(This was originally a thread on mastodon. The discussion there led me to extend the thread into this post. Thanks to all involved.)
We’ve all had to suffer their company at one time or another. If you’ve worked in tech, then that suffering is almost a daily gauntlet of irritation, a background radiation of hyper-confident babble that regularly threatens to drown out everything else.
You know the type: the sort of person who seems to alternate between surprisingly acute observations and blatant idiocy, both with equal confidence. The Schroedinger’s cat of intelligence, who simultaneously inhabits the two, incompatible states of being both extremely smart and extremely not so, both at the same time.
You can ignore these types on social media but, in the workplace, their confidence and blind spots can easily become a major liability.
The core reason why they can be harmful is that they are playacting genius. Their style of reasoning and presentation is performative and misrepresents both their own accomplishments (which are often non-trivial) and the validity of their methodology (which is where it can get harmful).
Playacting genius
The first confusion is that a lot of the time what they call “reasoning from first principles” is more accurately “being a fast learning autodidact”. Going from zero to expertise in six months is not in any real way discovering a field by reasoning from first principles. They’ve just speedrun a field from a logical starting point. This is a major accomplishment, one that is enormously useful in the right context, as long as you are aware that this is what you did.
Anybody can reason from first principles or basic truths (if you are a regressive Aristotelian intent on ignoring all progress in human thought over the past two millennia). It’s generally useless because for any given field you just end up running into a series of ‘turkey’ problems on a compressed timeline without the benefit of history.
The turkey problem
As presented by N.N. Taleb, if I recall correctly, the turkey problem is as follows: a turkey who has never experienced Thanksgiving thinks it lives in a paradise managed by a benevolent god. It has observed the basic truths of its existence, the first principles of both social and natural order and come to the inescapable conclusion that humans are benevolent and kind and have the turkey’s best interests at heart. There is nothing observable in its environment, history, or culture that could lead it to discover from first principles that they are being bred for eatin'.
The event that will give them a true understanding of the nature of their existence lies in the future and can’t be predicted from observation or first principles. The turkey who distrusts its environment and the caretaker is being irrational.
The history of almost every non-trivial field is full of dead ends and bad ideas that were impossible to predict before the fact. Genuinely reasoning about the field from first principles will regularly lead you into one of these dead ends.
This is why that annoying coworker alternates between interesting observation and inane idiocies. Their speedrun through various fields sometimes leads them into interesting corners and sometimes into dead ends. Your problem is that you don’t know whether or not their interesting observations are dead ends that you simply don’t recognise because it’s in a field unfamiliar to you.
This is also why it isn’t uncommon for people who genuinely engage in reasoning from first principles to end up as cranks or conspiracy theorists. You need a justification of why your intellectual dead end isn’t actually a dead end to assuage your cognitive dissonance. This is a recipe for falling for all sorts of nonsense pseudo-medicine.
Fast learners
Most people who claim to be reasoning from first principles are just really fast learners. They are regularly capable of referencing texts in the field to back up their opinion, which is how they avoided the turkey barriers, and isn’t something you’d be able to do if you stuck to pure deductive reasoning. They do read a field’s texts (unlike what you would with genuine first-principles reasoning) but they read selectively.
What these dudes (generally dudes) usually mean by ‘reasoning from first principles,’ is they use those first principles to guide their reading and learning: observe what look like basic truths in a field and use that to choose your reading. Couple that with a decent capacity for memorisation and you can quickly get to a surprising level of expertise. You pick and choose your reading and research based on what is, honestly, half-baked reasoning based on observations that you’re gambling aren’t just baked-in personal biases.
The problem here is that this only works if you’re a quick reader and if you have the semantic memory capacity of your average twenty-something. Raw reasoning has precious little to do with it.
The other big problem is blind spots.
Those blind spots
This is the big issue with this crowd. It’s what leads otherwise accomplished programmers to be unaware of the halting problem or P=NP and leads them to build systems that are rife with unsolvable problems and architectural flaws.
It also frequently leads them to be toxic managers who have picked up a narrow range of good ideas and practices but missed a bunch of abuse and dysfunction dynamics because that’s exactly the sort of stuff you can only learn from history.
Abuse dynamics are an emergent characteristic of a complex system. That’s why the same behaviour can be abusive in one context but harmless in another. The methods that work for preventing abuse in the workplace and for properly handling the abuse you can’t prevent have been developed from practice and experience, and are generally impossible to discover from first-principles deduction.
Paradoxically, these practices are generally discovered and established through adversity and conflict. Most of the progress we’ve seen over the past century in minimising workplace abuse of all kinds has been through collective action: where those with less power in a workplace band together to establish boundaries for those who have power.
Reducing abuse by deliberately being more adversarial.
So, picking up a new field can be done quickly if you use a bit of forethought on which first principles to guide your reading and research but you need to be aware that this approach is fundamentally risky and can lead you to have a bunch of blind spots.
And, boy howdy, can those blindspots be bad. They’re exactly the sort of nonsense that held scholarly thought back for centuries.
Aristotle
—“But Aristotle reasoned from first principles. He was one of the greatest philosophers in history, so you must be wrong!"
Yeah, sorry, no. Aristotle…
- Was a dick.
- Was wrong about pretty much every observable phenomenon he had an opinion on.
- Was a massive dick.
Truly a dick of epic proportions. A massive pendulous cock that pissed all over scholarly practice for over two thousand years. Harmful at least in part because of the specific precedence he set with first-principles thinking:
That Aristotle espoused these views matters, because his imprimatur imbues them with authority and an air of dispassionate reason. Many readers of the “Politics” have concluded that Aristotle lays out first principles, the indisputable facts prescribed by nature, before reasoning from them to arrive at his political theories. If nature, including hierarchies and natural slavery, is simply a fact, then society can function properly only if it is ordered with this fact in mind.
Facts don’t care about feelings, after all. Aristotle therefore envisioned a hierarchical society, where everyone had their proper place, from fully enfranchised citizens all the way down to slaves. Given how nature has endowed different people differently, such a society would benefit all.
(From Aristotle, father of scientific racism, an article that then waters down its observations. It weasels out of fully condemning Aristotle by stating that, even though he supported slavery and thought it was natural, he also viewed nature as changeable, so a natural slave, in his terms, might stop being a natural slave over time. Which are just weasel-words to gloss over the fact that they’d still be slaves and Aristotle still supported slavery.)
It took academic thought centuries to recover from the malpractices of Aristotle and Plato.
Moreover, Aristotle was generally wrong about everything because he reasoned from first principles.
From an article that I linked to above, Aristotle Was Wrong—Very Wrong—But People Still Love Him:
What is the difference? The Greeks approached ideas with some common and assumed truth and then deduced the details from that. For instance, take a heavy object and a light object (like a rock and a feather) and drop them. Which will hit the ground first? I think we can all agree that the rock will fall faster than the feather. So, this is something we assume to be true and the rest of the details can be deduced from that.
This led to misguided notions on gravity that derailed us for centuries: first-principles thinking, yay!
The reason why this frequently leads people to disaster is that it’s quite hard to tell which observable phenomenon is a first principle (or basic truth), which is just a biased observation on your part, and which is an emergent property of a complex system you cannot observe and could change at any moment for reasons that are invisible to you.
We need a process to validate those first principles as an actual basic truth you can be reasonably confident about (say over 90% confidence).
That’s where the scientific method comes in.
The Scientific Method
—‘But first-principles thinking is how science works! It is the scientific method!!'
No, it’s not. If you think that then you’ve fallen for the performative nonsense of arrogant dicks playacting genius.
The scientific method is all about invalidating theses, not about deductive reasoning. It’s the process of proving all sorts of misguided nonsense wrong until you are left with stuff that may or may not be misguided nonsense but hasn’t been proved to be wrong yet, so is at least correct enough to maybe build on.
For example:
“Are ducks aliens?" is not a scientific thesis. It’s a nonsense question that, on its own, can’t be answered and, if you try to debate it, you’re just going to get spun around in circles. It’s too general, too vague, and not specific enough to be a useful scientific thesis.
However, “is the duck species Aix galericulata an alien phenomenon that evolved separately from other bird species on earth?" is a scientific question that you can test.
- Take genetic samples from a representative sample of the duck population
- Map their genome
- Compare that to another bird species like chickens or something. Chickens are birds, right?
The result should give you reasonable confidence (like, say 98%) that this particular duck species co-evolved with other bird species here on earth. With additional research, you might even be able to say that with 99% or even 99.5% confidence. Science!
It isn’t reasoned from any first principles, it’s a thesis born entirely out of “hey, wouldn’t it be funny if…" fridge logic, but it’s scientific and also, surprisingly, not entirely a nonsense thesis. It’s can be a useful question to try to answer.
But the point is that the question of the genetic relationship ducks have with other species on the planet and the question of whether gravity works the way we think it does are equally ‘scientific’ questions.
What they both have in common is that we are using the scientific method to see if the first principles we have assumed about the world around us are the basic truths we think they are.
In the case of gravity, checking our assumptions has led to some of the greatest advances in the history of science and engineering. Because it turns out that Aristotle was a dick who was fundamentally mistaken about how large parts of the natural world work.
Science is driven by a scepticism about first principles, not by first-principles thinking. Questioning your basic truths or first principles is usually much more interesting than building a pyramid of deductions on top of what may or may not just be a puddle of biases.
First principles aren’t the building blocks of genius. Unless they have been validated by research they are roadblocks in the way of science. Over time, those validations and the practice of those validations become a history and context that is the true foundation of science. This shit takes time but it’s so worth it. Because it lets you spot the stuff that looks funny.
‘That’s funny’
The most exciting phrase to hear in science, the one that heralds new discoveries, is not “Eureka!” (I found it!) but “That’s funny …”
No matter the actual provenance of the quote, it’s a central truth in scientific progress.
The history of innovations in science generally comes in one of two forms:
- Painstaking research builds our confidence and knowledge in a field inch by inch.
- Somebody notices something strange or out of place in an experiment they’ve witnessed.
Both of these depend on science as a practice and neither functions properly if you speedrun your way through the field. To recognise the odd and out-of-place, you need a familiarity with the place, the normal, the usual. Practice is a prerequisite to progress.
This isn’t the end of the story for scientific progress or the scientific method, which has been discussed and debated in detail over the years.
Although I do find Feyerabend’s Against Method charming, I’m more of a believer in Kuhn’s The Structure of Scientific Revolutions.
He argues that since worldviews and theories can only ever be partial explanations of scientific observations, as those observations accumulate, inconsistencies will mount until they make the initial worldview untenable.
Then once somebody comes up with a worldview that properly accommodates all of those observations a scientific revolution will take place. That then guides new research and new observations and the cycle repeats itself.
The hope is that each new worldview is a more accurate representation of the order of things than the previous one. What is or isn’t a first principle or basic truth will vary, change, and evolve with each successive worldview (or paradigm).
But, I digress. All you need to know is that an insistence on first-principles thinking is anti-progress and frequently regressive. It’s a system of reasoning that builds intricate structures on the baked-in assumptions of the dominant, orthodox worldview. Which is the reason why it regularly leads you to reactionary and conservative conclusions.
(I omitted a note here about how speedrunning into expertise in a field might have creative benefits based on Mihaly Csikszentmihalyi’s Creativity because 1. who cares? and 2. I can’t be arsed. and 3. Csikszentmihalyi has a tendency for anecdotal ‘just so’ woo-woo observations so he might not be the best reference for this particular essay.)
Software is dominated by dilettantes playacting genius
The popularity of first-principles thinking is a problem because of its enduring appeal to emotionally stunted programmers who were repeatedly praised during childhood for being the smartest person in the room.
This, in turn, is a problem because, to quote somebody who definitely thought he was being the smartest person in the room: software is eating the world.
If that software is being written and designed by people who are—because of an internet brainworm they encountered at a vulnerable stage in their life—oblivious to a wide range of serious problems and issues they are creating in the fields they are disintermediating…
…bad things happen.
You end up with social media sites that are used to coordinate hate, abuse, and even genocide. You get hosting providers defending paedophiles and neo-nazis.
You end up with a software industry that’s increasingly becoming a tool that serves various repressive, reactionary, and even fascistic politics all over the world.
So, the rest of us need to figure out ways to work around these people when they appear in our workplaces and mitigate the damage they do to the software we make. We need to make sure that expertise and an awareness of history are applied to minimise abuse and harm. We need to make sure that our software isn’t actually annihilating some preexisting field or industry we are only vaguely aware of.
In short, we need to make sure that their arrogance isn’t reflected in the work we do and that the software we make faces the world with a touch of humility.