Stop waxing philosophical

Waxing philosophical and its features

Maybe I’m just a narrow-minded bag of spite but as a philosopher there’s nothing I hate more than the moment when people who aren’t philosophers decide to wax philosophical.

Don’t get me wrong, I love it when non-philosophers think about and do philosophy. However, there’s a particular mode of “engagement” with the discipline that I hate. Its pernicious features are:

1. Deflecting from hard thinking about specifics and domain knowledge

2. Detaching from anything with empirical or practical content often at the moment these things are most needed.

3. Keying on words and concepts with no shared understanding. Many of these words, often gestures towards complexity and heterogeneity, are coded “good” while others, often markers of simplicity and homogeneity, are coded “bad”.

4. Waxing philosophical is a real favourite of bigwigs and effectively a status flex. “Being allowed to wax philosophical” is often a sign you’ve “made it”, and it can also serve a protective function allowing those of high status to avoid specific commitments.

5. Fundamentally, it is a sign that the topic is not being taken seriously, even if the speaker themselves doesn’t realise it. With some exceptions, people don’t tend to “wax philosophical” about topics they’re genuinely anxious to get right.

The first two features are almost always present. The third feature usually is present. The last two are often absent but tend to travel with the pack as a function of who has most use of these tricks, and who we permit to use them.

Examples

Some examples:

The editor of a journal is giving a keynote talk. They touch on some recent results on the unreliability of peer review and proposals for change. Across many examples and with much waffling they effectively give the line: “but what does quality even mean, really?” Not as a prelude to an answer, followed by sketching out implementation steps. Rather they say it as reassurance that it’s all terribly difficult to think about but there’s probably nothing to worry about.

Someone on Substack is asked a question about the economic impacts of AI. Rather than answering it they start talking about whether AI is “truly intelligent” referencing ideas that look a bit like Searle’s Chinese Room argument, plus an argument that AI can never be “truly creative” because it recombines things. Stochastic parrots and so on are invoked. All this would be fine if there were specific empirical claims that haven’t yet been falsified being made. However, no specific predictions regarding the task limitations of AI are given.

A senior advisor to the minister is quizzed on some technical points of a policy which, while tightly defined, might have enormous leverage. Rather than engaging he invokes a set of vaguely management adjacent ideas, talks about policy defensibility, and invokes strategy aligned language like “the government’s core priorities”. [I recognise the need for realism here: we have put very strong expectations on the minister’s assistant to not say anything too stupid. He’s just rationally responding to those incentives. It doesn’t make this pattern less dangerous though.]

Someone has a concrete and quite compelling case they’ve been treated unfairly. The response, from those accused of unfair treatment is, hey, haven’t many people been treated unfairly? Isn’t the world unfair? Aren’t the causes all very complex?

Another field bigwig is giving a keynote in, say, psychology. He gives a series of box-and-arrow-diagrams between abstractions, relating to say, methodology in biological psychology. None of it is new but the bigger problem is that none of it allows anyone to sink their teeth into reality.

A self-appointed methodologist of social science talks at length about emergence, holism, embodiedness, reflexivity and so on. No specific claims are made about what works except perhaps that we should be more emergentist, holistic, embodied and reflexive. Kuhn’s probably going to be in there somewhere. Stretch goal is a mention of “ways of knowing”, “plural knowledges” etc. Sophistication is both the validation metric and the goal.

The physical scientist talks about Popper. We’ll just leave it at that- don’t even get me started on this.

“When you think about it, experts have often been wrong, who is to say vibes aren’t better”. Given as an argument with no specific theory about why the experts might be wrong in this case, no clear articulation of inductive features associated with being right and wrong among experts, no clear expression of the rival theoretical basis to expertise in this case etc.

Part of the danger here is that the reflex often clicks in precisely when it shouldn’t. It clicks in when the stakes are high but it’s hard to see the good or bad outcomes. If our editor at a journal was accused of fraud, sexism, collusion or provable incompetence in who he published, he’d take it very seriously. But because he’s giving a talk on tracking quality, he doesn’t care. This is because it’s really hard to prove you’ve fucked selection quality up and even if you have so has everyone else. When people are talking about the future- e.g. with my large language model example- they often turn to philosophising. This is because they can only be held accountable for their mistakes here in the future which is A) time discounted and B) people will have usually forgotten you said it by then anyway.

How to avoid it

In sum waxing philosophical is bad because 1) It’s typically bad philosophy 2) Given when philosophy isn’t what’s called for. Philosophy is helpful for almost any complex task but doing high level philosophy instead of what’s in front of you is far from helpful.

How to avoid waxing philosophical? Generally speaking:

1. Drawing a lot out of a small number of abstract principles is a bit like drawing blood out of a stone. It can be a sign of a great miracle - some of the greatest ideas in the canon are like this - or it can be a sign of fraud.

2. If you are talking about an empirical topic, your argument should generally have implications for what we will observe, and the more specific the better. If you’re talking about a practical topic, your argument should generally have implications for what we should do. These implications should genuinely follow from what you’ve said and not just be shoehorned on.

3. Look for the dance. Generally, good reasoning that uses abstractions moves back and forth between the concrete and the abstract.

4. You should keep track of how vague your terms are. There are no clear rules here. All terms are vague: there are no practically discoverable necessary and sufficient conditions for anything except stipulated terms outside mathematics is about as close to a substantive and established result as anything philosophy has ever proved. Adding little potted definitions generally won’t help. The “Webster’s dictionary defines…” guy is reviled for a reason. But try to keep an eye on your vagaries and clarify where necessary.

5. Break the status reflex where people are allowed to wax philosophical if they’re high enough status. As an audience member, you should ask yourself “would I tolerate this level of high-falutin abstraction and manner of thought from a postdoc, less famous writer, or less senior personage”. If the answer is no, that’s at least a sign that you shouldn’t be tolerating it from the more senior personage. Yes, in certain contexts there might sometimes be good reasons to allow greater abstraction from the more senior figures. Vague impressions about what works and what doesn’t might be more valuable from someone with a lifetime of doing things. For the most part, however, symmetry is the best rule. Sometimes, to be sure, you should resolve the inconsistency in the other direction and develop greater tolerance towards high-flautin rhetoric from the postdoc, the less famous writer or the less senior personage.

6. You should be clear about the marginal claim you are making - what is distinct about your position or argument. A lot of the time there’s not going to be anything strictly original in your argument but that doesn’t mean it lacks a marginal contribution “we all know X, but I don’t think we’ve fully metabolised its implications”, “position Y has been unjustly neglected” are all perfectly valid marginal contributions as is “I don’t think enough people understand Z, so I’m explaining it”.

7. Recognise that, unfortunately, the undesirable pattern I am speaking of is vague and hard to pin. There are plenty of things that look superficially like they have the features of waxing philosophical but are intellectually important. Picking an example at random, the elaboration of the various classes of construct reliability in early psychometrics could have been skewered as vague pseudo philosophy by a cruel critic at the time (I’m sure some operationalist rear guarder did exactly this). However, they weren’t pseudo philosophy, they were incredibly important and concrete contributions to methodology. My own PhD thesis has a section on Babylonian versus Euclidean methodology, and I think it’s loadbearing enough to earn its place. The finest test is “does it do something” but do and something are vague terms here. Ultimately you just have to git gud.

8. Also, confusingly, there’s an opposite failure mode where you avoid dealing with the real questions by bogging down in irrelevant concreta. In this failure mode you erroneously focus on considerations that answer a question very well and specifically but the wrong question. Good luck out there!

Discussions around AI are particularly bad when it comes to this sort of thing, they formed one of my examples but could have formed dozens. In working out what AI is going to do, it is vital not to substitute philosophy of mind or language for concrete thought about the architecture, capacities and trajectory of large language models. There are critics out there- I won’t name them, but people connected to the field will know who I’m talking about- who have predicted on more or less a priori grounds that large language models will never do anything interesting. This might have been borderline acceptable in 2022 before Language Models are Unsupervised Multitask Learners came out. Since then, large language models have been able to do things moreover they have been able to do more and more over time. “I am making a concrete prediction that large language models will never X for reasons Y” is a totally acceptable thing to say. “I don’t think they’re going anywhere for vaguely Chinese room adjacent reasons” is nonsense, they’ve already done useful things so they’ve already gone somewhere. The interesting question is what well-specified bounds there are on their capacities- trying to draw the perimeter of what they will and will never be able to do. Some of these critics even make a point of not using the models.

I’d also say waxing philosophical can be an incredibly dangerous thing to do on a personal level, although it’s hard to prove. How many times in a movie does the protagonist justify some pivotal choice with a string of pseudo-profound bullshit? Some deepity or maxim that sounds nice for two seconds, if that. Indeed, the person who tries to think hard about important choices is satirised. I believe, though I cannot prove, that this disastrous approach is not just present on the silver screen, but has become ingrained in the habits of many.

It’s the end of the article so it’s time for the conflict-of-interest declaration. As a philosopher, I must concede a conflict. Of course, it bothers me that what I have trained my entire life to do is treated as an evasive manoeuvre, a bit of frisson, or a sophistication signal. It bothers me even more that the market for philosophy is flooded with this junk. Am I then just an angry cartel member pushing for occupational licencing through social censure? No, I don’t think so. I love it when people do philosophy. Properly, carefully, not jumping from similar vibe to similar vibe, not as a low effort flourish and not as a replacement for trying to improve our world.