jpeloquin

I'm having difficulty reconciling these two opinions:

1. This ethics guidance is bad because it limits, or even censors, scientific investigation related to socially sensitive topics.

2. Social science has a reproducibility crisis and is mostly useless (a viewpoint commonly expressed in response to other articles).

Does anyone hold both of these opinions, or is it just different people talking at different times?

> You could write a book on this point. In fact a few have and they are worth reading.

If you have time, could you share some specific recommendations?

The practice of plotting case maps like this started during the 1854 cholera epidemic. The cholera case map helped discredit the theory that cholera was transmitted by miasma, and supported the theory of fecal-oral transmission. Just some related history that you might find interesting.

https://en.wikipedia.org/wiki/1854_Broad_Street_cholera_outb...

We just call this "literature review", and everyone does it at times regardless of seniority level. But the aim is not a sentiment graph because it doesn't really matter whether author A said paper B was good or bad. What would we even use that information for? We instead want a comprehensive list of papers that are relevant to what we're trying to do right now, with a note on each saying why it is relevant.

Not sure how to interpret the question. Using a robot vs. a person doesn't change anything about the physics of how tissue responds to perturbation. You can of course simplify the simulator and correspondingly limit its use to specific learning objectives, probably novice-level. To return to the thread's starting point—we understand flight well enough to make general-purpose flight simulators, we don't understand tissue well enough to make general-purpose surgical simulators. You can still make simple interactive training tools to help someone rehearse the motions of a procedure, but I wouldn't call such a tool a simulator.

Yes, you can produce a model incorporating nonlinear elasticity, viscosity, plasticity, frictional contact, subfailure damage, and fracture / cutting. Repeat for each tissue involved. Some of those methods are well-developed, others are not. Damage and failure in particular is poorly developed, and simulating deformable body contacts can be tricky. Then you have to estimate parameters for all of that, validate that the model produces realistic outcomes, and get it to run in real time. You can ignore most of the complexity and make approximations, like treating everything as an isotropic linear elastic material, setting cutting resistance to a fixed value per tissue, treating the scalpel like a lightsaber, and accepting that contact will be imperfectly enforced (surfaces may interpenetrate). A crude simulation can still be good enough to be useful as a learning tool though, but it tends to work more like a video game than a true physics-based simulation. We'll still eventually get the kind of accurate simulation you're asking for, probably.

Yep, acknowledgments in the 2017 paper linked by the MIT article check out as referencing the grant I linked, and the paper does describe work with chimeric viruses.

> during 2014-2019 NIH actually paid $3.6M to EcoHealth Alliance under the grants clearly specifying the coronavirus GoF experiments in Wuhan with the "Human subjects included" checked.

Is this is the grant you're referring to? https://reporter.nih.gov/search/me9X23pZWEigATS-4XTBfA/proje...

From your comment it was unclear to me what the human subjects component was. It appears to have been population surveillance using PCR and antibody testing, with the intent to detect prior infection by previously unrecognized coronaviruses. Which is low-risk.

From the public project details, it is not clear that the scope of the work as described to the NIH included gain of function experiments. Perhaps you were referring to a different project description? And of course EcoHealth Alliance may have done gain of function experiments that were not described in their proposal.

> I’d be very interested in a link, that sounds cool!

Paris' Law, which predicts fatigue crack growth rate, was published in "The Trend in Engineering", which is the University of Washington College of Engineering alumni newsletter.

“Though Paris’s ideas presented a way to quantify and predict structural integrity, his approach was considered unconventional, and none of the leading journals published it. So, in January 1961, after leaving the UW to return to Lehigh for his Ph.D., Paris published his paper, ‘A Rational Analytic Theory of Fatigue,’ in The Trend.”

https://www.engr.washington.edu/news/article/2021-05-24/pari...

https://www.engr.washington.edu/news/trend

Start your own journal? It doesn't necessarily matter that citations to it won't be tracked in Scopus or whatever when it's first launched. Even the gatekept academic journals aren't automatically accepted into citation databases and academic search engines, it usually takes several years after they're launched to be fully accepted. Or discard the journal framework entirely—as long as people who are interested in the work can find the work, it's fine.

Edit: One of the most important papers in fracture mechanics was published in a campus newsletter. The publication venue doesn't matter that much.

The "Unreasonable Effectiveness of Multiple Dispatch" talk is a good example of how multiple dispatch is special in a good way, in that everything (should) work together as new types and functions are added to the ecosystem. However, this also means the scope of potential integration bugs encompasses the entire ecosystem. The Julia manual has a small section about special composibility pitfalls arising from multiple dispatch: https://docs.julialang.org/en/v1/manual/methods/#man-method-...

As best as I can summarize it: Multiple dispatch is supposed to dispatch a function call to the implementation with the most "specific" call signature. This means that you must design your functions with an eye to what everyone else has implemented or might implement so whatever function gets called does the "right" thing, and also that your implementation doesn't block someone else from writing their own implementation specialized to other types. This requires some coordination across packages, as shown in one of the manual's examples.

The rules defining type specificity (subtyping) are complicated, and I think not in the manual. They have been inferred by observation: http://janvitek.org/pubs/oopsla18a.pdf. To quote from that paper, "In many systems answering the question whether t1 <: t2 is an easy part of the development. It was certainly not our expectation, approaching Julia, that reverse engineering and formalizing the subtype relation would prove to be the challenge on which we would spend our time and energy. As we kept uncovering layers of complexity, the question whether all of this was warranted kept us looking for ways to simplify the subtype relation. We did not find any major feature that could be dropped." Julia's multiple dispatch allows a high degree of composibility, but this does create new complexity and new problems.

In the US, federally funded research is already required to made available without paywalls within 1 year of publication. Open-access advocates online generally seem to be unaware of this, and that current goalposts are therefore (a) immediate open access, and (b) open-access publication of all scientific research regardless of funding source.

https://obamawhitehouse.archives.gov/blog/2013/02/22/expandi...

https://www.nih.gov/health-information/nih-clinical-research...

Yeah, when I speak or write out my thoughts I often find they weren't as fully formed as they seemed while they were just in my head, and maybe shouldn't have counted as thoughts in the first place. This is one reason it's beneficial in a work context to keep a project journal or regularly discuss work with other people.

It's a personal statement from someone with relevant lived experience. Exploring others' viewpoints is interesting, if only to find out that they exist. The article touches on questions that do not currently have an objective answer: Why do people do the things they do? Should they be doing something different? Science may someday provide an objective ("substantive") answer to the first, but it can't choose our goals (make value judgments) for us, so articles like this are valuable. And social media works a lot better for opinion-based than objective discourse because everyone can participate on equal footing.

I remapped control to the bottom row, whichever key my thumb naturally rests on. Usually alt. Also moved \, [, ], enter, and backspace to avoid reaching for them with my right pinky. I think if your fingers start feeling even a little bit weird, you should keep remapping keys until they feel good all the time. Doesn't make sense to wait for it to get worse before taking action.

Thanks for your reply. Not providing numerical estimates is fine with me, my opinion is that they're not productive anyway. It's like the Drake equation but worse, or trying to estimate the probability of divine creation vs. a natural origin for the universe. Too many unknowns, no ability to test counterfactuals. We are not even close to inventorying all relevant natural coronaviruses, so we don't have a good estimate for the number of "draws" to assume against the natural origin, and we don't do the routine monitoring that would allow a good estimate of probability for potentially pandemic pathogen → actual pandemic. So an armchair analyst ends up with a solid "maybe", with massive uncertainty intervals, for any explanation.

I completely agree that WIV's practices were dangerous and irresponsible, and that they are a plausible origin at all is a crushing condemnation. There's an old piece of advice: imagine that what you're doing ends up on the front page of the newspaper, and consider whether you would be proud or ashamed. If the latter, do something else. WIV clearly didn't follow that advice.

The facts in your post are all correct to my knowledge, I only quibble with the emphasis on infectiousness. The uniquely dangerous aspect of SARS-CoV-2 was its several-day period of transmissibility in the absence of symptoms, not its infectiousness, which for early strains was not that remarkable. Without asymptomatic transmission it would probably have been contained and quickly forgotten, like SARS-CoV-1. We need to adopt methods that will detect and contain viruses like SARS-CoV-2 so we're better prepared for when SARS-CoV-3 shows up.

One aspect of the intellectual approach that I struggle with is decision paralysis. That is, second-guessing any expenditure of resources I'm considering as possibly better spent on something else. And there are innumerable options. Empathy or (perhaps more usefully?) a sense of duty to specific people or organizations helps break that indecision, at least for me.

Sure. P(lab leak origin | last bullet) = P(last bullet | lab leak) * P(lab leak) / P(last bullet). There isn't much evidence to constrain each term, and your post doesn't make any attempt to estimate them, but let's make some stuff up. The values change a lot if you take P(lab leak) to be any lab leak, or specifically a modified RaTG13 from WIV.

Let's go with the more specific scenario. Then P(last bullet | WIV lab leak scenario) becomes P(outbreak | WIV lab leak &c.) because we've conditioned on everything but the outbreak part. I'll guess this is 0.001, which is generous because it's very unlikely for a mere exposure to turn into a pandemic. Exposures to natural potentially pandemic viruses happen every day, most don't become a pandemic. P(WIV lab leak scenario) is difficult. There are a lot of different lab leak scenarios that could have happened instead, but we also don't care that much about trivial differences. Let's be really generous and say P(WIV lab leak scenario) = 0.25 because it's one of I think 4 plausible covid origin sites in Wuhan. Probably it should be lower because it's plausible that Wuhan was only the site of the first superspreader event, not the origin. And let's say P(outbreak) = 0.2 because we seem to get a potential pandemic every 5 years or so. Then P(WIV origin | outbreak) = 0.00125.

There are all sorts of problems with my analysis of course. Even if I haven't made any mistakes, the numbers I conjured up are entirely unconvincing. With only one datum, we have hardly any objective support for our probability estimates, and Bayesian analysis only gives a semblance of objectivity to our prior opinions.

Edit: I have no desire to support either side here, just to point out that probabilistic reasoning doesn't help much here. Whatever happened, happened, regardless of how (un)likely it was.

Gell-Mann Amnesia Effect? Although that's more about how news and social media are unreliable sources of information on all topics. I've heard "Engineer's Disease" or "Engineer Syndrome" used to describe the tendency of so-called technical people to think their expertise generalizes to topics in which they have no experience.

> If an idea is successful, whether correct or not, then as in business people will blindly support it to 'get a piece of the pie'.

I wonder whether articles like this are driven in part by the same copycat effect described above. People love to complain about their workplace. Although the article identifies real problems, it is lean on evidence for the magnitude of their impacts. Many of the citations are to other opinion pieces. It's certainly worth improving things where we can, but the sky is not falling either.

For example, the abstract leads with "If a critical mass of scientists become untrustworthy, a tipping point is possible in which the scientific enterprise itself becomes inherently corrupt and public trust is lost, risking a new dark age with devastating consequences to humanity." Which sounds serious. But based on articles posted to Hacker News previously, 1–10% of scientists (excluding countries with direct cash payments for accepted articles) fabricate or falsify data, which is a typical percentage of bad actors in most groups. The other 90% still appear able to serve as quality control in the usual way. People can tell if an article is likely to be b.s. with about 70% accuracy based only on its description, not even reading it. I think any defects in science feel more severe than they really are because science has been so successful that our expectations are now sky-high.

I do emphatically agree that promoting altruistic and ethical norms should be a focus of scientific institutions. Major discoveries will arise from curiosity and serendipity, as they always have, they cannot be predicted in a proposal's Gantt chart.

Apologies for implying that scientific employment outside of the US is easy to obtain. I should have clarified that my comment pertained only to the US scientific job market, which enjoys considerable support from public and private funds.

Everything on this list is nice in general; my favorite is #5. I suspect that UBI would not make any difference to the "Karikó problem" though. Labs are almost always desperate for people who can get things done. If you don't mind not being a PI and are ok with a relatively small paycheck (but still much greater than what UBI would provide), it's fairly easy to maintain scientific employment that still allows you to pursue your own side projects. Perhaps we should call persisting as an individual contributor, not in a PI / management role, the "Karikó solution".

> I am trying to get good data to determine if vaccination is risky or not.

Rather than giving you a list of work that you may suspect of being curated, I will show you how to quickly find what you're looking for.

One of the great things about the National Library of Medicine is that a human looks at each and every paper and assigns standardized semantic tags to it. These are called "MeSH terms" and you can filter the entire library by them.

You mentioned interest in myocarditis, and the following would be an appropriate search query in pubmed.ncbi.nlm.nih.gov for this topic:

myocarditis[mh] AND vaccines[majr] AND COVID

https://pubmed.ncbi.nlm.nih.gov/?term=myocarditis%5Bmh%5D+AN...

The "mh" selects articles that were tagged with that MeSH term. The "majr" selects articles that were tagged with that MeSH term and were mainly about that term. In this case, the structured query drills down from thousands of potential results to 61 results, which is manageable to triage manually.

Edit: Here's how you can figure out wish MeSH terms exist: https://meshb.nlm.nih.gov/search

> T-cell immunity was considered borderline conspiracy theory until recently.

This surprised me, so I checked my cell biology books. The role of memory T-cells in the adaptive immune response is clearly described in Garland, "Molecular Biology of the Cell", (c) 2002. Not sure in what sense T-cell immunity is a conspiracy theory, unless it's being used as code for something else.

In my experience this is very rare, but that's my plan going forward. I can't guarantee that seed will stay up if I'm not paying attention to it though, so in practice it still means emailing the corresponding author. And actual transfer may still require sending hard drives through the mail. But publishing a torrent at least indicates willingness and preparation to share data, which is worth something.

> One of the questions I've had to really grapple with is that I just cannot personally host all the terabytes of data I'm collecting.

Hosting is a big challenge for open science. Whenever I see someone calling for open data, I feel half enthusiasm for the movement to cooperate more and half dread at the prospect of trying to comply. I used OSF to publish one of my datasets, but they instituted storage limits that would prevent doing the same in the future. And exposure to thousands of dollars in surprise fees from personal archival in cloud hosts isn't acceptable. Which leaves the status quo of email the author for data, hope they respond, as disappointingly the best option.

Open code is trivial to provide in comparison, but also less useful.

As a researcher, although I love the idea of a pressure-free job, I'm also keenly aware that I'm being paid by tax money and therefore should be accountable for my productivity (or lack thereof). I would rather have more robust review of research activities, and reliable consequences for misconduct, than freedom from financial pressure.

I am also somewhat skeptical that removing funding pressure will reduce fraud. It may reduce the amount of show publications, i.e, work that doesn't really make any progress but isn't wrong either. But in my experience outright falsification of results comes from entitlement. A fraudster feels that they entitled to a certain level of achievement and recognition for it, and are willing to make something up when they're not getting that recognition. I believe that the pressure that drives a person to commit fraud is internal, not external.

> Give universities a blank cheque to employ scientists to work on whatever for sciences sake, not for business.

Personally, I favor the opposite—stop awarding universities the "indirect costs" part of grant awards unless they agree to cap the hiring of grant-seeking investigators. The number of grant applicants needs to be balanced with the supply of public funds, and I don't think the public is willing to pay much more than they already are. And I say this as someone who is currently grant-funded.

The use of 7–9 year training periods (5 years PhD + 2–4 years postdoc) means a journal article's author and their scientific readership have a lot of shared knowledge and culture, which probably increases the reliability of communication. I agree that pre-publication feedback from peers also helps.

It's worth noting that the two provided references disagree on their definitions of reproducibility. Plesser quotes ACM as reproducibility being tested by a "Different team, different experimental setup", consistent with use in the GP's comment, but ultimately seems to favor ignoring "replicability" and instead using "reproducibility" in combination with various adjectives. The common understanding in the physical sciences is also that merely re-running the same code is not a test of reproducibility.