cgiles

> I learned how to be a good multiplier: how to help people out, help them grow, unblock procedural bottlenecks in a lasting way...

So, here you perfectly describe a 10x middle manager. Or just a "good" middle manager, because I agree this 10x business is kind of silly.

I will never understand why people take all these soft skills unrelated to programming and say they are more important to a programmer than skill at programming. It's not that these skills are undesirable. It's just that, to continue the sports analogies, it is like saying that it is more important for a basketball player to be fast and be a good team player rather than be skilled at shooting and blocking.

If anything it is the other way around, people with good soft skills and bad technical skills are an absolute menace and plague if they try to get involved in anything technical. They have no idea how much they screw up everything they touch.

If you aren't a programmer, and you're a manager of programmers, then fine. A coach needs a totally different skillset than a player and you can be a fine coach even if you are a mediocre player.

Because, at best, it is a very dubious and hamhanded way to solve a problem that is ill-defined.

Is the problem advertising, or advertising to minors, of nicotine products? Then ban that.

Is the problem that minors are able to buy vapes even though they shouldn't be able to? Then strengthen enforcement (ah, but that costs money).

Is the problem that some vape products are too easily smuggled into schools and look like USB sticks? Fine, then set a minimum size requirement on vaping devices. Heck, mandate they be colored a bright primary color.

Who cares what Juul or the tobacco industry says? They will say whatever they need to say to sell stuff, we know that. What matters is that vaping is substantially less harmful than tobacco by all accounts.

Personally, I could potentially get behind a total ban on tobacco, as it is very harmful and costly to society, although I'm generally anti-prohibition. But what makes no sense in any possible universe is banning vaping while keeping cigarettes legal.

What I mean is that he produces an amount and quality of work that is comparable to other people at his experience and title level. In some ways better, in other ways worse, but on average, he is producing at least as much as his peers are. He's quite talented but lazy.

Obviously there is never a situation where you are out of work to do but some would say it is unfair to expect a talented person to produce more for the same salary and title as less talented people (I have no clear opinion on that, I'm only concerned about the morale issues). After all, the whole premise of being salaried rather than hourly is that you are paid to produce a certain amount of work, not work a certain amount of hours.

Normally a solution could be promotion but that is not an option right now.

> Why are they not motivated enough by their work maybe? Are they not aware maybe how a higher than "normal" engaging in distracting activities will hinder their development and ultimately their career?

Great point, and I think it needs elaboration. Particularly if increased slacking/FB use is something that has increased over time for a person, I would also take it as a sign that they are losing motivation. Banning FB or chastising them when you see them use it won't restore that motivation.

A lot of people go through phases like this. It can be symptomatic of:

- Overwork, stress, potential burnout

- A workload the person finds repetitive, not challenging, or uninteresting (maybe they need a change in task type)

- A feeling that work is not rewarded, possibly that person needs more praise or recognition, or even just interest from manager/others about what they are doing

- Person is lost or stuck on what to do next, the task may be too hard for them or they need to talk it over with someone. Many introverts especially will not ask for help when stuck.

> engineer's brains being constantly on fire

The elusive flow. In my experience flow states are most common when: (A) there is sufficient freedom to explore the problem space, (B) fairly rapid feedback, and (C) people are impressed by or interested in what I am doing. Success breeds success. I am convinced that one of the most important jobs of a manager is to cheerlead for the team and remind them why their work is interesting and valued. Who else is going to do that if not the manager? Feelings of success and confidence breeds more success.

Having people do quick demos or small green-field projects, even if it isn't exactly the top priority at the moment, then complimenting the result, is a good way to boost morale and interest in my experience.

This is actually a difficult problem, because there are two kinds of slackers: (1) the kind that work less but produce roughly the same as, or sometimes even more than, everyone else (i.e., smart but lazy), and (2) the kind that don't work or produce.

Dealing with (2) is absolutely the manager's job and such a person should be fired if attempts to identify and fix an underlying problem fail.

But (1) is really difficult. We have one of those right now, let's call him Bill. Our manager is hesitant to do anything about him, because the manager wants to incentivize productivity, not butt-in-chair time (we're all salaried, with no set hours). But Bill comes in 4 hours a day, seemingly works 1 of them, and spends the other 3 trying to distract everyone else with off-topic conversations. If he comes in at all.

Amazingly Bill does get work done, somehow. But it is terrible for morale. Worse, some of the junior members are starting to think they can get away with the same thing Bill is doing, and they can't. As the senior dev, the "big brother" to our manager's "dad", it falls to me to do something about it, but I have no idea what, if anything.

Manager fears that if he rewards finishing tasks quickly with more work, it will also harm morale and productivity. And there is a decent argument to be made for that. Having experienced that firsthand, it is highly irritating and incentivizes people to work slowly.

Additionally, in reply to GP, a manager that has great technical skills (10) and good managerial skills (1-9) must be as rare as a unicorn. Even a manager who started with good technical chops inevitably will lose them over time when they are an Outlook jockey all day. OTOH, I have seen plenty of otherwise decent managers who think their technical chops are good when they aren't. Pretty much the whole premise of Dilbert. The best realistic alternative seems to be managers who know their technical limits, delegate, and seek advice from their subordinates.

I agree with your basic premise, but I think the complaint about ME3 was that it didn't reflect any previous decisions you made in the ending, not that the story failed to fork exponentially at every decision point. Indeed, IMO, ME3 did a really good job of making it feel like your decisions matter...until the very end.

Star Ocean: The Second Story is yet another example of an RPG that does this; it has 70-80 endings, many forking paths including mutually exclusive party member choices, relationships between characters, etc. The trick for making it feasible seems to be to make the ending modular so that it includes/removes different cutscenes based on what happened before. As I understand it there are quite a few old-school RPGs that do this.

Of course it is harder if you are talking about voice acting and super-detailed 3D assets. But assets aren't that hard considering you already made most of them for the rest of the game, and realistic voice synthesis is perhaps not that far away. So I don't think most people were expecting model (B), but even that could be viable if you controlled the number of forks carefully.

> I wouldn't generalize about the Bay area from a few days in a particular corner of San Mateo County, though.

I'm very wary about it too, but the persistent claims that SF/Bay Area has some peculiar kind of dysfunction, rendering it hard to live in, that doesn't exist in other major metros has always piqued my curiosity. It seems like an extraordinary claim requiring extraordinary evidence.

The only real evidence I've seen supporting this idea is very high housing prices. I sat next to a woman on my flight who paid $5K/mo for a 2-bedroom apartment, shared with 2 others. Sounds insane. And of course I've seen more systematic evidence of this.

But after my visit, I wonder whether she, and perhaps others on this board, don't consider the SF suburbs as worthy for a potential living space. Or alternatively whether the claims are true that there is very high demand and these suburbs would be demolished and replaced by high-rise apartments but for zoning restrictions.

> Then you quite possibly didn't actually see any of the City and County of San Francisco.

Very possible. I got a good sampling of a 5 mile radius around SFO and sporadic sampling elsewhere. I wouldn't claim to know 1/100th of what a resident does about the area.

However, I was interested in testing to the extent I could the implicit hypothesis often floated on HN, "the Bay Area is borderline uninhabitable". If an area 15 miles away from downtown seems quite low-density, and assuming a 15 mile commute is reasonable (assisted by BART, whose analogue doesn't exist in many comparable cities), I concluded complaints on HN are overblown, especially considering everyone doesn't commute to downtown. For example, YouTube headquarters were pretty close to SFO. Stanford is not near downtown either. Not sure where all the other headquarters are located.

I don't have high confidence in that conclusion obviously because my information is very limited. But I am comparing to cities I know better like DC and NYC.

> it is not the main public transit in the city, Muni is

I assume "Muni" means buses or cable cars, and I will only say that during wandering around for ~20 hours, I didn't see a single bus, so I conclude Muni doesn't service suburbs.

I visited the Bay Area for the first time (as an adult) a few weeks ago and was astonished by the difference between what I was expecting based on HN comments, and reality.

All I saw of the city made it seem like a giant suburb. BART stations were miles apart and therefore practically useless. It is really not that dense and traffic is not bad.

I am sure some areas are different. I was mainly in the area around SFO, and my conference was not hosted in downtown because apparently the homeless problem is bigger there, but at that level of low density and congestion, I fail to see how public transit is much of a win. People will not realistically take public transit if it doubles their travel time. And this case applies doubly to medium-sized suburbanized cities in the central US.

On the other hand, I can see what people mean about a housing problem when there are virtually no buildings taller than 3 stories except hotels in such an urban area.

It is just like anything else, for example politics. Many politicians have strong convictions and care about the electorate and outcomes (I have several politician friends).

But when push comes to shove, keeping their job and food on the table for their families comes first. So ethical compromises are made.

Similarly, for tech people and privacy. I would wager there are dozens if not hundreds here on HN who have raised concerns with management about the privacy implications of something they're being requested to build. Management says, build it anyway. Programmer reluctantly complies because they like their job.

WRT science it is obvious that it works over the long term. But in the short term it can be quite nasty. So I would say if you are a member of the public, you can safely believe something scientists have said for a long time without controversy, like the helical structure of DNA, but new or controversial findings should be taken with a large grain of salt.

It is honestly hard to tell in these situations when they are lying by omission and when they are plain stupid. I think it is usually a form of apathy whereby what is produced is bullshit (in the technical meaning of the essay, "On Bullshit", whereby people don't care whether a thing is true or false so long as it's published).

> On a bigger scale, this just fuels public anti-science

Yep... But people have careers to maintain and grants to get.

I totally agree with you and take your side on this issue but the systemic issues are huge and I don't know what exactly it would take to fix them.

Ha. I have a lot of those papers too, but none that are BS in such a high tier, congrats! (or "congrats")

At this point I am cynical enough to believe people include analysts on their papers as an insurance policy to have the "licensed" person to give the seal of approval, and take the blame if necessary, for their shitty data. Which they will never listen to concerns about before publication.

Just today I am working on a qPCR array (yes people still use those) with one "housekeeping" gene replicated 4 times for ddCt normalization. 80 miRs go up, none go down, out of 400. No one sees any problem with this, the data is the data.

Them: Oh, and can you, cgiles, also tell us about the genes regulated by these miRs?

Me: Not really in any principled way, especially when the underlying data is incredibly fishy, and even if I did, we would be saying all the target genes go down. Don't you think people will find that odd? Do you think perhaps there could be a problem with your "housekeeping" gene?

Them: Who cares? Just run the algorithm.

Thank you for this tale. I'm glad I'm not the only one in this situation.

Funny, I'm a postdoc (in biomedical research), and I and my boss, a PI, had this exact debate last night.

I took the side of 781, and he was justin66. My perspective is that you cannot take data, or a concept, that is fundamentally broken, add a little effort, and fix it.

For example, if a collaborator sends data for analysis where the sample size was too small for the effect, or worse, asked to find statistical differences between 1-2 samples per group, you can't fix that with any amount of algorithmic mumbo-jumbo. You can only fix it by redoing the experiment properly. Although we are asked to do this sort of thing almost daily (not always this blatantly).

His perspective is, and maybe I'm being a bit uncharitable, you take what you're given, do your best, clearly describe what you did and honestly note the limitations, and move on. Even if the result is something highly dubious, that is for the reviewers to decide.

But I personally think a little loud whining is called for in the sciences to discourage totally half-baked ideas and data from ever seeing the light of day. But the trouble is the pressures are so intense and funding is so tight, people feel compelled to halfass many things, which is the real problem.

EDIT: I should add I am not talking about data falsification or major ethical breaches like that. I'm talking about things that more amount to a form of pressure-induced willful stupidity, no dishonesty involved. I would speculate this is much more of a cause of the replication crisis than outright dishonesty.

People don't like to outright lie because it could cost them their careers. But they certainly can conveniently forget some basic scientific principles at times if it suits them and they can't get a decent story while being rigorous. And their excuse will be -- well, we clearly documented everything we did in the paper, and the reviewers thought it was OK (because the reviewers do the same sorts of thing). And this, folks, is how you get a lot of shitty papers.

But if you're trying to merge on, whether or not you have to stop before merging is not totally in your control. Specifically, if the right lane is packed, you may have to stop if there is no opening as you pull up. Then it will be harder to merge in, not only because you are starting from zero velocity but also because the lane is packed.

That is why I think people should not get in the right lane on a highway unless they are about to get off.

As a slower-than-average driver, this is exactly what I do on any highway except an interstate.

My logic is that if there are 3 or more lanes, the furthest right lane should be for getting on or off only, further left is for fast long-distance driving, and the slowest lane should be second-from-the-right.

I do not see how it makes sense to encourage slower drivers to take the far right lane in this scenario. It is actually pretty dangerous for people trying to merge on to have the right lane heavily occupied.

However, indeed the counterargument is that this leads to a lot of lane swapping when people get on and want to cross over to the far left to travel, and then all the way back over to the right when they want to get off. Especially since speed of travel is not necessarily related to whether the car is making a short or long trip on the highway.

The real question, I think, is what should a fast driver do if they get onto a highway, and are planning to get off in 1-2 miles. Should they bother crossing over to the left or stay in the right lane?

> I'm not arguing that we can't address other societally-imposed disadvantages...<AA opponents> are arguing that one form of assistance is verboten.

I didn't think you were. But, as a (weak) AA opponent, I would say that the fundamental problem with AA is that it needlessly replaces data about individuals, which we have, with group-mean data. Throwing away information like this naturally leads to all sorts of perverse results, where a poor white is considered to be doing well because the average white is, and vice-versa for blacks, etc.

So I don't say AA is "verboten", but that AA is not an alternative to means-testing, it IS means-testing, except it is means-testing of groups instead of individuals. It is also a far worse intervention than means-testing, on its own terms, for a variety of reasons [1].

It is possible to argue that AA is about justice, but in practice the benefits are always allocated based on current bad outcome. Otherwise, Asians would get preference at Harvard.

---

[1] AA disadvantages:

- AA gives advantages to people based on ethnicity, regardless of whether an individual's ancestors actually suffered the abuse in question.

- Means-testing turns the philosophical and intractable question of "how much should past suffering be worth? (in terms of admission preference, etc)" with the concrete and measurable question "how poorly is this individual doing now?"

- Means-testing is far less politically controversial.

- The AA mindset encourages people to invent an ever-expanding list of disadvantaged groups so that people can get a slice of the redress pie, leading to disputes over the relative value of different kinds of oppression.

- Even wealthy, beautiful, well-educated members of a disadvantaged group get the AA benefits. AA "subsidizes" individuals who don't need it.

Of course. There is however a subtle but important distinction between talking about (race-based) "past discrimination" (RBPD) versus any "past disadvantage" (such as grandpa with polio).

It is this limitation of focus to RBPD that I'm objecting to. First, I think it is a form of selection bias, as I said.

Selection bias is a big issue in politics -- I have found that climate activists like my wife hate hearing "we can deal with climate later once we've solved poverty" -- because they very reasonably know that "later" in politics means "never", and they interpret this statement to mean "poverty is more important than climate".

Interestingly, I think this is what is going on in reverse when affirmative action advocates hear people say "We should do something about RBPD, but poverty is even worse, we should focus on that first". AA advocates hear "RBPD is unimportant and we'll do it later (never)".

Perhaps you could help me with the following summary of every discussion I've had about this topic:

- Why do we care about RBPD?

Because it has lingering bad effects.

- But so does any form of past disadvantage. Why then do we focus on RBPD?

Because we have to start somewhere, and we don't have enough bandwidth to focus on everything.

- OK, but wouldn't it be simpler then to target disadvantaged individuals rather than disadvantaged groups because then we wouldn't have to solve the impossible problem of quantifying everyone's group-based net advantages and disadvantages? And also, if you adjust outcomes based on group outcomes, rather than individual outcomes (poverty), won't it be impossible to ever balance the ledger, because every time you give an advantage to one group, you disadvantage another, some members of which are disadvantaged on other axes?

And this is where it breaks down. The answer usually given is "Don't you believe in RBPD? It is horrible and there is plenty of evidence it exists!" But this doesn't answer the question at all. I've already agreed RBPD exists.

The only answer I can think of is perhaps that we should distinguish between problems caused by society (slavery) and problems not caused by society (polio), and to say that society only has an obligation to fix the former. But in this case, it becomes an issue of justice rather than outcome, and we are no longer justifying affirmative action on the basis of current bad outcomes but rather on past injustice. And if you do that, then you lose the ability to ever say when the ledger has been balanced and consequently remedies will have to go on forever.

> you're left with the mere observation that there are multiple factors that work against different groups of people. Perhaps we should account for more of those factors. But it doesn't follow logically that we should stop accounting for the factors that hold black families back.

Indeed there are. But in practical terms, there are so many things that happened in the past to people that disadvantage them today (e.g., grandpa had polio) that it is impossible in practice to account for all of them.

If we however take some factors, generally the ones that are easy to observe such as race, and try to address them without addressing the rest, that is textbook selection bias. So the logical argument against (semi-exclusively) accounting for the factors that hold black families back is that it would be selection bias to do so.

Furthermore, advocates of affirmative action rarely answer the very important question "how will we know when the policy is successful and past discrimination has been offset by present advantage?"

The only reasonable answer I can think of is "when members of the discriminated group have equal outcomes on average to nondiscriminated groups on average". But if this is true, then given the above, the logical way to proceed is to target underprivileged people individually, as that is the only way you will ever get such equity over all the different factors.

I like this way of framing it, although I find it questionable whether SATs, grades, etc measure C exclusively.

Because almost all candidates are the same age when applying, C also carries information about m, in the sense that the same equation applies to their time in K-12, so candidates with a high C did so either because they had a high learning rate (m) or went to a high-quality high school or had a high-quality environment (x), or some combination.

Thus I would say that SAT score measures mx. C at time t = mx + C at t - 1, or dC/dt = mx. Schools are interested in m, which is not easily measurable directly. Affirmative action is a very crude way to look for candidates that have a high m, low x, and moderate to high C at application time.

My conclusion is that if it were possible, colleges would ideally look for candidates that performed substantially better on SATs etc than their peers at each high school, as that would control for x and allow you to estimate m. Affirmative action, on the other hand, is a cruder but partially functional way to do this, as it uses race as a bad proxy for x.

> I wonder to what extent it is possible to slow the damage. For instance, if you're bleeding, you may be able to slow it - but if you don't stop it completely, eventually you will bleed to death.

Yes, exactly. Well, the simplistic model is that you have a rate of damage (RD) and rate of repair/replacement (RP), and if RD > RP, the system will eventually fall apart. "Repair" consists of both your body's endogeneous repair or replacement processes and exogeneous things we might do.

Most researchers focus on RD, because it seems easier for us to do things about that. Calorie restriction probably mostly affects RD instead of RP, for example. In high theory, one could stop aging purely by lowering RD enough so that it is below the body's endogeneous RR.

But in reality, it is easier for the body to detect and repair some kinds of damage than others. It is quite hard for the body to detect DNA mutation in a cell before it starts actively causing aberrant behavior in a cell, although such mechanisms exist.

> I think it is ridiculous to believe that in a universe where galaxies die, human beings can keep going like the Duracell bunny for as long as we like

Well, at THAT timescale, we are talking about problems of physics and thermodynamics, not biology. But if you consider that in some sense life on Earth has proceeded in an unbroken chain for ~4B years, you can see that it is possible to go quite far by decreasing local entropy at the expense, of course, of global entropy. I do not see how physics prevents human lifespan of at least 10K years, but the technical problems are immense, and I do not predict major success within our lifetimes.

Whether cancer is special in this regard depends on what you mean by aging. I work in this area and consider it to mean "damage (maladaptive changes) that accumulate over time, first degrading function and ultimately leading to catastrophic failure" . This is not the only model but it is widespread.

So for something like the cardiovascular system, chronic stress, inflammation, ROS etc are the triggers, atherosclerosis and endothelial weakening is the damage, and stroke or heart attack is the failure. In the case of cancer, it is radiation or replication error, DNA mutation, and metastasis, respectively.

The key for "curing" aging would be the ability to reverse or repair damage (in this case, to selectively kill cells that are mutated or precancerous). To extend life it is sufficient to merely slow damage accumulation rate.

Except for Mendelian diseases, not really. The sad fact is that for the so-called "complex diseases", we have very little idea what is going on inside the black box. I won't presume to guess whether we know less about them or about behavior, but the ignorance level is very high in both cases. They are called "complex" because we don't understand them, and because they seem to have many causes, both individual genetic loci and environmental variables, each contributing a little bit rather than something like sickle-cell anemia which has one big cause.

However, I think our techniques are quite sufficient right now to say X% of the variance in phenotype P is attributable to genetics (i.e., genetics partially cause P). That is a totally different thing from explaining how the genotype causes the phenotype.

The typical path to determining causality for a variant is: Variant -> (Gene expression or some other molecular-level phenomenon) -> Phenotype. Actually, associating a variant with a phenotype is the easy part, determining what happens in between is much harder. Another hard part is determining which variant does something since variants are highly correlated with each other.

I am not sure I understand your question exactly so I'll give a few different answers. One is that I see no difference whether the phenotype is behavioral or some disease occurring below your neck. It all depends on the statistics and effect sizes of the phenotype and how good a predictor "race" is.

Another is that actually it is much more difficult, obviously, to determine causality than a correlation, regardless of what you are correlating with what.

And a final answer is that we generally consider a causal relationship between genetics and a phenotype proved with acceptable levels of error if the two are correlated, and we control for all the environmental variables we can, and the correlation remains. You can also look at heritability with twins separated at birth and raised in different environments and other such experiments to control for environment.

I have no idea whether the overall pendulum is swinging back from "race is a purely social construct". I hope so, because it has always seemed absurd to me, for the very reasons the author cites, just as absurd as the idea that race is a hard determinant of anything.

But what I can say is that in molecular biology, working scientists have completely ignored this politically charged debate and continued to use the concept this whole time. In one example I am directly familiar with, researchers have continued to use and stratify genome-wide association studies by self-reported race/ethnicity in the search for variants causal for lupus. This is important not only because as with many diseases, there is a difference in likelihood of getting the disease between "races" not attributable to lifestyle, but also because it is entirely possible, even probable, that the mechanisms causing the disease are somewhat different between ethnicities.

Actually, therefore, the understanding that there are real differences between "races" in the context of disease is actually helpful for the smaller groups because it means that special attention is paid to the etiology of their disease apart from the general etiology. In the same way that, for many years, most biology research was done on males and it was just assumed that the findings would always apply to females. That was incorrect, and now studies are done to determine gender differences in treatment and disease etiology.

IMO this whole debate has dragged on as long as it has because too many in the public are seemingly incapable -- or unwilling -- to understand basic concepts about population means and variances, and in particular that in a situation like this where population means are very real but usually small, and the variance is high, knowing what "race" you are usually conveys little information about some other attribute of interest. Usually, but not always, as is particularly the case with many diseases.

I have grown particularly tired of the argument that, because a taxonomy, like any clustering, is fuzzy and the number of clusters is somewhat arbitrary, the whole thing conveys no useful information. It is patently false.

I wasn't really attempting to express an opinion on how valid or likely this claim is, just to clarify what I took to be GP's point.

But for fun, I'll try to address the question. I live in a conservative state and do some work with public health policy, working with local politicians and have friends in LEO and military. Here are the common arguments/viewpoints (I am not endorsing them):

- The biggest proportion of gun deaths are suicides, so focusing on random massacres or inner-city violence, which are tiny blips on the mortality radar, is not a productive or a good-faith basis for banning guns.

- There is a (somewhat justified IMO) fear of a slippery-slope approach to gun regulation. It is not controversial or speculative to say there is a large contingent on the left that would like to totally ban them if it were possible.

- Guns are very limited in the damage they can do. A bomb in a high-school assembly could do a lot more damage than the most well-prepared shooter, and with much less risk to the perpetrator. At best, guns are not the most dangerous thing to be focused on and at worst, banning them could push psychopaths into more dangerous alternatives.

LEOs and military folks lean conservative. That means they tend to generally distrust or even fear the very government they work for. LEOs in particular are well acquainted with the damage guns do on a daily basis but frequently do not see banning them as a solution because:

- It is impractical

- The people killing each other with them are for the most part poor or (by definition) criminals, so no one cares about them

- The original intention of the 2nd Amendment was to protect the people from the government that, again, they distrust and fear, so even if banning guns would save lives, which they would dispute, it is a worthwhile cost to pay to protect against tyranny.

- There is a viewpoint that gun control actually increases violent crime, and therefore deaths, because guns serve as a deterrent. I have been in an extremely long-running debate with a local politician about whether this is the case.

As a result of all these arguments, I think a serious effort to confiscate guns would make many, many people in conservative states, including LEOs and military, very highly irritated. Irritated enough to rebel? I have no idea.

Again, I am not endorsing most of this, but there is one argument that resonates with me from a public health perspective. Gun homicide is an extremely rare cause of death. I personally feel that it would be orders of magnitude more helpful if the left would channel its energy into increasing funding for your friendly neighborhood National Institutes of Health.

I think the implication was that a very large fraction of police/national guard/US armed forces would at least refuse to enforce such a law, if not join the rebellion itself. Indeed, I've never met a more adamant group of 2nd amendment supporters than my military friends.

But should such a dire scenario occur, I have always wondered how these small arms would fare against UAVs. The trouble is that we have precedent in this country that in a large-scale rebellion, it is taken out of law enforcement's hands and put into the military's hands.

I agree. However, nothing much better is available. I loved reading work from the Santa Fe Institute and Stuart Kauffman and so on, but the reality is that despite their best efforts, they provided nothing better. General statements about the system-as-a-whole, but no specifics about where to look for the etiology of a specific disease or process. Nobels await those who could do more.

Until such time as someone does bring theory to biology, for example, wet-labbers (not me) struggle on. I am in a little doubt as to whether general theories about complex systems could provide useful predictive frameworks, but if so, great. But until then, pre-study odds are zero if you don't do the study.

It is solid probabilistic reasoning if you accept:

- A publication bias term u that is pulled out of thin air for different scenarios.

- A parameter R that reflects the unknowable proportion of true relationships/hypotheses in a field compared to the universe of possible hypotheses. The pre-study odds he calculates are determined entirely by R.

Actually this paper provides a good framework for determining what factors will affect the PPV of a field. In that regard it was a good contribution. But the enormous leap to the title -- which can only be achieved by massive assumptions about R and u -- was baseless fearmongering and demagoguery.

If one wants to address these issues at the 50,000 feet level, it would be far better to look at something concrete, such as reproducibility rates, as indeed many have done, rather than models based on pre-study odds. If we actually knew the pre-study odds, or IOW the proportion of hypotheses in hypothesis space that are true, then we could just correct our p-values for that and be done with it. Although I suppose that many Bayesians would not see any problem with abstracting away everything we don't know into priors R and u, making wild guesses about them, and drawing conclusions.

We do have a replication crisis on our hands. But the pre-study odds, whatever they are, are unchangeable. If this paper had been framed in terms of "the lower the pre-study odds, the higher the power will need to be to compensate to get an acceptable level of reproducibility", I would accept it wholeheartedly, although then it would have been a simple and obvious statement rather than a citation-grabber.

No doubt it is a target-rich environment. However, his arguments are so generic that they could apply to any field studying anything multifactorial or with high variance.

Fundamentally, however, I don't believe, for example, that physicists are more scrupulous/intelligent/honest than biologists, who are more so than psychologists. The irreproducibility rates in a field have to be primarily caused by the nature of the system studied in that field rather than some inherent attributes of the researchers.

So any field will have some constant level of irreproducibility which which essentially cannot be changed. But some researchers do better than others. My fundamental point is that if you want to change what can be changed, and genuinely improve a field, you have to focus the criticisms on specific bad actors rather than target an entire field at once and leverage a near-tautology (more complex systems yield more irreproducible results) into a 20-year tirade/career.

Also, helpfully for people like Ioannidis, a statement like "field X has lots of errors and irreproducibility" is non-falsifiable, and indeed will be true no matter what, so he is taking no risk with his criticisms -- even though they do more damage to public perception, morale, and funding, and help less to change things than targeted criticisms would do.

If I sound irritated by this sort of thing, it is because my graduate advisor dabbled in "meta-science" / "error correction" and I stoutly refused to have anything to do with those projects. I thought, and still think, that it is considerably more valuable to try to build my own solid lego tower despite the difficulties rather than knocking down other people's, despite the latter being considerably easier and a fairly simple and reliable route to a high citation count. In the end, irreproducible or erroneous papers in science just get ignored rather than retracted, and I think that's fine.

I would use Andrew Gelman and John Ioannidis as prototypes for the Good Critic and Bad Critic, respectively.

Gelman does real, solid work and confines his righteous takedowns to a side hobby. And his criticisms are directed at individual, specific cases, with evidence, and he reserves most of his wrath for repeat offenders rather than one-off mistakes, which are unavoidable given enough projects. He confines criticisms to his area of expertise, which is statistics.

Ioannidis' claim to fame is writing a hand-waving, philosophical argument in order to cast doubt on all research at once. Followed up shortly by a paper that implied (but didn't directly state!) that regression to the mean on replication implies error or fabrication in the original. And he has ridden this sort of tired argument into a lucrative and prestigious career and Stanford appointment. Never mind the collateral damage on funding and morale. My theory is that people pay attention to Ioannidis and wring their hands about these highly generic issues for the same reason most people pretend to be highly, highly concerned about sexual harassment/discrimination in the workplace -- they're praying that if they act concerned enough, the hammer will fall somewhere else. But indiscriminate criticism like this is highly toxic to science (or anywhere else, for that matter, and has the side effect of letting real bad actors off the hook).

A further moral of the story is that, for better or worse, I don't trust and generally ignore criticisms that do not come from someone who has invested time into a field and made useful contributions to it. 'Full-time critics' strike me as generally sleazy and opportunistic, and I doubt the sincerity of their desire to 'improve' a field if all they can do is tear down others' work and not provide any of their own.