Deflationism isn’t the solution to philosophy’s woes

[epistemic status: thinking out loud; reporting high-level impressions based on a decent amount of data, but my impressions might shift quickly if someone proposed a promising new causal story of what’s going on]

[context warning: If you’re a philosopher whose first encounter with LessWrong happens to be this post, you’ll probably be very confused and put off by my suggestion that LW outperforms analytic philosophy.

To that, all I can really say without starting a very long conversation is: the typical Internet community that compares itself favorably to an academic field will obviously be crazy or stupid. And yet academic fields can be dysfunctional, and low-hanging fruit can sometimes go unplucked for quite a while; so while it makes sense to have a low prior probability on this kind of thing, this kind of claim can be true, and it’s important to be able to update toward it (and talk about it) in cases where it does turn out to be true.

There are about 6700 US philosophy faculty, versus about 6000 LessWrong commenters to date; but the philosophy faculty are doing this as their day job, while the LessWrong users are almost all doing it in their off time. So the claim that LW outperforms is prima facie interesting, and warrants some explanation.

OK, enough disclaimers.]

A month ago, Chana Messinger said:

Rob says, “as an analytic philosopher, I vote for resolving this disagreement by coining different terms with stipulated meanings.”

But I constantly hear people complain that philosophers are failing to distinguish different things they mean by words and if they just disambiguated, so many philosophical issues would be solved, most recently from Sam and Spencer on Spencer’s podcast.

What’s going on here? Are philosophers good at this or bad at this? Would disambiguation clear up philosophical disputes?

My cards on the table: I understand analytic philosophers to be very into clearly defining their terms, and a lot of what happens in academic philosophy is arguments about which definitions capture which intuitions or have what properties, and how much, but I’m very curious to find out if that’s wrong.

Sam Rosen replied:

Philosophers are good at coming up with distinctions. They are not good at saying, “the debate about the true meaning of knowledge is inherently silly; let’s collaboratively map out concept space instead.”

An edited version of my reply to Chana and Sam:

Alternative hypothesis: philosophers are OK at saying ‘this debate is unimportant’; but...

  • (a) … if that’s your whole view, there’s not much to say about it.

    Sometimes, philosophers do convince the whole field in one fell swoop. A Bertrand Russell comes along and closes the door on a lot of disputes, and future generations just don’t hear about them anymore.

    But if you fail to convince enough of your colleagues, then the people who think this is important will just keep publishing about it, while the people who think the debate is dumb will roll their eyes and work on something else. I think philosophers in a given subfield tend to think that a large number of the disputes in other subfields are silly and/​or unimportant.

  • (b) … there’s a culture of being relaxed, or something to that effect, in philosophy?

    Philosophical fields are fine with playing around with cute conceptual questions, and largely feel no need to move on to more important things when someone gives a kinda-compelling argument for ‘this is unimportant’.

    Prominent 20th-century philosophers like David Lewis and Peter van Inwagen acquired a lot of their positive reputation from the fact that all their colleagues agreed that their view was obviously silly and stupid, but there was some disagreement and subtlety in saying why they were wrong, and they proved to be a useful foil for a lot of alternative views. Philosophers don’t get nerd-sniped from their more important work; nerd-sniping just is the effective measure of philosophical importance.

We’ve still ended up with a large literature of philosophers arguing that this or that philosophical dispute is non-substantive.

There’s a dizzying variety of different words used for a dizzying variety of different positions to the effect of ‘this isn’t important’ and/​or ‘this isn’t real’.

There are massive literatures drawing out the fine distinctions between different deflationary vs. anti-realist vs. nominalist vs. nihilist vs. reductionist vs. eliminativist vs. skeptical vs. fictionalist vs. … variants of positions.

Thousands of pages have been written on ‘what makes a dispute merely verbal, vs. substantive? and how do we tell the difference?‘. Thousands of journal articles cite Goodman’s ‘grue and bleen’ (and others discuss Hirsch’s ‘incar and outcar’, etc.) as classic encapsulations of the problem ‘when are concepts joint-carving, and when are words poorly-fitted to the physical world’s natural clusters?’. And then there’s the legendary “Holes,” written by analytic philosophers for analytic philosophers, satirizing and distilling the well-known rhythm of philosophical debates about which things are fundamental or real vs. derived or illusory.

It’s obviously not that philosophers have never heard of ‘what if this dispute isn’t substantive?? what if it’s merely verbal??’.

They hear about this constantly. This is one of the most basic and common things they argue about. Analytic philosophers sometimes seem to be trying to one-up each other about how deflationary and anti-realist they can be. (See “the picture of reality as an amorphous lump”.) Other times, they seem to relish contrarian opportunities to show how metaphysically promiscuous they can be.

I do think LW strikingly outperforms analytic philosophy. But the reason is definitely not ‘analytic philosophers have literally never considered being more deflationary’.

Arguably the big story of 20th-century analytic philosophy is precisely ‘folks like the logical positivists and behaviorists and Quineans and ordinary language philosophers express tons of skepticism about whether all these philosophical disputes are substantive, and they end up dominating the landscape for many decades, until in the 1980s the intellectual tide starts turning around’.

Notably, I think the tide was right to turn around. I think mid-20th-century philosophers’ skepticism (even though it touched on some very LW-y themes!) was coming from a correct place on an intuitive level, but their arguments for rejecting metaphysics were total crap. I consider it a healthy development that philosophy stopped prejudicially rejecting all ‘unsciencey’ things, and started demanding better arguments.

Why does LW outperform analytic philosophy? (Both in terms of having some individuals who have made surprisingly large progress on traditional philosophical questions; and in terms of the community as a whole successfully ending up with a better baseline set of positions and heuristics than you see in analytic philosophy? Taking into account that LW is putting relatively few person-hours into philosophy, many LWers lack formal training in philosophy, etc.)

I suspect it’s a few subtler differences.

  • Something to protect” is very much in the water here. It’s normal and OK to actually care in your bones about figuring out which topics are unimportant—care in a tangible “lives are on the line” sort of way—and to avoid those.

    No one will look at you funny if you make big unusual changes to your life to translate your ideas into practice. If you’re making ethics a focus area, you’re expected to actually get better results, and if you don’t, it’s not just a cute self-deprecating story to tell at dinner parties.

  • LW has a culture of ambition, audacity, and ‘rudeness’, and historically (going back to Eliezer’s sequence posts) there’s been an established norm of ‘it’s socially OK to dive super deep into philosophical debates’ and ‘it’s socially OK to totally dismiss and belittle philosophical debates when they seem silly to you’.

    I… can’t think of another example of a vibrant intellectual community in the last century that made both of those moves ‘OK’? And I think this is a pretty damned important combination. You need both moves to be fully available.

  • Likewise, LW has a culture of ‘we love systematicity and grand Theories of Everything!’ combined with the high level of skepticism and fox-ishness encouraged in modern science.

    There are innumerable communities that have one or the other, but I think the magic comes from the combination of the two, which can keep a community from flanderizing in one direction or the other.

  • More specifically, LWers are very into Bayesianism, and this actually matters a hell of a lot.

    E.g., I think the lack of a background ‘all knowledge requires thermodynamic work’ model in the field explains the popularity of epiphenomenalism-like views in philosophy of mind.

    And again, there are plenty of Bayesians in academic philosophy. There’s even Good and Real, the philosophy book that independently discovered many of the core ideas in the sequences. But the philosophers of mind mostly don’t study epistemology in depth, and there isn’t a critical mass of ‘enough Bayesians in analytic philosophy that they can just talk to each other and build larger edifices everywhere without constantly having to return to 101-level questions about why Bayes is good’.

  • This maybe points at an underlying reason that academic philosophy hasn’t converged on more right answers: some of those answers require more technical ability than is typically expected in analytic philosophy. So when someone publishes an argument that’s pretty conclusive, but requires strong technical understanding and well-honed formal intuitions, it’s a lot more likely the argument will go ignored, or will take decades (rather than months) to change minds. More subtly, the kinds of questions and interests that shape the field are ones that are (or seem!!) easier to tackle without technical intuitions and tools.

    Ten years ago, Marcus Hutter made a focused effort to bring philosophers up to speed on Solomonoff induction and AIXI. But his paper has only been cited 96 times (including self-citations and citations by EAs and non-philosophers), while Schaffer’s 2010 paper on whether wholes are metaphysically prior to their parts has racked up 808 citations. This seems to reflect a clear blind spot.

  • A meta-explanation: LW was founded by damned good thinkers like Eliezer, Anna, Luke M, and Scott who (a) had lots of freedom to build a new culture from scratch (since they were just casually sharing thoughts with other readers of the same blog, not trying to win games within academia’s existing norms), and (b) were smart enough to pick a pretty damned good mix of norms.

    I don’t think it’s a coincidence that all these good things came together at once. I think there was deliberate reflection about what good thinking-norms and discussion-norms look like, and I think this reflection paid off in spades.

    I think you can get an awful lot of the way toward understanding the discrepancy by just positing that communities try to emulate their heroes, and Anna is a better hero than Leibniz or Kant (if only by virtue of being more recent and therefore being able to build on better edifices of knowledge), and unlike most recent philosophical heroes, LW’s heroes were irreverent and status-blind enough to create something closer to a clean break with the errors of past philosophy, keeping the good while thoroughly shunning and stigmatizing the clearly-bad stuff. Otherwise it’s too easy for any community that drinks deeply of the good stuff in analytic philosophy to end up imbibing the bad memes too, and recapitulate the things that make analytic philosophy miss the mark pretty often.

Weirdly, when I imagine interventions that could help philosophy along, I feel like philosophy’s mild academic style gets in the way?

When I think about why LW was able to quickly update toward good decision-theory methods and views, I think of posts like “Newcomb’s Problem and Regret of Rationality” that sort of served as a kick in the pants, an emotional reminder “hold on, this line of thinking is totally bonkers.” The shortness and informality is good, not just for helping system 1 sit up and pay attention, but for encouraging focus on a simple stand-alone argument that’s agnostic to the extra theory and details you could then tack on.

Absent some carefully aimed kicks in the pants, people are mostly happy and content to stick with the easy, cognitively natural grooves human minds find themselves falling into.

Of course, if you just dial up emotional kicks in the pants to 11, you end up with Twitter culture, not LW. So this seems like another smart-founder effect to me: it’s important that smart self-aware people chose very specific things to carefully and judiciously kick each other in the pants over.

(The fact that LW is a small community surely helps when it comes to not being Twitter. Larger communities are more vulnerable to ideas getting watered down and/​or viral-ized.)

Compare Eliezer’s comically uncomplicated “RATIONALISTS SHOULD WIN” argument to the mild-mannered analytic-philosophy version.

(Which covers a lot of other interesting topics! But it’s not clear to me that this has caused a course-correction yet. And the field’s course-correction should have occurred in 2008–2009, at the latest, not 2018.)

(Also, I hear that the latter paper was written by someone socially adjacent to the rationalists? And they cite MIRI papers. So I guess this progress also might not have happened without LW.)

(Also, Greene’s paper of course isn’t the first example of an analytic philosopher calling for something like “success-first decision theory”. As the paper notes, this tradition has a long history. I’m not concerned with priority here; my point in comparing Greene’s paper to Eliezer’s blog post is to speak to the sociological question of why, in this case, a community of professionals is converging on truth so much more slowly than a community of mostly-hobbyists.)

My story is sort of a Thiel-style capitalist account. It was hard to get your philosophy published and widely read/​discussed except via academia. But academia had a lot of dysfunction that made it hard to innovate and change minds within that bad system.

The Internet and blogging made it much easier to compete with philosophers; a mountain of different blogs popped up; one happened to have a few unusually good founders; and once their stuff was out there and could compete, a lot of smart people realized it made more sense.

LW is academic philosophy, rebooted with better people than Plato as its Pater Patriae.