Hi Erik, the information you provide about NIH indirects is incorrect. The indirects are not taken out of the funds that pay for the research, they are separate and pay for the infrastructure that supports the research. For example, if I propose a study that costs 1 million for the research (direct costs), the university receives an additional 500,000 to pay for the infrastructural and administrative costs (indirect costs) that support the research. Capping indirect costs at 15% will require universities to immediately fire staff since the change affects funds that have already been budgeted for the remainder of this fiscal year. The impact is not just on Ivy League schools, it’s on all universities (state and private) that support NIH research.
Is this just changing the emphasis or is there something factual I'm missing? It would certainly seem to make funding confusing for all involved, e.g., if I apply for a million dollar grant, and then everyone has to keep track of the fact that somehow this is really for a 1.5 million dollar grant. I think it's more sensible to think about it as you apply and are awarded 1.5 million dollars, and then 33% percent goes to the university, as that would be the listed award, e.g., "The Young Investigators Award for 1.5 million." To take an example: for me it was usually 56% in indirect costs (I just went and checked). So for me, if I got $100,000 grant, my only direct cost is salary, since I'm a theorist. So a $100,000 grant should have, judged by the total amount, paid for my salary for more than a year. But it wouldn't have done that. It would have instead put $43,000 toward my salary, like 6 months or so, and then the university pockets 56%. You could justify that percentage, e.g., by saying that 56% is required to keep the university running and the person at payroll needs 56% of the money I bring in in order to pay me salary, but I think it's a semantic argument to say this money isn't really coming from the grant, since it is, it's just either a justifiable high percentage or not.
No, the FOA caps for a grant are always about direct costs otherwise every Prof would rather work at University of Arkansas where the indirects are 20% vs Children’s National where they are closer to 90%! You would pick your University based on how much they were taking from your research budget, wouldn’t you?! Instead the indirects get added on top of whatever you requested so it isn’t money you had to justify any way, it was negotiated directly between the university and HHS and applies to all federal funding agencies. Also they cannot consider the indirects when reviewing your grant even if it means millions more that need to be paid out because of where you work. Probably your research office just took care of this for you and you were none the wiser;)
Here is a call for a grant up to $60,000 that I applied for and received. And I think I found the grant breakdown, where I end up getting only $27,000 to pay salary (although I think fringe might go to health care, I don't remember). The indirect costs came out of the advertised total. That is, If you apply to a grant with a total cost X, indirect cost comes out of the X. It's not some unlisted behind-the-scenes negotiation (and I can't imagine how complicated that would make things). https://imgur.com/a/0q1fLDL
Ahh, ok so the confusion is understandable and thanks for the transparency and example from your own research program. But there are actually two distinct models for how federal agencies handle indirect costs (F&A) in grants. Most federal research funding agencies, including NIH, NSF, NASA, DOE, etc., add indirect costs on top of the direct costs using institutionally negotiated rates. However, the Department of Defense (including ARO, ONR, DARPA) often operates differently, where total award amounts include both direct and indirect costs. And you will also note that the F&A that they pulled from your grant here is lower than the Tuft's rate that you cited previously (36% VS 56%)? This is because DOD already capped the rates that Universities could pull from your budget and evened out the playing field a bit for researchers.
Regarding funding distributions: DOD research funding to universities represents approximately 5-7% of total federal research funding to academic institutions, while agencies following the NIH model (NIH, NSF, DOE, NASA, etc.) collectively account for about 90-93%. This is why most researchers are more familiar with the NIH model where indirect costs are added on top.
So someone used to the NIH model can be quite surprised if they don't read the details in the FOA for something from DOD, or if you apply to a private funder and forget to add it to your budget (as I know of several new faculty who have made this mistake).
Finally, I think this points to a larger issue that new faculty never get adequate training about these grant budget quirks. Most of us learn about different agency rules the hard way.
This does explain it, thank you (and sorry, I've been visiting family this weekend so am slower on responses). I'm simply used to receiving either funding from nonprofits or from the governmental funding sources that do take indirect costs out of the total received. Well, that changes my opinion!
It's funny because the two fields that I work in, which is research on consciousness, as well as work on emergence, are not quite, well, conservative enough for most of the institutions following the NIH model. They're getting better about that sort of thing, but this explains why I'm a pretty peculiar case of being solely on one side of this split.
Erik, something factual that is worth rethinking is your claim that indirect costs can go towards "the new fancy gym for undergraduates". My understanding is that indirect costs can only be used to reimburse F&A (facilities and administration) costs that are related to the NIH sponsored research. See e.g. point 10 in this FAQ from the AAMC: https://www.aamc.org/media/81716/download?attachment. Do you dispute this or think it is misleading? Of course, you could argue that admin costs are too high, but that would require a more nuanced and technical argument. As it is, your characterization of the situation seems incorrect to me.
The minor issue is that $100k at 56% means you got about $64k directly (indirects are 56% of that on top) but I think the larger issue is that the costs legally can’t go to the gym, but to research infrastructure. A theorist probably uses less of that and gets screwed, but like I use an HPC system paid on indirects.
So can I start charging my HPC costs as directs? Yea probably. There’s a world where this is good if the staffing is sorted out and more grants get funded now and a world where staffing collpases and universities stop investing in research.
Yea came here to make the same comment. Some top school do have crazy high indirect rates, but the indirects are highly restricted to support research costs. My school has a computing center for example. Legally the money cannot be spent on heating the building, paying nonresearch staff, or say a new gym. A lot of research institutes will cease to exist if this happens, ie Woods Hole’s rate is 65% or so but they don’t collect tuition.
Second that the percentage rate is of direct costs, not of total. A 50% indirect means a $100k grant is bumped up to $150k, not that $75k of a $150k grant goes to the school. So the 70% rates are high but its most a majority of the grant unless indirects hit 100% (some national labs do that through different arragments, like nation security projects funded through select grants plus indirects, but not universities).
Bryan is correct. Universities negotiate the indirects with NIH and the budget you make for your research is what you get in directs. They slap on the indirects and pay that directly to the admin and it doesn’t come out of your research budget. And budget caps are always on directs not indirects as the % varies widely from 10-nearly 90% at some institutions. Nevertheless cutting down on indirects will free up more money for researchers !
PS I was wrong about the Trump administration allowing the money that previously was allocated for indirects going back into the NIH budget so actually this will just defund Universities and in no way benefit researchers. The indirects which never came out of researchers’ budgets will now just be cut from the NIH pot of money. Perhaps consider revising this part of your post since no researcher is benefiting from this policy. You could say it will dramatically reduce University administration if that is something you are happy with but it’s a huge blow to the current budgets of Universities.
The possibility that the AI wrote research papers at a PhD level is only possible is Cowen has terrible PhD students. My prior on that is insanely low.
I wonder if the "human search engine"/many eyeballs might work better than AI at finding you a bed. Assuming you would prefer the bed to a really good AI test question. :) Tell your readers what you are looking for and see if one of us can find something that matches.
Really disappointing that you frame indirect costs as being used mostly for frivolous things, like building new gyms for undergraduates. Do you know this to be factually correct? Most of the indirects are used to pay for the research buildings, utilities, and admin staff.
Very interesting to see #6... Yours is one of the few theories of emergence that has enough mathematical detail to support ongoing work. Many other discussions of emergence make vague assertions about ontology vs. epistemology without laying it down in formal details that can actually be applied. Whereas Zhang et al. are generalizing from your EI measure, my own work has focused on generalizing from your coarse-graining rules (and the black-boxing rules in later IIT papers) to show how a weaker set of rules for forming macro elements still produces interesting properties (i.e. still stronger than functional equivalence). I'm eager to see what new scientific ideas you're coming up with!
I found this working book claimed to introduce a novel perspective on value through the lens of granular interaction thinking theory, proposing an informational entropy-based notion of value. Could be interesting!
I apprecite the conversation about AI and the suggestion of a great illustrator. Illustrators should be less worried about midjourney and simply show depictions of good work. People who otherwise would never use illustrators are not going to use it--perhaps because of midjourney, perhaps because they have a substack etc--and there needs to be a better way to find them. I spent forever trying to find a kids illustrator...
My new book, 'The Future Loves You: Why and How We Should Abolish Death', is out now!
You've probably heard of the science-fiction concept of 'cryosleep', where people stay unconscious, unchanging, and unageing in pods for indefinite periods of time, only to later re-emerge into a strange new world. You may even have heard of the already-existing practice of cryonics, where the clinically dead are pumped full of antifreeze and suspended upside down in liquid nitrogen in the hope that they might be resurrected by the medicine of a more advanced age. What you may be unsure of, though, is whether this is all science-fiction nonsense, or whether in principle technology like this could somehow be made to work?
Well, my book runs through the very real science, medicine, philosophy and more required to analyse this question, and concludes that preservation - if well done - really could be used to stop someone from dying. After all, from ventilators to brain implants, modern medicine has already been blurring what it means to die. The past century of scientific and medical advances have shown that death is no longer the loss of heartbeat or breath, but of personal identity – that the core of our identities is our minds, and that our minds are encoded in the structure of our brains. On this basis, I explore how recently invented brain preservation techniques now offer us all the chance of preserving our minds to enable our future revival.
If you enjoy Erik's discussions of consciousness, neuroscience, philosophy and more, there's a decent chance you'd find my book interesting too. Erik's theory of falsification and consciousness even briefly features in one of my middle chapters, 'What is Consciousness?'
I’ve found Deep Research to be great at answering questions that are straightforward, but require a lot of grinding, such as looking up certain facts about each item in a list and presenting the information in a table for easy comparison. I’m shopping for a vehicle right now and it compiled a list of everything that they changed in each model year and what trim I needed in each year to get the features I was looking for. It’s helped me make decisions about medical treatment tradeoffs by gathering the quantitative information I wanted to look up, saving me hours of compiling information.
Compared with Wikipedia or general articles on the web, it is far better at answering a specific quantitative question.
It blows away other models for compiling info. They can often retrieve one fact, but never a spreadsheet’s worth of facts.
I have found the initial answer that it gives asynchronously to be dead accurate so far, while follow-up questions unfortunately do not seem to invoke the same model and are often inaccurate.
Thanks for testing Deep Research so I don't have to. I was thinking about having it look into homeowner's insurance, but if it can't even find a mattress, forget it.
I was pretty surprised in both a good and bad way at trying to use all recent LLMs to answer a pretty simple question as a test (explain the concept of the vocateur in Ada Palmer's Terra Ignota series) and how they resorted to a wild amount of bullshitting based on what the word seems to be and sound like. On one hand, come on; she's talked about this, even further prompting like "Jo Walton considers herself a vocateur reader, what would this mean then" resulted in more games. On the other, not bad to try to tie the whole thing into someone who talks, advocates; good guesses if you don't know what you're talking about, and maybe 80% of the way to figuring it out from first principles. But that last 20% is kind of the point of the question.
I believe you’ve written about microplastics before. Did you see the PlasticList.org research?
I’m working on a concept to let consumers (buy or crowdfund) identical testing on whatever foods or drinks they want, with the results being published openly.
What’s your take on civilian-funded research/testing like this, or on microplastics (really plasticizers or phthalates) in the CPG supply chain generally. Tractable problem? Too big? Too little utility?
I associate with many people who are involved with animal rescue. They mostly concern themselves with homeless or abused dogs and cats, sometimes with guinea pigs, rabbits, gerbils, etc. Not many of them worry about the plight of farm animals (though I think they should). And almost no one is concerned with suffering among wild animals. Bentham's Bulldog (see link above) puts it into perspective for us.
Hi Erik, the information you provide about NIH indirects is incorrect. The indirects are not taken out of the funds that pay for the research, they are separate and pay for the infrastructure that supports the research. For example, if I propose a study that costs 1 million for the research (direct costs), the university receives an additional 500,000 to pay for the infrastructural and administrative costs (indirect costs) that support the research. Capping indirect costs at 15% will require universities to immediately fire staff since the change affects funds that have already been budgeted for the remainder of this fiscal year. The impact is not just on Ivy League schools, it’s on all universities (state and private) that support NIH research.
Is this just changing the emphasis or is there something factual I'm missing? It would certainly seem to make funding confusing for all involved, e.g., if I apply for a million dollar grant, and then everyone has to keep track of the fact that somehow this is really for a 1.5 million dollar grant. I think it's more sensible to think about it as you apply and are awarded 1.5 million dollars, and then 33% percent goes to the university, as that would be the listed award, e.g., "The Young Investigators Award for 1.5 million." To take an example: for me it was usually 56% in indirect costs (I just went and checked). So for me, if I got $100,000 grant, my only direct cost is salary, since I'm a theorist. So a $100,000 grant should have, judged by the total amount, paid for my salary for more than a year. But it wouldn't have done that. It would have instead put $43,000 toward my salary, like 6 months or so, and then the university pockets 56%. You could justify that percentage, e.g., by saying that 56% is required to keep the university running and the person at payroll needs 56% of the money I bring in in order to pay me salary, but I think it's a semantic argument to say this money isn't really coming from the grant, since it is, it's just either a justifiable high percentage or not.
No, the FOA caps for a grant are always about direct costs otherwise every Prof would rather work at University of Arkansas where the indirects are 20% vs Children’s National where they are closer to 90%! You would pick your University based on how much they were taking from your research budget, wouldn’t you?! Instead the indirects get added on top of whatever you requested so it isn’t money you had to justify any way, it was negotiated directly between the university and HHS and applies to all federal funding agencies. Also they cannot consider the indirects when reviewing your grant even if it means millions more that need to be paid out because of where you work. Probably your research office just took care of this for you and you were none the wiser;)
Here is a call for a grant up to $60,000 that I applied for and received. And I think I found the grant breakdown, where I end up getting only $27,000 to pay salary (although I think fringe might go to health care, I don't remember). The indirect costs came out of the advertised total. That is, If you apply to a grant with a total cost X, indirect cost comes out of the X. It's not some unlisted behind-the-scenes negotiation (and I can't imagine how complicated that would make things). https://imgur.com/a/0q1fLDL
Ahh, ok so the confusion is understandable and thanks for the transparency and example from your own research program. But there are actually two distinct models for how federal agencies handle indirect costs (F&A) in grants. Most federal research funding agencies, including NIH, NSF, NASA, DOE, etc., add indirect costs on top of the direct costs using institutionally negotiated rates. However, the Department of Defense (including ARO, ONR, DARPA) often operates differently, where total award amounts include both direct and indirect costs. And you will also note that the F&A that they pulled from your grant here is lower than the Tuft's rate that you cited previously (36% VS 56%)? This is because DOD already capped the rates that Universities could pull from your budget and evened out the playing field a bit for researchers.
Regarding funding distributions: DOD research funding to universities represents approximately 5-7% of total federal research funding to academic institutions, while agencies following the NIH model (NIH, NSF, DOE, NASA, etc.) collectively account for about 90-93%. This is why most researchers are more familiar with the NIH model where indirect costs are added on top.
So someone used to the NIH model can be quite surprised if they don't read the details in the FOA for something from DOD, or if you apply to a private funder and forget to add it to your budget (as I know of several new faculty who have made this mistake).
Finally, I think this points to a larger issue that new faculty never get adequate training about these grant budget quirks. Most of us learn about different agency rules the hard way.
This does explain it, thank you (and sorry, I've been visiting family this weekend so am slower on responses). I'm simply used to receiving either funding from nonprofits or from the governmental funding sources that do take indirect costs out of the total received. Well, that changes my opinion!
It's funny because the two fields that I work in, which is research on consciousness, as well as work on emergence, are not quite, well, conservative enough for most of the institutions following the NIH model. They're getting better about that sort of thing, but this explains why I'm a pretty peculiar case of being solely on one side of this split.
Erik, something factual that is worth rethinking is your claim that indirect costs can go towards "the new fancy gym for undergraduates". My understanding is that indirect costs can only be used to reimburse F&A (facilities and administration) costs that are related to the NIH sponsored research. See e.g. point 10 in this FAQ from the AAMC: https://www.aamc.org/media/81716/download?attachment. Do you dispute this or think it is misleading? Of course, you could argue that admin costs are too high, but that would require a more nuanced and technical argument. As it is, your characterization of the situation seems incorrect to me.
The minor issue is that $100k at 56% means you got about $64k directly (indirects are 56% of that on top) but I think the larger issue is that the costs legally can’t go to the gym, but to research infrastructure. A theorist probably uses less of that and gets screwed, but like I use an HPC system paid on indirects.
So can I start charging my HPC costs as directs? Yea probably. There’s a world where this is good if the staffing is sorted out and more grants get funded now and a world where staffing collpases and universities stop investing in research.
Yea came here to make the same comment. Some top school do have crazy high indirect rates, but the indirects are highly restricted to support research costs. My school has a computing center for example. Legally the money cannot be spent on heating the building, paying nonresearch staff, or say a new gym. A lot of research institutes will cease to exist if this happens, ie Woods Hole’s rate is 65% or so but they don’t collect tuition.
Second that the percentage rate is of direct costs, not of total. A 50% indirect means a $100k grant is bumped up to $150k, not that $75k of a $150k grant goes to the school. So the 70% rates are high but its most a majority of the grant unless indirects hit 100% (some national labs do that through different arragments, like nation security projects funded through select grants plus indirects, but not universities).
Bryan is correct. Universities negotiate the indirects with NIH and the budget you make for your research is what you get in directs. They slap on the indirects and pay that directly to the admin and it doesn’t come out of your research budget. And budget caps are always on directs not indirects as the % varies widely from 10-nearly 90% at some institutions. Nevertheless cutting down on indirects will free up more money for researchers !
PS I was wrong about the Trump administration allowing the money that previously was allocated for indirects going back into the NIH budget so actually this will just defund Universities and in no way benefit researchers. The indirects which never came out of researchers’ budgets will now just be cut from the NIH pot of money. Perhaps consider revising this part of your post since no researcher is benefiting from this policy. You could say it will dramatically reduce University administration if that is something you are happy with but it’s a huge blow to the current budgets of Universities.
The possibility that the AI wrote research papers at a PhD level is only possible is Cowen has terrible PhD students. My prior on that is insanely low.
Think your second "is" should have been "if".
I wonder if the "human search engine"/many eyeballs might work better than AI at finding you a bed. Assuming you would prefer the bed to a really good AI test question. :) Tell your readers what you are looking for and see if one of us can find something that matches.
Really disappointing that you frame indirect costs as being used mostly for frivolous things, like building new gyms for undergraduates. Do you know this to be factually correct? Most of the indirects are used to pay for the research buildings, utilities, and admin staff.
Very interesting to see #6... Yours is one of the few theories of emergence that has enough mathematical detail to support ongoing work. Many other discussions of emergence make vague assertions about ontology vs. epistemology without laying it down in formal details that can actually be applied. Whereas Zhang et al. are generalizing from your EI measure, my own work has focused on generalizing from your coarse-graining rules (and the black-boxing rules in later IIT papers) to show how a weaker set of rules for forming macro elements still produces interesting properties (i.e. still stronger than functional equivalence). I'm eager to see what new scientific ideas you're coming up with!
Share a link to his work of yours?
Thanks for asking... I hesitated to share a link in a comment thread.
My thesis is available here: https://doi.org/10.46569/hq37vw99k
Here's a talk about it, though only covering through chapter 4: https://www.youtube.com/watch?v=7y5w-CBK928
I'm working with a collaborator on turning chapter 5 into a paper for publication, and I'll write a post about it here on Substack at some point!
Great post Erik!
1. Was actually talking about asteroids this morning, 6. Kudos on your Theory of Emergence - ELI5, 7. loved the new art! 9. Fun post!
I found this working book claimed to introduce a novel perspective on value through the lens of granular interaction thinking theory, proposing an informational entropy-based notion of value. Could be interesting!
https://books.google.com.vn/books?hl=en&lr=&id=8SVEEQAAQBAJ&oi=fnd&pg=PA8&ots=JRez8XfakU&sig=fpzZJ-9CrNbeAOEKP4CAgfdLF5Y&redir_esc=y#v=onepage&q&f=false
I apprecite the conversation about AI and the suggestion of a great illustrator. Illustrators should be less worried about midjourney and simply show depictions of good work. People who otherwise would never use illustrators are not going to use it--perhaps because of midjourney, perhaps because they have a substack etc--and there needs to be a better way to find them. I spent forever trying to find a kids illustrator...
2.3% is hardly different from the chances of double 6 with 2 dice (2.7777%). Uncomfortable.
My new book, 'The Future Loves You: Why and How We Should Abolish Death', is out now!
You've probably heard of the science-fiction concept of 'cryosleep', where people stay unconscious, unchanging, and unageing in pods for indefinite periods of time, only to later re-emerge into a strange new world. You may even have heard of the already-existing practice of cryonics, where the clinically dead are pumped full of antifreeze and suspended upside down in liquid nitrogen in the hope that they might be resurrected by the medicine of a more advanced age. What you may be unsure of, though, is whether this is all science-fiction nonsense, or whether in principle technology like this could somehow be made to work?
Well, my book runs through the very real science, medicine, philosophy and more required to analyse this question, and concludes that preservation - if well done - really could be used to stop someone from dying. After all, from ventilators to brain implants, modern medicine has already been blurring what it means to die. The past century of scientific and medical advances have shown that death is no longer the loss of heartbeat or breath, but of personal identity – that the core of our identities is our minds, and that our minds are encoded in the structure of our brains. On this basis, I explore how recently invented brain preservation techniques now offer us all the chance of preserving our minds to enable our future revival.
If you enjoy Erik's discussions of consciousness, neuroscience, philosophy and more, there's a decent chance you'd find my book interesting too. Erik's theory of falsification and consciousness even briefly features in one of my middle chapters, 'What is Consciousness?'
https://www.arielzj.com/the-future-loves-you
I’ve found Deep Research to be great at answering questions that are straightforward, but require a lot of grinding, such as looking up certain facts about each item in a list and presenting the information in a table for easy comparison. I’m shopping for a vehicle right now and it compiled a list of everything that they changed in each model year and what trim I needed in each year to get the features I was looking for. It’s helped me make decisions about medical treatment tradeoffs by gathering the quantitative information I wanted to look up, saving me hours of compiling information.
Compared with Wikipedia or general articles on the web, it is far better at answering a specific quantitative question.
It blows away other models for compiling info. They can often retrieve one fact, but never a spreadsheet’s worth of facts.
I have found the initial answer that it gives asynchronously to be dead accurate so far, while follow-up questions unfortunately do not seem to invoke the same model and are often inaccurate.
Thanks for testing Deep Research so I don't have to. I was thinking about having it look into homeowner's insurance, but if it can't even find a mattress, forget it.
https://x.com/ATabarrok/status/1888061547261386940
Alex Tabarrok is as small-government-libertarian as it gets and he thinks this NIH policy is idiotic.
I was pretty surprised in both a good and bad way at trying to use all recent LLMs to answer a pretty simple question as a test (explain the concept of the vocateur in Ada Palmer's Terra Ignota series) and how they resorted to a wild amount of bullshitting based on what the word seems to be and sound like. On one hand, come on; she's talked about this, even further prompting like "Jo Walton considers herself a vocateur reader, what would this mean then" resulted in more games. On the other, not bad to try to tie the whole thing into someone who talks, advocates; good guesses if you don't know what you're talking about, and maybe 80% of the way to figuring it out from first principles. But that last 20% is kind of the point of the question.
Really coming to love these posts!
I believe you’ve written about microplastics before. Did you see the PlasticList.org research?
I’m working on a concept to let consumers (buy or crowdfund) identical testing on whatever foods or drinks they want, with the results being published openly.
What’s your take on civilian-funded research/testing like this, or on microplastics (really plasticizers or phthalates) in the CPG supply chain generally. Tractable problem? Too big? Too little utility?
This is what I've found to be interesting (if disturbing).
https://benthams.substack.com/p/the-worst-thing-in-the-world-isnt
I associate with many people who are involved with animal rescue. They mostly concern themselves with homeless or abused dogs and cats, sometimes with guinea pigs, rabbits, gerbils, etc. Not many of them worry about the plight of farm animals (though I think they should). And almost no one is concerned with suffering among wild animals. Bentham's Bulldog (see link above) puts it into perspective for us.