promotes “meaningful social interactions” or (iv) that maximizes “engagement.” Those are rather different objectives that don’t necessarily have much to do with one another. You could probably design an algorithm to optimize for any one of those goals reasonably well. But if you don’t know what your goal is, you’re going to have a mess on your hands.
So … where does that leave us?
Well, the bad news is that these are really hard problems. The good news is that, mere moments from now, you’ll officially receive a degree from Kenyon College, in what I hope is the capstone to an amazing four years that you’ve had here. I’m a huge believer in a liberal arts education like the one you get at Kenyon because the goal is to develop critical thinking skills — and those skills will stay with you for a lifetime as you experiment with different passions.
If you’re pursuing a career in a statistics-related field, I hope you’ll use that Kenyon education to combine good data science with good judgment. And I hope you’ll remember that statistics are important because they help us to describe and explain the real world all around us — they aren’t that important unto themselves.
And if you’re not undertaking a career in data science, I hope you’ll nevertheless develop and maintain enough knowledge about statistics that you’re comfortable scrutinizing statistical claims when they’re made by powerful companies or governments or politicians. It’s an essential part of being a well-informed citizen.
Either way, most of you will start out your careers feeling like you’re on the outside looking in, like I felt 18 years ago. And many of you will become very successful and very powerful despite that. When that happens, don’t stop thinking critically. Don’t stop questioning the data, questioning authority — and questioning yourself.
Thank you, students, and congratulations!