There aren’t a lot of books whose genesis can be traced back to a Twitter argument. Russ Roberts’ Wild Problems may be one of them.
In 2019, Julia Galef made a tweet about how she wished she could run a study on people unsure of whether they wanted kids. She would try to find what questions predicted satisfaction with their eventual choice decades later.
Russ Roberts replied that having children changed his life in immeasurable ways, and would be skeptical of the merits of such a study.
In 2020, Roberts and Galef continued their debate on Pairagraph, without coming to much of a resolution.
Wild Problems is the book-length version of Roberts’ argument. And as much as I would like to imagine the impetus for the book started solely with that tweet, in reality, regular listeners to Roberts’ podcast EconTalk will note that he has long had skepticism towards what he sees as an overly quantitative mindset - a mindset he once held himself as an economist trained at the University of Chicago.
What Are Wild Problems?
>Whether to have a child is what I call a wild problem - a fork in the road of life where knowing which path is the right one isn’t obvious, where the pleasure and pain from choosing one path over another are ultimately hidden from us, where the paths we choose defines who we are and who we might become. Wild problems are the big decisions all of us have to deal with as we go through life.
These are contrasted against “tame problems” - problems with clear objectives, tried and true formulas, and rich amounts of data, like playing chess, making an omelette, or creating a crypto fraud. Roberts thinks the standard approaches of rationality work just fine for those. What he thinks is a mistake is applying the techniques that work well for “tame problems” towards “wild problems”.
So is this book a how-to guide on approaching your own “wild problems”? While Roberts has a few ideas (in the last chapter, for instance, he suggests “living like an artist”, aiming to explore in a way that enhances serendipity) the very definition of a wild problem implies that a “how-to guide” is of limited assistance, or may even lead you astray.
What’s the point of this book then? More than anything else, I think it was written as a how-not-to-guide for the highly quantitative. This is not a book you’d give to your aunt to help her plan vacations, this is a book aimed squarely at telling rationalists off for thinking that they can solve everything with a spreadsheet.
Roberts dedicates the first half of the book to gently mocking people who attempted to use a “rational approach” to wild problems. There is Charles Darwin, who tried to decide whether to marry by using a pro vs con list (an example of a pro: a wife would be an “o__bject to be beloved & played with. — better than a dog anyhow”). There’s Katja Grace, an AI researcher and rationalist who was undecided about having children - she decided it’d help if she tried caring for a robot baby. And there’s Persi Diaconis, a mathematician who was agonising about whether to move to Harvard from Stanford. A colleague suggested: “You’re one of our leading decision theorists – maybe you should make a list of the costs and benefits and calculate your expected utility.” Diaconis blurted out: “Come on, Sandy, this is serious.”
Roberts thinks that these sorts of approaches lead to a focus on “narrow utilitarianism” (counting day-to-day pleasures and pains) rather than what he thinks is more important: “flourishing” (the overall sense of meaning in your life).
Couldn’t you simply include “flourishing” as just another “benefit” in your cost-benefit analysis? (“I estimate the NPV of additional lifetime flourishing for having one more child to be the equivalent of $675,000 cash today”).
Roberts rejects this idea. He sees flourishing as something that is fundamentally unquantifiable, and not something you should think about trading off against day-to-day pleasures and pains.
But there’s also another reason why Roberts rejects the quantification of wild problems.
The Vampire Problem - Everyone Who Drinks Blood Says They Love It
Imagine that Dracula came up to you and offered you the chance to permanently turn into a vampire. You’ll get the chance to be immortal and look awesome in black. However, you’ll have to spend your nights searching for innocent victims to suck blood from, and also won’t ever be able to go to a sunny beach again. Would you take up Dracula’s offer?
Perhaps you’re leaning towards no. But Dracula has come prepared with evidence. He pulls out his laptop, its screen already open to a web page from a prestigious polling firm. “Poll: 98% of vampires enjoy drinking blood”, reads the headline. You read further. “In a nationally representative poll of 378 vampires aged between 18-623, 370 reported they had no regrets about becoming vampires.” Intrigued, you scroll down to the methodology section. Aside from the mysterious disappearance of one of the researchers, it seems flawless.
And yet, you’re not convinced. Vampires may report that they enjoy drinking blood, but how much credence should you give to the opinion of a vampire?
This is a thought experiment created by philosopher L.A. Paul, expanded in her book Transformative Experience, and also discussed in her EconTalk episode with Roberts. Astute readers who always read through every link in a post will notice that Roberts invoked this concept in his debate with Julia Galef. And it makes up a core aspect of Roberts’ argument in this book: that “wild problems” often involve decisions that fundamentally change who we are. In this analogy, becoming a parent (or making other major life decisions) is analogous to becoming a vampire - you’re turning into someone whose life and preferences are quite alien to the person you once were. As such, trying to model them is somewhat of a fool’s errand.
But Is Roberts Right? What Lessons Should Rationalists Draw?
I reviewed this book precisely because as an economics-degree holder and a self-identified rationalist, I found myself in instinctual disagreement with the book’s thesis. What do you mean, you shouldn’t measure some things?
But it’s for that reason I think it’s an important contribution to the debate over the applicability of “applied rationality”. Roberts is no crank that thinks numbers are stupid. He’s well-versed in the theory of rationality and thinks there are many appropriate uses for it. He just thinks some people go too far. Well, perhaps they do. But perhaps Roberts goes too far himself?
There’s an idea, perhaps best encapsulated in Scott Alexander’s post Should You Reverse Any Advice You Hear? that some advice can be good advice for people towards one end of a spectrum, but bad for people at the other end of that spectrum. “Be more selfish” might be good advice for a self-flagellating hardcore altruist, but it’s probably bad advice for Bernie Madoff.
I think Wild Problems’ advice is probably good on the margin for the sort of person, like myself, who instinctually disagrees with its thesis and secretly feels like they could make every life decision correctly with an app that predicts what goals you should have. “Be a little less quantitive” is probably good advice for such people. Don’t fall for the false precision of toy models. Live a little. Go with your gut (or heart) sometimes, and not just because Kahneman said there were legitimate uses for System 1 thinking.
But is Roberts ultimately correct on his own margin? That’s where I find it hard to agree.
Take Robert’s analysis of Charles Darwin’s dilemma in deciding whether to marry. As the main focus of the book’s second chapter, it serves as perhaps his flagship case study. (Interestingly, Julia Galef also made a major case study of Darwin in her book The Scout Mindset. Sounds like this Darwin guy might have some good ideas).
Roberts talks about how Darwin makes up a pro-con list of marriage (you can read the whole thing here) that seems to have a lot more cons (“loss of time”, “fatness and idleness”) than pros (“charms of music and female chit-chat”). Nonetheless, Darwin eventually decides to “Marry—Mary [sic]—Marry Q.E.D”. Roberts concludes that the pro-con list wasn’t truly decisive in Darwin’s decision - at best, the process helped steer him into realising what was truly meaningful for him.
But is Darwin really a good example? According to Roberts, Darwin had never been in a romantic relationship before. The “pro” he writes of a wife being an “o__bject to be beloved & played with. — better than a dog anyhow”) is funny because it comes from a man profoundly ignorant of women. Asking him to predict how marriage will go is like asking five-year-olds to predict what career they will be happiest in. The experience is so alien to them that they just can’t give meaningful answers.
In fairness to Darwin, the social mores of his day probably didn’t allow him to casually date before deciding on marriage. But in Western societies today, that’s very much an option. So are other forms of exploration and self-discovery - before embarking on a career, you can figure out what you’re likely to be good at (did you do better in English class or woodworking?) and try a few different jobs out to see what you think of them. Indeed, that’s something 80,000 hours, an EA career advice organisation, recommends.
But isn’t exploration (to the extent he recommends anything) something Roberts recommends? Sure. It’s good advice. But it can also make your wild problems a fair bit less wild. You might not truly understand what it’d be like to have kids from babysitting a bunch of times. But you probably have a better sense of it than you do of what it’s like to be a vampire.
And there’s another thing that bugs me. Roberts makes a big deal out of the idea that you shouldn’t trade off “flourishing” for “narrow utilitarian” values. Except he concedes that he himself would make this tradeoff in certain circumstances. For instance, he says he wouldn’t have made his move to become President of Shalem College in Jeruselum (which he said was a calling for him) if it had resulted in financial hardship for his family, or been located in a country that wasn’t the Jewish homeland. He calls these “exceptions that prove the rule”, but that seems rather hand-wavey to me. Doesn’t that mean there might be some level of hardship that might have been just enough to make Roberts decide not to go? Or an alternative job that was only a partial calling? Or some combination thereof? What would stop us from constructing indifference curves for Roberts?
Maybe there’s a way to accommodate this concept in a way that allows for trade-offs that still differentiates itself from standard economic analysis. But I don’t feel that Roberts has quite given us this.
I finished Wild Problems feeling a little bit dissatisfied, with questions like the above rattling through my head. The book is engaging, and I think Roberts gestures at something important and often overlooked in rationalist circles. But the case he makes is so strong that it’s hard to agree with him, even when I feel like he might be directionally correct.
No matter though. We’ll settle it once and for all with an experiment. I will assign 1,000 people at random to read Wild Problems and another 1,000 to read Algorithms to Live By. Each group will promise to try to adhere to the book’s lessons as best they can over the next 10 years. After that, I’ll ask everyone to rate their happiness on a 1-5 scale. Roberts couldn’t possibly fail to be persuaded by that.