I've been thinking about how we actually figure out what's true when trying to do the most good. Not how ideal utilitarians with perfect information would reason, but what actually works for humans trying to make high-stakes decisions under uncertainty.
The core argument: we already implicitly rank epistemic methods in our daily lives (we trust thermometers over theology, Google over gut feelings), but sometimes people pretend these hierarchies don't exist when theorizing about knowledge.
The point isn't the specific rankings but making our implicit hierarchies explicit. When we do this, we can better understand why some domains have more consensus than others, and why certain types of arguments are more persuasive. Further, I think my overall framework for incorporating different epistemic methods and "ways of knowing" is more adaptive and fluid and less likely to impose artificial top-down order on humanity's many disparate sources of knowledge. As a wise prince once said, "There are more things in heaven and earth, Horatio, than are dreamt of in your philosophy."
I think this type of reasoning is in many ways critical to the effective altruism project. Many discussions here implicitly involve weighing different types of evidence (from RCTs in global health and factory farming to theoretical arguments in AI safety and wild animal welfare), and this community overall favors systematization.
Hopefully, by having a detailed article that nudges people away from subtly (or not-so-subtly!) wrong frameworks into a framework that's more conducive to clearer thinking, we can make less mistakes in systematization and become more effective at cause and intervention prioritization.
Questions
Curious what people think about:
Which methods you'd rank differently and why
How you navigate domains where high-tier methods aren't available
Whether certain cause areas systematically rely on different tiers of evidence
What methods I've missed entirely
How you'd adapt my framework more explicitly for your preferred projects or cause areas
Would be interested in understanding why people downvoted this. I spent nontrivial effort conveying something quite subtle and difficult in a fair amount of detail, and I think ought to be helpful for improving a lot of people's thinking a little.
I didn't downvote or disagreevote, but I'm not sure the logic of the rankings is well explained. I get the idea that concepts in the lowest tiers are supposed to be of more limited value, but I'm not sure why the very top tiers are literacy/mathematics - seems like literacy/mathematics by themselves almost never point to any particular conclusions, but are merely prerequisites to using some other method to reach a decision. Is the argument that few people would dispute that literacy and mathematics should play some role in making decisions, where as the value of 'divine revelation' is hotly disputed and the validity of natural experiments debatable? That makes sense, but it feels like it needs more explanation.
Here are the arguments for each of the tiers, for reference and ease of quoting:
What Makes a Method Great?
What separates S-tier from F-tier? Three things: efficiency (how much truth per unit effort), reliability (how often and consistently it works), and track record (what has it actually accomplished). By efficiency, I mean bang-for-buck: literacy is ranked highly not just because it works, but because it delivers extraordinary returns on humanity's investment compared to, say, cultural evolution's millennia of trial and error through humanity’s history and pre-history.
A key component of this living methodology is what Taleb calls "Wittgenstein's ruler": when you measure a table with a ruler, you're learning about both the table and the ruler. Every time we use a method to learn about the world, we should ask: "How well did that work?" This constant calibration is how we build a reliable tier list.
S+ Tier: Literacy/Reading
The peak tool of human epistemology. Writing allows knowledge to accumulate across generations, enables precise communication, and creates external memory that doesn't degrade. Every other method on this list improved once we could write about it. Whether you’re reading an ancient tome, browsing the latest article on Google search, or carefully digesting a timeless essay on the world’s best Substack, the written word has much to offer you in efficiently transmitting the collected wisdom of generations. If you can only have access to one way of knowing, literacy is by far your best bet.
S Tier: Mathematical Modeling
Math allows you to model the world. This might sound obvious, but it is at heart a deep truth about our universe. From the simplest arithmetic that allows shepherds and humanity’s first tax collector to count sheep to the early geometrical relationships and calculations that allowed us to deduce that the Earth is round to sophisticated modern-day models in astrophysics, quantum mechanics, and high finance, mathematical models allow us to discover and predict the natural patterns of the world with absurd precision.
Further, mathematics, along with writing and record-keeping, allows States to impose their rigor on the chaos of the human world to build much of modern civilization, from the Babylonians to today.
A Tier: [Intentionally empty]
Nothing quite bridges the gap between humanity’s best tools above and the merely excellent tools below.
B Tier: Mimicry, Science, and Engineering
Three distinct but equally powerful approaches:
Mimicry: When you don't know how to cook, you watch someone cook. Heavily underrated by intellectuals. As Cate Hall argues in How To Be Instantly Better at Anything, mimicking successful people is one of the most successful ways to become better at your preferred task.
Ultimately, less accessible than reading (you need access to experts), less reliable than mathematics (you might copy inessential features), but often extremely effective, especially for practical skills and tacit knowledge that resists verbalization.
Science: Hypothesis-driven investigation.RCTs, controlled experiments, systematic observation. The strength is in isolation of variables and statistical power. The weakness is in artificial conditions and replication crises. Still, when done right, it's how we learned that germs cause disease and DNA carries heredity.
Engineering: Design under constraints. As Vincenti points out in What Engineers Know and How They Know It, many of our greatest engineering marvels were due to trial and error, where the most important prototypes and practical progress far predates the scientific theory that comes later. Thus, engineering should not be seen as merely "applied science": it's a distinct way of knowing. Engineers learn through building things that must work in the real world, with all its fine-grained details and trade-offs. Engineering knowledge is often embodied in designs, heuristics, and rules of thumb rather than theories. A bridge that stands for a century is its own kind of truth. Engineering epistemology gave us everything from Roman aqueducts to airplanes, often before science could explain precisely why it worked.
Scientific and engineering progress have arguably been a major source of the Enlightenment and the Industrial Revolution, and likely saved hundreds of millions if not billions of lives through engineering better vaccines and improved plumbing alone. So why do I only consider them to be B-tier techniques, given how effective they are? Ultimately, I think their value, while vast in absolute terms, are dwarfed by writing and mathematics, which were critical for civilization and man’s conquest over nature.
Wind tunnel of the Wright Brothers Source. The Wright brothers built their own wind tunnel and tested many designs through rigorous trial-and-error; despite then-scientifically flawed models of lift, they were able to achieve the first heavier-than-air flight. Airplanes were ultimately possible due to advancements in both science (Bernoulli's principle, Newton’s Laws) and practical engineering.
Solid tools with a somewhat more limited scope. Statistics help us see patterns in noise (and sometimes patterns that aren't there). Natural experiments let us learn from variations we didn't create. Both are powerful when used correctly, but somewhat limited in power and versatility compared to epistemic tools in the S and B tiers.
C Tier: Expert Intuition, Historical Analysis, Frameworks and Meta-Narratives, Forecasting/Prediction Markets
Often brilliant, often misleading. Experts develop good intuitions in narrow domains with clear feedback loops (chess grandmasters, firefighters). But expertise can easily become overwrought and yield little if any predictive value (as with much of political punditry). Historical patterns sometimes rhyme but often don't, and frequently our historical analysis becomes a Rorschach test for our pre-existing beliefs and desires.
I also put frameworks and meta-narratives (like Bayesianism, Popperism, naturalism, rationalism, idealism, postmodernism, and, well, this post’s framework) at roughly C-tier. Epistemological frameworks and meta-narratives1 refine thinking but aren’t the primary engines of discovery.
Finally, I put some of the more new-fangled epistemic tools (forecasting, prediction markets, epistemic betting in general, other new epistemic technologies) at roughly this tier. They show significant promise, but have a very limited track record to date2.
D Tier: Thought Experiments, Pure Logic, Introspection, Non-expert intuitions, debate.
Thought experiments clarify concepts you already understand but rarely discover new truths. Pure logic is only as good as your premises. Introspection tells you about your mind, not the world. Vastly overrated by people who think for a living.
In many situations, the philosophical equivalent of bringing a knife to a gunfight. Thought experiments can clarify concepts you already understand, but rarely discover new truths. They also frequently cause people to confuse themselves and others. Pure logic is only as good as your premises, and sometimes worse. Introspection tells you about your own mind, but the lack of external grounding again weakens any conclusions you can get out of it. Non-expert intuitions can be non-trivially truth-tracking, but are easily fooled by a wide range of misapplied heuristics and cognitive biases. Debate suffers from similar issues, in addition to turning truth-seeking to a verbal cleverness contest.
These tools are far from useless, but vastly overrated by people who think for a living.
F Tier: Folk Wisdom, Cultural Evolution, Divine Revelation "My grandmother always said..." "Ancient cultures knew..." "It came to me in a dream..."
Let's be specific about cultural evolution, since Henrich's The Secret of Our Success has made it trendy. It's genuinely fascinating that Fijians learned to process manioc to remove cyanide without understanding chemistry. It's clever that some societies use divination to randomize hunting locations. But compare manioc processing to penicillin discovery, randomized hunting to GPS satellites, traditional boat-building to the Apollo program.
Cultural evolution is real and occasionally produces useful knowledge. But it's slow, unreliable, and limited to problems your ancestors faced repeatedly over generations. When COVID hit, folk wisdom offered better funeral rites; science delivered mRNA vaccines in under a year.
The epistemic methods that gave us antibiotics, electricity, and the internet simply dwarf accumulated folk wisdom's contributions. A cultural evolution supporter might argue that cultural evolution discovered precursors to what I think of as our best tools: literacy, mathematics, and the scientific method. I don't dispute this, but cultural evolution's heyday is long gone. Humanity has largely superseded cultural evolution's slowness and fickleness with faster, more reliable epistemic methods.
F - - Tier: Arguing on Twitter, Facebook comments, watching Tiktok videos, etc. Extremely bad for your epistemics. Can delude you via presenting a facsimile of knowledge. Often worse than nothing. Like joining a gunfight with a SuperSoaker.
Ultimately, the exact positions on the tier list doesn’t matter all too much. The core perspectives I want to convey are a) the idea and saliency of building a tier list at all, and b) some ideas for how one can use and update such a tier list. The rest, ultimately, is up to you.
Thanks! Can you be more specific about which of my arguments you disagreed with? Your hypotheses seemed sufficiently far away from my actual arguments that I have trouble understanding where you're coming from.
I also think the actual tiers themselves matter much less than the logic behind the tiers and how to use them, which I emphasized multiple times in the post. Maybe my mistake was pasting the infographic on the forum post so people think that's the bulk of the argument? EDIT: I've deleted the infographic.
I thought I was reasonably clear in my post but I will try again. As far as I understand .your argument is that the items in the tiers are heuristics people might use to determine how to make decisions, and the "tiers" represent how useful/trustworthy they are at doing that (with stuff in lower tiers like "folk wisdom" being not that useful and stuff in higher tiers like RCTs being more useful)
But I don't really see "literacy" or "math" broadly construed as methods to reach any specific decision, they're simply things I might need to understand actual arguments (and for that matter I am convinced that people can use good heuristics whilst being functionally illiterate or innumerate). The only real reason I can think of for putting them at the top is "many people argue against trusting (F-tier) folk wisdom is bad, there are some good arguments about not overindexing on (B-tier) RCTs, there are few decent arguments on principle against (S-tier) reading or adding up, despite the fact that literacy helps genocidal grudges as well as scientific knowledge to spread. I agree with this, but I don't think it illustrates very much that can be used to help me make better decisions as an individual. Because what really matters if I'm using my literacy to help me make a decision is what I read and what things I read I trust; much more than whether I can trust I've parsed it correctly. Likewise I think what thought experiments I'm influenced by is more important than the idea that thought experiments are (possibly) less trustworthy than at helping me make decisions than a full blown philosophical framework or more trustworthy than folk wisdown.
FWIW I think the infographic was fine and would suggest reinstating it (I don't think the argument is clearer without it, and it's certainly harder for people to suggest methods you might have missed if you don't show methods you included!)
Your linkpost also strips most of the key parts from the article, which I suspect some of the downvoters missed
I think as an individual reading and mathematical modeling is more conducive to learning true things about the world more than most other things on the list. Certainly I read much more often than I conduct RCTs! Even working scientists have reading the literature as a major component of their overall process.
I also believe this is true for civilization overall. If we imagine in an alternative civilization that is incapable of RCTs but can learn things from direct observation, natural experiments, engineering, etc, I expect substantial progress is still possible. However, if all information can only be relayed via the oral tradition, I think it'd be very hard to build up a substantial civilization. There's a similar argument for math as well, though less so.
Likewise I think what thought experiments I'm influenced by is more important than the idea that thought experiments are (possibly) less trustworthy than at helping me make decisions than a full blown philosophical framework or more trustworthy than folk wisdown.
Sure, the article discusses this in some detail. Context and discernment definitely matters. I could've definitely spent more effort on it, but I was worried it was already too long, and am also unsure if I could provide anything novel that's relevant to specific people's situations anyway.
FWIW I think the infographic was fine and would suggest reinstating it (I don't think the argument is clearer without it, and it's certainly harder for people to suggest methods you might have missed if you don't show methods you included!)
I think the infographic probably makes it more likely for people to downvote the post without reading it.
Your linkpost also strips most of the key parts from the article, which I suspect some of the downvoters missed
Yeah the linkpost is just an introduction + explanation of why the post is relevant to EA Forum + link. I strongly suspect, based on substack analytics (which admittedly might be inaccurate) most people who downvoted the post didn't read or even skim the post. I frankly find this extremely [1]rude.
(Less than 1% of my substack's views came from the EA Forum, so pretty much every single one of the clickers have to have downvoted; I think it's much more likely that people who didn't read the post downvoted. I personally only downvote posts I've read, or at least skimmed carefully enough that I'm confident I'd downvote upon a closer read. I can't imagine having the arrogance to do otherwise.)
