What is an Explanation?

The last three posts were a series of investigations into the Golden Handcuffs phenomenon. First, a review of existing arguments and dismissal. Then, a series of ground-up empirical estimates. And finally, crowdsourced comments from readers. [1]

In the end, I conclude that it’s largely visa issues and selection effects.

Now we ask, is the distinction actually important?

If Golden Handcuffs are not about Google [2] bribing its employees with kombucha, but still prevent engineers from leaving corporate jobs to pursue their dreams, why should we care what the specific mechanism is?

For years, I have listened to people lament the “brain drain” out of socially important fields into tech. Supposedly, if Google didn’t pay such high wages and offer such good perks, the geniuses who work there would instead be doing renewable energy research.

But if it’s all selection effects and visa issues, this isn’t true at all! If Google engineers are just people who were not going to do interesting things to begin with, then it doesn’t matter at all to the rest of society. Their alternative was not solar panels, it was finance and consulting.

Note that this is only true for that particular purpose. If you’re dating around and looking for someone with a stable income, a Google engineer is still a great choice. It doesn’t matter one lick if they’re stuck at Google because they’re addicted to kombucha or because they’re an intrinsically risk-averse person. Either way they’re not leaving.

So the language we use does have this kind of fuzzy “by what definition” feature to it, but that doesn’t mean it’s arbitrary.

Postmodernism was about the context-dependence of facts [3]. Who’s asking matters as much as the question itself. If you’re trying to study The Eastern World it matters very much if you’re an insider. Some judgements will be wrong, others merely fantastic anachronisms that are simply irrelevant to how someone sees themselves.

In contrast, meta-rationality is about purpose-dependence. Rather than fixate on who is posing the question, meta-rationality asks why they want to know, and then provides an answer that is actually useful.

Is that too fanciful? Let me put it this way: if you’re asking purely out of intellectual curiosity, any number of things can be true. Is minimum wage bad? Good? It depends on what you want. Are you minimizing short term unemployment? Maximizing long term economic growth? Are you a policy maker, or a voter? If so, then you should be asking about minimum wage in a very particular context, not the phenomenon in general. If not, then why does any of this matter, and what kind of answer would even be satisfying?

To be clear, this isn’t about “skin in the game” in the sense of investing as a dojo for rationality. Betting on your beliefs is a fine exercise for proving that you can think clearly, but it also means you only get to think about bettable claims. At first that seems like a pretty large universe, but by design, it systematically excludes any domain where you might have an actual impact. Do you have a strong belief about UBI as an economic stimulus? It doesn’t matter because you won’t get to test it. And if you do make it your life’s mission to implement UBI, you’re now insider trading.

Purpose-dependence is about clarifying why you’re asking in the first place, figuring out which facts matter for that purpose, and using language in a way that’s useful for discovering those facts.

Accordingly, if you’re a reader skimming blogs out of pure intellectual curiosity, it is very much worth asking why.

[1] These map onto three popular ways of approaching questions. You can either consider common knowledge, reason from first principles, or ask broadly.

[2] If you read the previous posts closely, you might ask “how can there be Golden Handcuffs at Google when the average tenure is just 3 years?” The answer is that I’m using “Google” as shorthand for “big tech companies”.

[3] Words can mean different things in different contexts, which means first, you shouldn’t trust anything too much and second, it’s critical to ensure that language is used in a beneficial manner. And so you get theorists applying the Sapir–Whorf hypothesis to everything, and trying to rename everything (minority to minoritized, homeless to experiencing homelessness, etc). This feels absurd from the outside, but once you give up on language as a fixed thing, it makes sense to leverage it in a way that suits your goals.