Musk's still here
November 21, 2018
Back in August, my friend Ernest threw it down:
I believe Elon Musk is out at Tesla in the next 3 months and will bet a six pack of good beer on it.
I haven't gotten the six pack yet, though I know he's good for it. I'm more interested in what made him see it that way—why did he believe what he did, and how was he so certain? His mistake is one I've made myself, countless times: forgetting that, more than anything else, journalists are storytellers.
My favorite journalism is the kind that has no obvious precipitating event. Take this recent story, from The Economist, "Teenagers are better behaved and less hedonistic nowadays":
AT THE gates of Santa Monica College, in Los Angeles, a young man with a skateboard is hanging out near a group of people who are smoking marijuana in view of the campus police. His head is clouded, too—but with worry, not weed. He frets about his student loans and the difficulty of finding a job, even fearing that he might end up homeless. “Not to sound intense,” he adds, but robots are taking work from humans. He neither smokes nor drinks much. The stigma against such things is stronger than it was for his parents’ generation, he explains.
This isn't a story that "breaks". And yet, it talks about a deeply important trend: across many countries including the United States, teenagars are drinking, smoking, and having sex much less than their parents.
You wouldn't know it, reading a lot of what passes for journalism today. Across countless topics—violent crime, poverty, literacy rates—the most important trends rarely appear in the news. How does one write about falling suicide rates? Where's the story?
But it's actually worse: the important long-term stuff gets left behind, but in its place, a series of articles written about what's breaking today; not only is that irrelevant, but much journalism tends to grossly overestimate both the frequency and severity of major events, everything from recession to impeachment.
I've seen myself fall into the trap too many times: I read too much print journalism and find myself dramatically overestimating the likelyhood of non-status quo outcomes. I dramatically overestimate the probability of a revolution, versus things basically staying the same. Because the fact is, the world doesn't change much day-to-day.
But that doesn't make for great journalism.
I've gotten better at predicting things over the past years. It's a combination of a few things, both intentional and accidental:
- I'm older. Each day is sort of a generalized training set, as I've written before, everything gets less surprising
- I'm a better consumer of the news; I've watched politics and current events particularly closely since I was about 16, reading Newsweek cover-to-cover and doing debate throughout high school
- I'm not a heavy user of facebook and follow people pretty selectively on Twitter; that puts me in my own little bubble but I tend not to mind. There's too much selective editing, vitriol, and straight-up trash getting shared.
- I've tried hard to follow Paul Graham's advice: "This is isomorphic to the principle that you should prevent your beliefs about how things are from being contaminated by how you wish they were. Most people let them mix pretty promiscuously."
- I've read a good number of books specifically on the topic of judgment and behavior:
- Taleb's Incerto
- Thinking Fast and Slow, by Kahneman. This book taught me what's perhaps the most important general rule of judgment, the base rate fallacy: focus more on base probabilities (the right thing to do) and less on situational specifics (the fallacy). Example: many are certain Trump will be impeached. The historical odds of that are perhaps 4% (2 in 40-60 depending on how you count it). Even at 5x likelyhood, that's only a 1 in 5 chance; that seems far from "certain" to me.
- Jonathan Haidt's work on decision-making, particularly relating to morality
- Many others: Ariely, EconTalk, Farsighted, Emanuel Derman, Moneyball, Dalio's Principles and Big Debt Crises, some of Tetlock's work on prediction
I'm not saying I'm perfect, but I've found the above helps.