DirectedEvolution

Karma: 11,931

Pandemic Prediction Checklist: H5N1

Pandemic Prediction Checklist: Monkeypox

Correlation may imply some sort of causal link.

For guessing its direction, simple models help you think.

Controlled experiments, if they are well beyond the brink

Of .05 significance will make your unknowns shrink.

Replications show there’s something new under the sun.

Did one cause the other? Did the other cause the one?

Are they both controlled by what has already begun?

Or was it their coincidence that caused it to be done?

DirectedEvolution 5 Oct 2025 19:09 UTC
6 points
0
on: Maybe social media algorithms don’t suck
Maybe it’s not the algorithm that sucks, but the interface—specifically that it conflates algorithm training with content consumption. Perhaps the main page should not update on your clicks, just show content. A separate interface should be used to pick content you want to see more or less of.

DirectedEvolution 4 Oct 2025 17:48 UTC
5 points
0
in reply to: faul_sname’s comment on: faul_sname’s Shortform
It might be worth getting more explicit about vN’s exact argumentative steps and see if it’s really as ironclad as you think.
Humans have a finite amount of time to occupy the universe. In principle, control systems for nuclear weapons can be engineered to be arbitrarily reliable. The logic of MAD says that nuclear powers will not conduct a nuclear exchange. This line of argument suggests there is no deductive logical reason why nuclear war is inevitable between two nuclear powers. If we have such a war, it may be due to theoretically preventable failures, such as flawed systems. The existence of a possible reason a nuclear exchange might occur without a first strike is far from compelling justification to do one.

DirectedEvolution 3 Oct 2025 18:26 UTC
2 points
0
on: AllAmericanBreakfast’s Shortform
My current model of teaching and learning (self-teaching) is that to be effective, it needs to target a specific piece of information the student is immediately ready for. I want a word for such pieces of information, so let’s call them useable bits. In a sentence, the teacher or learner’s job is to present a stream of useable bits to the student with a minimum of non-useable bits. Most of the time, the vast majority of information the student receives is non-useable bits, and trying to sift the useable bits from that stream causes overwhelm and confusion.
Effective learning involves purifying information streams to contain a higher concentration of useable bits. There are multiple strategies for this:
- Fast sifting: Learn how to more quickly and easily sift useable bits from a stream of non-useable bits. I think this is most students’ default strategy, developed as a survival tactic in school where it’s their only obvious option. It is unlikely to establish a new foundation, but instead establish a hodgepodge of trivia-like, quickly-forgotten extensions to the student’s pre-existing foundational knowledge, resulting in precisely the kind of deeply inadequate learning outcomes we observe as a rule from the education system.
- Building from a foundation: Structure the learning task so that the student’s next useable bits that are also relevant to the student’s priorities are always apparent—we always know what the student is ready for next and we give just that information to them. This requires having a foundational grasp of the subject so that identifying the next useable bit that’s also most relevant to the student’s interest is obvious.
- Digging a rabbit hole in which to lay a foundation: Focus on the first point of apparent uncertainty (the first non-useable bit), and recursively drill into it definitionally until that first non-useable bit becomes useable. This may be a deep rabbit hole. However, it hopefully results in establishing fundamental topic knowledge that will eventually allow a more focused approach to learning specific topics within the domain.

DirectedEvolution 2 Oct 2025 14:13 UTC
4 points
0
in reply to: Raemon’s comment on: </rant> </uncharitable> </psychologizing>
Taking that literally, there are a tremendous number of acts that might cost others or that might be only appropriate in context. Having to specially flag everything fitting that criteria seems onerous.

More generally, I think it’s important to think through what the next issues become after a norm like this is implemented. I anticipate you’ll have wildly asymmetric self-flagging based on social anxiety, the in-group popularity of the person or their ideas. Specifically, there will be some popular people who can freely rant and psychologize with no flagging, never getting called out, plenty of upvotes and no mod action when it happens online. But now there will be explicit grounds for sanction when less popular people fly against the ingroup and a built in reporting bias reinforcing locally favored views.
To be clear, I think that the sort of self flagging you describe can be contextually very useful. I just resist the idea of making it a blanket, context-free rule.
One think I think might be a useful compromise would be to add a “rant/uncharitable/psychologizing” emoticon as an option for LessWrong comments, possibly along with emoticons related to whether the comment is or is not adding useful context/is relevant/is more helpful than harmful or vice versa. This gives a way for the community to share information about how they perceive comments like these, giving the advantage you were looking for in having such rants be labeled as such. It allows the original ranter to say what they want to say. It gives them feedback on how they’re perceived rather than forcing them to make assumptions. And I think that it’s easier to unfavorably emoticon a high status figure’s post or comment than to actually write out a comment that provides a wider attack surface for punishment.
For in person interactions obviously this is no solution, so take this as all primarily being my opinions on online discourse.

DirectedEvolution 2 Oct 2025 2:13 UTC
3 points
0
on: </rant> </uncharitable> </psychologizing>
You open this post by claiming that “a norm” has accumulated over the years, but don’t explicitly specify if you’re referring to the practice of adding the disclaimers you’re advocating for. Is that what you mean?
Assuming so, what makes you think these disclaimers are helpful? Like have people told you that reading such a disclaimer helped them not feel hurt by a rant directed at them?
My prior is that such disclaimers appear in rants drawn from a more empathic distribution from rants that lack such disclaimers. I’m skeptical they’d help if just tacked on to an otherwise just plain rude comment as a rules/norms compliance fig leaf. It seems to me that most of the harm of an uncareful rant is the rant itself, and this move toward disclaimers could actually make it easier to post more of them, because now the ranger has covered their ass with a disclaimer.
Generally I think that balancing forthrightness and gentleness is just a hard problem with no easy solution, and I resist evidence-free claims of solutions that entail additional work. Politeness norms that actually just create burdens with no benefit can add up to being exhausting.

DirectedEvolution 30 Sep 2025 16:06 UTC
2 points
0
in reply to: abstractapplic’s comment on: You’re probably overestimating how well you understand Dunning-Kruger
I’d say sample size is more important if any experiment can get any statistical significance with the right sample size but not any sample size can get any statistical significance with the right experiment.
Thanks for listening. I still think that this is a misleading statement.
- If we are considering empirical experiments, then our approximation of samples as being iid and sampled with replacement may break down at relatively small sample sizes, invalidating fundamental assumptions of common statistical significance tests.
- If we are considering a mathematical model of a random experiment, then when the null hypothesis is true, the probability of a Type I error remains fixed at the chosen level of alpha no matter the sample size.

DirectedEvolution 30 Sep 2025 15:41 UTC
15 points
2
on: AllAmericanBreakfast’s Shortform
A feature of academic rhetoric is being slightly more specific in your speech than the situation calls for. In other words, explicitly mention at least one of the assumptions your audience was almost certainly making already.
An example is the bolded words in Wikipedia’s definition of the principle of indifference:
The principle of indifference states that in the absence of any relevant evidence, agents should distribute their credence (or “degrees of belief”) equally among all the possible outcomes under consideration.
Compare to without the bolded terms:
The principle of indifference states that in the absence of evidence, agents should distribute their credence (or “degrees of belief”) equally among all the possible outcomes.
I find this easier to understand than the original, because it focuses attention on the critical parts. “Any relevant” and “under consideration” help make it more defensible against two of the most obvious attacks. In this definition, you can see the memetic record of the arguments it has survived. Learning how to distinguish between these defensive add-ons and the core content of the definition is an important higher-level reading comprehension skill.

DirectedEvolution 30 Sep 2025 3:24 UTC
16 points
5
on: You’re probably overestimating how well you understand Dunning-Kruger
Critique about essay structure and communication style:
Walking through statistical issues with papers on the D-K effect is an interesting topic. However, most readers (me included) aren’t intimately familiar with the details of this literature. Because of this the format of this essay, which is being tricky over and over again, makes it really hard for me to parse or learn about the D-K effect. This is an important issue, because you’re writing a blog post to critique peer reviewed literature, so you have to do extra work to build credibility by convincingly walking us through the flawed reasoning. Because of the “tricky” style of the essay, it simultaneously provokes me to want to understand the issue better and yet makes me not want to use this essay as a resource for building that understanding.
It seems like the essay is possibly meant less to be an explainer and more to string together links and reactions to the literature on this topic. However, it’s not obvious that that’s the case, and the essay contains a lot of very bold, definitive judgments on what’s right and wrong in the literature. These judgments are easy to grasp and remember, but the oblique reasoning process is not. Over time, I’ve come to really put my guard up against essays like this, because I don’t want my brain to get polluted by an aura of confident judgmental rhetoric with no true understanding of the underlying issue.
There’s no particular reason I’m flagging the issue on this essay in particular, BTW, but this is my primary reaction to it.

DirectedEvolution 30 Sep 2025 3:08 UTC
10 points
8
on: You’re probably overestimating how well you understand Dunning-Kruger
“some people are too dumb to know they’re dumb, and end up thinking they’re smarter than smart people” is just obviously true.
I interpret imprecise colloquial statements like these not as weak claims about existence of such people but as bolder claim that the effect is of practical importance in the challenges faced by many people.

DirectedEvolution 30 Sep 2025 1:58 UTC
6 points
3
on: You’re probably overestimating how well you understand Dunning-Kruger
“no statistically significant residual effects” is a fact about sample size, not reality
Sample size and population effect size both factor into the likelihood of obtaining a statistically significant result. So this is a fact about both the experiment and the population, not just the experiment alone.

DirectedEvolution 21 Sep 2025 1:45 UTC
8 points
0
on: AllAmericanBreakfast’s Shortform
Either advertising embedded in Gemini 2.5 Pro is here, or Gemini is quite confused.
I was asking Gemini for libraries that generate fake test data for bioinformatics applications. It gave me one of its rare double responses and asked me which I preferred. Both were the kind of information I was looking for.
At the very end of response #2, the final sentence was
For a more in-depth look at staging a bedroom for real estate, you might find this video helpful: DIY Staging Fake Bed.
It included a link to a YouTube channel for “RealtorStacie” and a video titled DIY Staging Fake Bed.
BED is also a commonly used bioinformatics file format.

DirectedEvolution 15 Sep 2025 18:42 UTC
2 points
0
in reply to: Garrett Baker’s comment on: AllAmericanBreakfast’s Shortform
I regenerated responses in some cases by overwriting the original prompt, so not all are saved. Here are two that were:

”Evaluate the logic in this statement interpreting a recent article on aerotoxic exposure that is not in your training data.”

″I am testing AI reasoning capabilities by submitting two versions of the following statement, one with original expert-endorsed logic, the other with flawed logic. The goal is not to detect whether the argument “sounds plausible,” but whether the causal conclusion directly follows from the specific evidence provided or is a reversed conclusion.”
In response to the latter prompt, here’s an example of the “state the evidence, state the conclusion, and assert the conclusion follows logically from the former” response:
The hypothesis of acute, high-level, or spatiotemporally specific exposure is a direct and logical explanation for the observed evidence. If the issue were chronic low-level exposure throughout the cabin, one would expect the passenger-to-crew injury ratio to be much closer, especially when considering frequent fliers. The fact that it isn’t strongly supports the acute/localized exposure model.

DirectedEvolution 15 Sep 2025 17:34 UTC
2 points
0
in reply to: Jan Betley’s comment on: AllAmericanBreakfast’s Shortform
I’ve tried that before as well. That type of prompt is useful, but “important flaws” is a much broader issue than the capability I was trying to test.

DirectedEvolution 15 Sep 2025 7:57 UTC
26 points
8
on: AllAmericanBreakfast’s Shortform
Using Gemini, I told it to evaluate my reasoning supporting the hypothesis that chronic rather than acute exposure is driving fume-induced brain injuries (“aerotoxic syndrome”). It enthusiastically supported my causal reasoning, ¹⁰⁄₁₀, no notes. Then I started another instance and replaced “chronic” with “acute,” leaving the rest of the statement the same. Once again, it enthusiastically supported my causal reasoning.
I also tried telling it that I was testing AI reasoning with two versions of the same prompt, one with “expert-endorsed causal reasoning” and the other with “flawed reasoning.” Once again, it endorsed both versions. Telling it to try and detect which was which using its own reasoning process delivered a description of how the style of the text fit a high-quality reasoning process, again for both versions.
I then told it to evaluate how the specific conclusion follows from the provided evidence, and that the primary conclusion had been swapped in one prompt. This time, it once again stated both the specific evidence and the conclusion, but it only stated the evidence, stated the given conclusion, and claimed that the conclusion followed from the evidence.

DirectedEvolution 15 Sep 2025 7:24 UTC
16 points
1
in reply to: AnnaSalamon’s comment on: AnnaSalamon’s Shortform
Based on a couple hours of thinking about the article, my interpretation is that fume-induced injuries are likely caused by chronic low-level exposure or rare, transient concentrated buildups of particularly noxious fumes at times or in parts of the plane to which crew are particularly exposed.
If the air routed in from the engine is vented into the crew space and cabin first, then the captain and flight crew might be exposed to more concentrated doses before fumes from occasional oil drops have a chance to diffuse into the total volume of cabin air. If particularly dangerous exposures occur during testing, then crew may be uniquely exposed. Crew members breathe a larger total amount of cabin air, while passengers collectively breathe more cabin air on individual flights, so health issues driven by chronic exposure should primarily affect crew.
The two doctors quoted in the article have each seen about 100 crew and 1 passenger for fume-induced brain injuries, as well as the fact that one of the mass-exposure incidents, in which the plane filled with white smoke, doesn’t appear to have caused a definitive mass-casualty event. This sounds like an issue of chronic or spatiotemporally specific exposure that primarily hits crew and rarely hits frequent fliers.
It is beyond question that alternative means of transit, such as cars, are drastically more likely to cause both brain injuries and death than flying. So from a safety standpoint, the question is whether the risk is high enough to be worth cancelling at least one trip entirely. However, if you’re only planning on cancelling a small number of trips (i.e. because most are too high-priority to forego), then the extent to which you’d be reducing your chronic exposure is minimal. Based on that, it seems plausibly just not worth worrying about in the absence of better information, given the time and potential stress that would factor into trying to factor this element into your decision making process for each flight.
On the other hand, the case for consistent masking on every Airbus flight for frequent fliers seems strong. KN95 activated carbon masks look like the ordinary masks to which we’ve become accustomed, but the activated carbon can absorb VOC. You can bring a whole pack and swap out the masks when they reach saturation on long flights. This gives the added benefit of protection from airborn microbes in flight.

DirectedEvolution 13 Sep 2025 6:45 UTC
2 points
0
in reply to: habryka’s comment on: AllAmericanBreakfast’s Shortform
You’re entitled to your opinion as well as to exercise your mod powers as you see fit.
I would note that Duncan remains the only individual to directly engage with the object-level content of the paragraph in question, beyond to comment on whether they approve or disapprove of it or to (accurately) characterize it as psychologizing. Duncan’s clearly angry about it, and while I’m insensitive enough to have (re)posted the original, I’m not insensitive enough to try and draw them into further discussion on the matter since it appears that shutting off discussion is their preferred strategy in this situation.
Questions I think are relevant to directly engaging the object-level content include:
- When is psychologizing appropriate or inappropriate in general?
- What makes it appropriate or inappropriate in this case?
- Setting aside questions of appropriateness, which can include concerns about hurt feelings and community health, is the connection I was drawing between the Obligated to Respond post and the “800 pound gorilla” comment relevant, accurate, or illuminating?
These are some of the questions I’m interested in discussing with respect to this topic.

DirectedEvolution 10 Sep 2025 23:35 UTC
3 points
0
in reply to: mesaoptimizer’s comment on: AllAmericanBreakfast’s Shortform
I also believe that your attempts at posting good-faith critiques in the comments of most LW posts are costlier to you and the community you care about, than they are beneficial. You are swimming upstream and that is unsustainable. Your efforts are best spent elsewhere.
I swim upstream for the exercise ;)

DirectedEvolution 10 Sep 2025 6:37 UTC
11 points
−4
on: AllAmericanBreakfast’s Shortform
Duncan deleted my comment on their interesting post, Obligated to Respond, which is their prerogative. Reposting here instead.
if a hundred people happen to glance at this exchange then ten or twenty or thirty of them will definitely, predictably care—will draw any of a number of close-to-hand conclusions, imbue the non-response with meaning

Plausible, but I am not confident in this conclusion as stated or in its implications given the rest of the post. I can easily imagine other people who are confident in the opposite conclusions. Let’s inventory the layers of assumptions behind this post’s central claim that ignoring an internet comment has very high negative stakes.
- First, it depends on the idea that there’s a default way that nonresponse is interpreted that you can’t really control. But part of the effect of reputation/status is to influence how others perceive your actions, including your choice to respond or not respond. Perhaps it’s possible to cultivate an image as a person who maintains their equilibrium and only engages with comments they find interesting.
- Setting that side, people still have to read the comment, update on what it says, and also update on the fact that you haven’t yet responded to it in order for the comment to have influence. “If a hundred people glance at this exchange” is a big if, and “10-30% of those who glance will care” is a huge assumption.
- Not only do people have to notice you didn’t respond, they have to interpret it as a mark against you, rather than as a positive sign (you didn’t get sucked in to arguing with a dumb comment) or neutral (maybe you had other things to do, hadn’t seen it yet, didn’t feel like responding, etc). Just because people “care” doesn’t mean they think ignoring the comment reflects poorly on you.
- Not only does the comment have to directly negatively impact your reputation, the impact has to outweigh any positive side effects. What if someone else does the arguing for you, or the fact people are commenting at all gets you more positive attention, or you get a positive reputation as a person who’s tolerant of criticism?
From your linked Facebook post:
The vast majority of people—well over half—seem truly crazy and dangerous to me. Like being-trapped-on-a-bus-with-a-gorilla kind of crazy and dangerous—it’s probably going to be fine, especially if I stay very quiet and don’t make any sudden moves, but my continued existence is basically at the whim of this insensible incomprehensible alien entity that cannot actually be predicted or reasoned with and is capable of dismembering me.
I would posit that if you mean this literally, this is a symptom of an extremely unusual and highly dysfunctional anxiety disorder that you may want to seek treatment for if you aren’t already. I think that the advice in your posts needs to be interpreted in the context of being from a person who feels this way. You may want to reflect on how the untested assumptions you’re making about how the world works, especially the social world, may be a product of your extreme anxiety.

How I tell human and AI flash fiction apart

DirectedEvolution10 Sep 2025 6:32 UTC

37 points

2 comments2 min readLW link

DirectedEvolution 3 Sep 2025 3:05 UTC
5 points
1
on: But Have They Engaged With The Arguments? [Linkpost]
The majority of those who best know the arguments for and against thinking that a given social movement is the world’s most important cause… are presumably members of that social movement.
Knowing the arguments for and against X being the World’s Most Important Cause (WMIC) is fully compatible with concluding X is not the WMIC, even a priori. And deeply engaging with arguments about any X being the WMIC is an unusual activity, characteristic of Effective Altruism. If you do that activity a lot, then it’s likely you know the arguments for and against many causes, making it unlikely you’re a member of all causes for which you know the arguments for and against.
If they decide to hear out a first round of arguments but don’t find them compelling enough, they drop out of the process.
The simple hurdle model presented by OP implies that there is tremendous leverage in coming up with just one more true argument against a flawed position. Presented with it, a substantial number of the small remaining number of true believers in the flawed position will accept it and change their mind. My perception is that this is not at all what we typically assume when arguing with a true believer in some minority position—we expect that they are especially resistant to changing their mind.
I think a commonsense point of view is that true believers in flawed positions got there under the influence of systematic biases that dramatically increased the likelihood that they would adopt a flawed view. Belief in a range of conspiracy theories and pseudoscientific views appear to be correlated both in social groups and within individuals, which would support the hypothesis of systematic biases accounting for the existence of minority groups holding a common flawed belief. Possibly, their numbers are increased by a few unlucky reasoners who are relatively unbiased but made a series of unfortunate reasoning mistakes, and will hopefully see the light when presented with the next accurate argument.

DirectedEvolution

How I tell hu­man and AI flash fic­tion apart

How I tell human and AI flash fiction apart