More Recent Comments

Monday, January 20, 2014

Not enough authors?

There's a big difference between publishing the complete sequence of a genome and having a highly accurate "finished" version that's fully annotated. You may be surprised to learn that there aren't very many high quality genomes of eukaryotes—especially vertebrates.

That's why I was interested in a paper published last April on the zebrafish genome. The authors have produced a high quality reference genome that will serve the scientific community (Howe et al. 2013).

Sequencing and assembly are highly automated and there are several programs that will find genes and other interesting bits of a draft genome. It's a lot more work to finish off the sequence by filling the gaps and it's even more work to annotate and check the sequences. Much of this work is labor intensive and expensive and that's why there are so many unfinished sequences in the literature.

I wasn't surprised to see that the original paper on the annotated zebrafish genome had 171 authors although that did seem a bit excessive. It meant that each author contributed an average of 0.6% to the final result. Some of them must have made a much smaller contribution. I wonder if every author read and approved the paper before publication?

Apparently there weren't enough authors. The January 9, 2014 edition of Nature contains a Corrigendum (correction) to the zebrafish paper. Five other authors were "inadvertantly ommitted" from the list bringing the total to 176 authors. In addition, the names of three other authors were spelled incorrectly in the original publication last April. I don't know why it took eight months before anyone noticed.

That just goes to show you that modern scientists have to deal with problems that us old fogies never encountered. I never had to spent more that a few seconds writing down the names of the authors on any of my papers. Today you need data management software to keep track of your authors.


Howe, K. and 171+5 others (2013 The zebrafish reference genome sequence and its relationship to the human genome. Nature 496:498–503. [doi: 10.1038/nature12111]

38 comments :

Veronica Abbass said...

"Today you need data management software to keep track of your authors," and the people you have to thank or acknowledge.

Ed Darrell said...

Help me out here: The chart shows the genes in common between these genera, generally? So, from this chart, I could say that the average human shares 10,660 genes with chickens, mice and zebra fish?

But saying the better human genes are those 10,660 is too much of a judgment call to make from these data, right?

Faizal Ali said...

There's another issue that has come up at the centre where I work. Compensation is partly based on the number of publications you produce and the impact factor of the journal in which they were published. So we have the situation where someone might get more money for being one of 170 authors in a Nature article, than another individual might receive for an original piece of research that required much more work but appeared in a less prominent journal.

Pedro A B Pereira said...

Are you in the US?

Faizal Ali said...

No, Canada. And the situation I describe may be atypical. I work as a medical doctor at a teaching hospital. I do no research myself, but part of my pay (for clinical work) is deducted and put into a pool that is used to compensate doctors for research and other academic activities. There are a number of formulas that are used to determine how this is divided up among those eligible for it, and what I describe above is one of these.

Georgi Marinov said...

OK, so what is your suggestion for dealing with these kind of situations?

Sure, something like a third of these authors may not have done much of substance for this paper. Let's make it a half. That still leaves 80+ authors, and if 170 is too big of a number, it would seem to me that 80 would be a big one too.

Perhaps they should have had single-digit number of authors for the HGP too?

Yes, these days a single person, who is capable of both building his libraries and running assembly software (they exist) can get a crappy highly fragmented assembly of a vertebrate-sized genome. But this is not what we're talking about here.

Pedro A B Pereira said...

Seems atypical to me, alright, but maybe it's common practice these days in medical research conducted at hospitals? Somehow I doubt it. I work in a more traditional academic setting, and that kind of thing is totally non-existent (at least until some suit-wearing idiot from some business school decides otherwise).

Uncivilized Elk said...

"OK, so what is your suggestion for dealing with these kind of situations?"

I didn't get the impression that Dr. Moran was saying they had to be dealt with in some sort of fundamental way. Hell, he wrote "I wasn't surprised to see that the original paper on the annotated zebrafish genome had 171 authors although that did seem a bit excessive".

What's with the victimized/sarcastic tone?

Alex SL said...

There is no way that 176 people have all contributed enough to the same paper to deserve authorship unless that term is made a mockery of.

Larry Moran said...

I understand why some work needs dozens of authors who have made substantial contributions. On the other hand, I believe that people can make minor contributions by just helping out or providing advice without deserving to be authors. I also think that if you hire people to do annotating and pay them a salary, it doesn't mean that they should be authors on the paper.

If you had nothing to do with designing the experiments, deciding what to do next, analyzing the results, discussing the conclusions, raising the money, managing the project, or writing the paper, then you probably shouldn't be an author.

If you didn't even notice that your name wasn't on the manuscript or that it was spelled incorrectly then that's a clue that it shouldn't have been there.

Georgi Marinov said...

If you didn't even notice that your name wasn't on the manuscript or that it was spelled incorrectly then that's a clue that it shouldn't have been there.

We're in agreement on this. You can also add "not being aware o the existence of the paper", which has also happened

caynazzo said...

Larry:
"If you didn't even notice that your name wasn't on the manuscript or that it was spelled incorrectly then that's a clue that it shouldn't have been there."

I disagree and not just because not noticing your name on a paper is a superficial metric. A lot of those authors are probably technicians working in large high-throughput labs. The work they do is highly specialized and a part, an integral part, of the entire sequencing/validating pipeline. Or maybe they're bioinformaticians who wrote some unique code that's being published for the first time.
Here are the NIH guidelines for authorship.
http://sourcebook.od.nih.gov/ResEthicsCases/guidelines_for_authorship_Contributions.pdf

Jonathan Badger said...

I haven't been on a paper with quite so many authors, but certainly a couple dozen is par for the course in genomics. And in regard to Larry's, "I also think that if you hire people to do annotating and pay them a salary, it doesn't mean that they should be authors on the paper.", no -- it means *exactly* that. Annotation is a major part of a genome project and it *must* be credited to be fully ethical. Actually, these days I find the quality of annotation is going down largely because there isn't enough manual annotation going on and too much of it is just running computational annotation methods (and I say that as a author of such pipelines; the correct way to use them is to aid manual annotation, not to replace them).

Jonathan Badger said...

For proposals to the US NSF, you have to include a biosketch consisting of your education, relevant papers, and the names and affiliations of everyone you've co-authored a paper with in the last 48 months. And it has to be done in two pages. Which makes life hard for anyone in a highly collaborative field like genomics.

Alex SL said...

caynazzo, Jonathan Badger,

The word 'author' has meaning; it is usually understood to have something to do with writing. In the case of scientific papers, one would also expect the authors to be the scientists who came up with the study design and/or did the analysis.

If we are talking about a technician or informatician, well, as Larry Moran mentions there are salaries for that and acknowledgement sections.

Sure, this project would not have been possible without technician A running the sequencers and that one would not have been possible without specialist B building a crucial piece of novel equipment for the particle accelerator.

But if you think about it, it would not have been possible without train driver C bringing the main investigator to work every day, nor without admin staffer D bookkeeping the grant money, nor without sweeper E keeping the lab clean. I don't see these three on the author list either, and for a good reason.

Georgi Marinov said...

The word 'author' has meaning; it is usually understood to have something to do with writing. In the case of scientific papers, one would also expect the authors to be the scientists who came up with the study design and/or did the analysis.

How exactly do you suggest that 200+ people write a paper together? I have seen how that works with 20+ people. It is not a pretty sight. Once you get past 3-4 actively participating in the writing process authors, the process just breaks down and it becomes like pulling teeth - it gets passed around for multiple revisions and everyone hold onto it for at least a day so each round takes weeks, it's impossible to agree on everything so people start arguing over each and every word and sentence. It's a nightmare.

The only way to actually do this is to have a few people write up the results that have been produced collectively. Are you suggesting that the people, who actually did the bulk of the actual work but did not participate in the writing should not be authors just because they did not put words on paper?

Alex SL said...

If asked, I would suggest that (a) only people who made a significant intellectual contribution to the study should be authors, (b) only people who made a significant contribution to writing should be authors, and (c) everybody who made a significant intellectual contribution should participate in writing. Problem solved.

More generally, having that many authors is simply ridiculous for the reasons that LM outlined right from the start: you end up with everybody having made a contribution of 0.6%. And that is on average; given that a handful of them will have done much more than that there will be dozens whose contribution will be less than 0.1%.

And then you have dozens of people strutting around going "I am on a Nature paper, and it was cited a hundred times!" because they held a pipette at some point. At the same researchers from other areas of science and scholarship are looked down upon because they publish single or two author papers that only get seven or twelve citations. (Divide citations by authors for both cases and contemplate what that means!) It is ludicrous.

We should not ask ourselves, did this person contribute something to the paper? We should ask ourselves, did this person make a significant intellectual contribution to the paper?

Again, what do people in those areas think acknowledgement sections were invented for?

Georgi Marinov said...

If asked, I would suggest that (a) only people who made a significant intellectual contribution to the study should be authors, (b) only people who made a significant contribution to writing should be authors, and (c) everybody who made a significant intellectual contribution should participate in writing. Problem solved.

Then good luck finding people willing to work on these kind of projects..

And then you have dozens of people strutting around going "I am on a Nature paper, and it was cited a hundred times!" because they held a pipette at some point.

This is assuming those evaluating these people at later stages are stupid enough to weigh the huge collaborative papers the same way as single-lab first-author papers. This is a false assumption.

We should ask ourselves, did this person make a significant intellectual contribution to the paper?

As strange as it might sound, there are papers to which hundreds of people indeed made a significant intellectual contribution. Even if it is not reflected in the actual text of the final product. You don't go from point A to point Z without going through B, C, D, E, etc.

Alex SL said...

Then good luck...

It seems to work in other areas. It is mostly genomics and particle physics where they believe that everybody who has vaguely looked into the direction of the project should be made an author.

This is a false assumption.

As a career scientist I have seen enough Nature! Science! PNAS! fetishism to make more than an assumption here. Sure, being the first or last author is much more impressive than #63 of 120. But how often have I heard "he has published three Nature papers in the last five years"...

And that is not even mentioning how citations and h-index are used for evaluating candidates and staff. With the single exception of the biological faculty of my alma mater I have yet to see an institution that takes author numbers into account. No, generally 20 citations on one of 'your' papers count 20 towards your stats regardless of whether you were the first of two authors or somewhere in the middle of dozens. This is a massive distortion disadvantaging people from fields with less inflationary authorship policies.

there are papers to which hundreds of people indeed made a significant intellectual contribution

I maintain that this is not physically possible without doing violence to the word 'significant'. Yes, we stand on the shoulders of giants, but if significant B, C, D and E are generally their own publications because in that case A-Z would make a manuscript five thousand pages long.

For comparison, I could do little in my line of work without the Polymerase Chain Reaction. That does not mean that Kary Mullis or Kjell Kleppe should be my co-authors on every second paper; they don't have anything to do with the design of the studies I am using their invention for.

Georgi Marinov said...

It seems to work in other areas. It is mostly genomics and particle physics where they believe that everybody who has vaguely looked into the direction of the project should be made an author.

Which other fields have huge collaborative projects of this kind?

No, generally 20 citations on one of 'your' papers count 20 towards your stats regardless of whether you were the first of two authors or somewhere in the middle of dozens.

Citations are a problem, that is correct. I personally have over a thousand, accumulated in little over 2 years. But we all know why I have that many.

However, focusing on this is missing the bigger point, which is that it is deeply misguided to count citations when evaluating individuals. I don't think the top institutions do that - I have the impression that they are above that sort of thing, but I may be wrong, I don't hang out in the circles that make these decisions so I don't know how it actually goes. But it definitely seems that there is an inverse correlation between the rank of the institution and the attention people pay to citations - because I do see citations prominently displayed on CVs and websites by some scientists and not by others (think how much time it takes to track those things on your own and constantly update them...)

I maintain that this is not physically possible without doing violence to the word 'significant'. Yes, we stand on the shoulders of giants, but if significant B, C, D and E are generally their own publications because in that case A-Z would make a manuscript five thousand pages long.

They are their own publications most of the time. But they are also essential for the big one and they would not have happened without it. I don't think you understand how these things work - there is a big project and as a part of that big project lots of people make a lot of small advances (each of which contributes to and is important for the whole). They publish their own papers on them, but then they also get to be authors on the big consortium paper, which summarize everything (and yes, we can debate what the point of having a big paper is if all the actual details of the work are elsewhere, but this is a different discussion). This is very different from your PCR example - none of the consortium-related work would have happened without the consortium.

Alex SL said...

It should be clear from what I have written that I am at least as sceptical of this number crunching as you are. Well, I have to keep track of my citations and h-index for annual evaluations, that's how it is.

I see where you are coming from but sadly you will not be able to convince me. Because of the understanding I have of the words 'significant', 'intellectual contribution' and 'author' I consider a claim such as there are papers to which hundreds of people indeed made a significant intellectual contribution to be self-evidently absurd, comparable only to hundreds of people fit into a space of two cubic metres. Maybe your understanding of these words is simply a different one.

It is way past midnight here so I'm out...

Jonathan Badger said...

"Again, what do people in those areas think acknowledgement sections were invented for?"

Acknowledgements are for minor things -- for example people who helped you run a program or taught you a lab technique -- so maybe 2-3 hours of effort *top*. It isn't ethical at all to merely acknowledge somebody who has worked full time on a project for months merely to keep the author count down. Particularly when it is doing something intellectual like manual annotation, which requires detailed knowledge of biochemical pathways and interpretation of the output of bioinformatics tools and not merely pipetting fluids like a mindless robot.

I've been on "big science" genome papers with 50 authors as well as "small science" systematics papers with 3, and there really is much more real intellectual effort to the genome papers (at least well done ones where the data is really analyzed rather than merely BLASTing it and calling it a day, which unfortunately is becoming more popular as genomics grants get smaller and manual annotation becomes impossible due to lack of labor support). A good genome paper has as much analysis behind it as a 100 small science ones -- even if it gets published in a "glamour mag" like Nature and most of the work gets hidden as "supplemental information"

"The word 'author' has meaning; it is usually understood to have something to do with writing."

In the humanities, yes, but not in science. Writing a manuscript is probably less than 5% the effort of a typical science project. It needs to be done, just like grant proposals need to be written, but it isn't what science is about.

caynazzo said...

Alex SL
"The word 'author' has meaning..."
Correct, and those who wrote the NIH guidelines have clearly given it more thought than you.

"http://sourcebook.od.nih.gov/ResEthicsCases/guidelines_for_authorship_Contributions.pdf"

Alex SL said...

Jonathan, caynazzo

You will find that there are more considerations than the ones you carefully chose to respond to. For 'working full time on a project for months' we also tend to get these things called 'salaries'.

Even if more intellectual effort goes into a genomics paper (and we could have a productive discussion whether annotation is really comparable in intellectual heft to study design) that still does not change the fact that it is physically impossible in principle for more than ca. twelve people to make a contribution to the same project that can be considered significant without first changing the meaning of the word significant.

None of your points address the distortions in evaluation of scientists' CVs between fields discussed above: this inflation in author numbers also inflates the perceived impact per researcher.

The linked NIH guidelines actually make my point for me: 'writing & other: none -> no authorship'. Could not be clearer. There is no way that, say, 176 people have all contributed to writing, much less something substantial.

Jonathan Badger said...

Nobody works in academic science for salaries alone; it just isn't worth it. There is a place where people do science without need for public credit. It's called the pharmaceutical industry -- I was there myself for a few years after my postdoc. But the salaries are much, much higher there to attract talent despite the lack of recognition.

That still does not change the fact that it is physically impossible in principle for more than ca. twelve people to make a contribution to the same project that can be considered significant without first changing the meaning of the word significant.

Where's the evidence to support this entirely handwaved assertion? I could just as easily say if you can only gather a measly 12 people to work on a project, it probably is a pretty dull or trivial one.

None of your points address the distortions in evaluation of scientists' CVs between fields discussed above: this inflation in author numbers also inflates the perceived impact per researcher.

I think you need to show it's inflation. Collaborative science is the future, like it or not, and not just in genomics and particle physics. In The US, both the NIH and NSF have begun to launch interdisciplinary programs with the goal to fund proposals involving massive collaboration across departments and institutions. Such proposals will probably have at least a dozen PIs, not to mention the numerous postdocs and grad students working under them.

Georgi Marinov said...

Collaborative science is the future, like it or not, and not just in genomics and particle physics.

This is true in general but in genomics, the trend is opposite - with increased automation and decreased cost of sequencing, fewer and fewer people are necessary to do what took whole genome centers in the past. That's why these days you see eukaryotic genome being published by just a handful of authors. The number of authors is creeping up and up otherwise though, that's entirely correct,

Alex SL said...

Well of course whether it is inflation is just the point of disagreement: Are there more people on the papers because two areas of science have magically managed to increase the number of percent that make up a whole or because two areas of science have lowered their standard for what constitutes a significant contribution?

I see that my arguments are not going to be addressed; instead you merely repeat that collaborations are necessary, a point I do not dispute. It boils down to this:

significant (sɪgˈnɪfɪk(ə)nt/), adjective, 1. sufficiently great or important to be worthy of attention; noteworthy. 2. having a particular meaning; indicative of something.

When talking author lists, the first definition applies. If 100 authors have contributed equally then each has done 1% of the work. In that case, there is no way to read the meaning of the term to avoid the conclusion that none of them has contributed anything significant.

In reality, a small number of researchers will have designed the study and written the paper but had various technical aspects (annotation, sequencing etc) done by other people. Only the first type of work is the actual intellectual core of a scientific study that an authorship should normally be awarded for.

But even assuming that all kinds of contributions should count (although in that case one wonders why university janitors and night guards don't make author list) that means, for example, that five researchers may have done 10% each (significant) and the remaining 95 people have each contributed ca 0.5% (too little to be worth mentioning). The math is not actually hard to grasp.

And once more, the fact that a certain minuscule contribution was essential doesn't change the situation either because the grant administrator was also essential and she doesn't make author list.

Georgi Marinov said...

When talking author lists, the first definition applies. If 100 authors have contributed equally then each has done 1% of the work. In that case, there is no way to read the meaning of the term to avoid the conclusion that none of them has contributed anything significant.

There is an obvious fallacy in this argument and it is that it rests on the assumption that the total amount of work that goes into each paper is constant and equal. Not only it isn't but there is orders of magnitude variation between different papers.

Alex SL said...

We have a failure to communicate. The question isn't how much work it is, the question is how much work it is relative to the whole.

Georgi Marinov said...

I addressed exactly what you said and meant.

People work on these consortium papers full time for years and do just as much work for them as many "small"-paper first authors. That it ends up being some small percentage of the whole is due to the fact that the whole is so much bigger.

Also, the people who write these papers are the PIs, who usually have done very little actual work other than the writing, and that includes the real intellectual contributions. Who do you think develops the algorithms and the experimental protocols? It is those people who did not put a single word on paper but instead only provided figures, and sometimes not even that.

You have to have a very twisted sense of fairness in order to claim that they should not be in the authors lists and only the PIs who got together and wrote the text should be...

caynazzo said...


Alex SL
"The linked NIH guidelines actually make my point for me: 'writing & other: none -> no authorship'. Could not be clearer. There is no way that, say, 176 people have all contributed to writing, much less something substantial."

That's not what the guidelines say at all.

It appears you're reading the categories (e.g., writing&other) in the NIH guidelines as interdependent, which does explain your views on authorship as an equal division of labor but it also doesn't make any sense. Some of those categories for authorship imply no manuscript writing, such as data analysis. If your contribution qualifies for Green in any category, boom, you're a co-author.

Alex SL said...

GM: I addressed exactly what you said and meant.

Yeeees... the fact that I had all those percentages in my argument clearly means that the size of the whole was what I cared about. You got me there.

caynazzo,

That interpretation seems rather odd. Are you suggesting that if I send samples to a company in Korea to be sequenced I need to put their techie onto the paper? Why? If not, why should somebody who has been hired by a researcher to do 0.5% of the work on a project necessarily be on the paper?

You also seem to be suggesting that somebody should be an author who has contributed absolutely nothing whatsoever to the writing, nor, and that is perhaps even more important, anything even remotely intellectual. Apart from making a mockery of the term 'author', have you thought through what that means for responsibility?

If you are the co-author of a paper you have basically signed the study off with your name. If the scientist who coordinated it all fudged the data to make them more interesting somebody who only contributes a minor technical step will have no way nor even the qualifications to realize that. But they might still have their career damaged when the fraud is discovered because, again, they signed it off.

Georgi Marinov said...

That interpretation seems rather odd. Are you suggesting that if I send samples to a company in Korea to be sequenced I need to put their techie onto the paper? Why? If not, why should somebody who has been hired by a researcher to do 0.5% of the work on a project necessarily be on the paper?

There is a big difference between sending something for sequencing and consortium work. Sure, some of it involves exactly that kind of work, But there is an enormous amount of real intellectual work that goes on, as I already mentioned. Algorithm design and development, development of new experimental protocols, etc. The people who do that work rarely contribute text to the final big paper - some of them make figures, but that's it. Are you suggesting they should not be on the paper?

Alex SL said...

Yes, I suggested that repeatedly. Do you know identification keys, as used to figure out the names of plants and animals? Here is one for figuring out what to do:

1a. Your contribution is a minuscule part of the whole project ... 2
1b. Your contribution is significant (>5%?) ... 3

2a. Your contribution nonetheless constitutes novel intellectual work ... No authorship but you should publish your innovation separately.
2b. Your contribution is purely technical and routine such as sequencing or annotating ... No authorship (or the janitor and the Korean lab should get one too).

3a. You participated in writing and understand the paper to the degree that you can vouch for the correctness of its conclusions ... Authorship.
3b. You did not participate in writing and cannot assess the validity of the conclusions ... No authorship, it would not be ethical for you to sign the paper off.

This is not always how it works out in reality, I am merely saying how it should work in an ideal world.

May I also suggest that you are confusing what is now established practice in genomics with what would make sense from the perspective of appropriateness, fairness, intellectual consistency and personal accountability?

Alex SL said...

I should add that it is a really dickish move if the coordinating scientist(s) exclude somebody from paper writing and then say they shouldn't get an authorship because they did not participate in writing. I have seen that happen. Obviously everybody who has made a significant contribution should be asked to help write the paper. But if they then don't, how can you possibly give them an authorship?

I have actually seen a PhD student being seriously pissed at her former supervisor for publishing something with her name on it although she had not participated in writing. She was aware of the problem of having your name on something that you weren't able to assess for its correctness.

caynazzo said...

Alex SL:
"Obviously everybody who has made a significant contribution should be asked to help write the paper. But if they then don't, how can you possibly give them an authorship?"

Since most journals are English language you can see the problem with statements like that.

Others have address why intellectual contribution doesn't mean what you think it means because it's not limited to banging on a keyboard.

Alex SL:
"Apart from making a mockery of the term 'author', have you thought through what that means for responsibility?"

Have you considered the possibility that your idea of authorship is outdated, restrictive and too literal?

Alex SL:
"If you are the co-author of a paper you have basically signed the study off with your name. If the scientist who coordinated it all fudged the data to make them more interesting somebody who only contributes a minor technical step will have no way nor even the qualifications to realize that. But they might still have their career damaged when the fraud is discovered because, again, they signed it off."

Such is modern science. (Yours will be a much better argument when it's backed by evidence.) There are risks in joining a consortium or any large collaboration that imply a certain level of trust in your fellow collaborators to not be unethical. No single author can be intimate with all the techniques employed in these types of papers. Are you really expecting a bioinformation grad student who analyzed the RNA-seq data to be able to adjudicate the validity of what FISH probes a tech used? Both are middle authors after all. Maybe you think students and trainees shouldn't ever publish papers since they don't have a complete mastery of their discipline and related disciplines.

Alex SL said...

Have you considered the possibility that your idea of authorship is outdated, restrictive and too literal? ... Such is modern science.

Comments like these are why I think that your and Georgi's position are based more on what is established practice in perhaps two fields of science than on what would make sense from the perspective of appropriateness, fairness, intellectual consistency and personal accountability.

Georgi Marinov said...

than on what would make sense from the perspective of appropriateness, fairness, intellectual consistency and personal accountability.

I can't believe my eyes. "Fairness" and "appropriateness"??

Do you even realize that you are arguing for a system in which between 50 and 500 people do the bulk of the work, including the intellectual heavy-lifting, and then only between 5 and 50 of them end up being authors. With those 5 to 50 being the PIs and everyone else who is left out being the grad students, postdoc and technicians. In what world is this fair? There are the following broad categories of people on these authors lists:

1) PIs - wrote the grants, played role in steering the research in a certain direction, oversaw things. Some have taken part in developing computational methods and designing and troubleshooting experiments. Wrote the paper in the end.

Trainees and research scientists who actually carried out the research, of which there are the following groups with respect to the final manuscript:

2) Made figures.This usually means they also had crucial role in the analysis of the data, including developing experimental and computational methods and tools. Sometimes some people in that category also write some text.

3) Also had crucial role in developing experimental and computational methods and tools, and producing data, however, ended up not contributing directly to the final manuscript in the form of text and figures

4) Technicians who usually do the bulk of data generation, where by "technician" I refer to the role they play, not to their official position (they may be grad students or postdocs too)

Then there is the last group

5) People who indeed contributed nothing but ended up on the paper because they went to meetings, were on the mailing lists, were included for political reasons, or nobody realized they had done nothing when compiling the authors list

That last group certainly does exist, and we can agree on whether it should or should not be on the paper. But you are claiming that groups 2, 3 and 4 should also be left out (except maybe for those in group 2 who did write some text). Which is insanity. Again, keep in mind who you are punishing in this way - the trainees who did most of the work. How is that "fair and appropriate"?

And who in their right mind would want to work on such a project under your conditions? Good luck finding qualified people willing to work just for money (remember that these are NIH-funded projects, which means NIH trainee-level salaries), which means good luck trying to actually do the project.