I agree with virtually everything in this and the previous post on indirect costs. However, part of intellectual marketing is recruiting members of the research community to assemble a research road map of topics that we believe are fundamentally important. Then selling this to the funding agencies and to our colleagues who will be reviewing our proposals. In this case, university researchers can set the agenda and direct research in the right directions.
That said, I think we should explore alternative funding models. For example, could we give every reasonably good faculty member a small annual grant (exempt from F&A charges) to spend as they see fit without requiring a proposal? Enough to fund one student and one month of summer salary. Call these "Innovation Grants". Under the current system, every reasonably good faculty member eventually gets such a grant (often as part of a larger team), but at great expense in time spent grant writing, selling, and reviewing. Faculty could re-qualify every 5 years based on measures of creativity rather than "impact" (i.e., publications). Explicitly reward people pursuing novel directions. Removing the F&A charges would remove the incentive for universities to add faculty just to get the F&A money (which was a big problem when NIH expanded its grant programs).
We could incentivize larger collaborations by telling faculty that if they combined their individual grants into a consortium to address some bigger question, the consortium could be eligible for additional funding for infrastructure, technicians/software engineers, etc. This would be a purely bottom-up system. (There may be some bugs to be worked out in this idea.)
I like it. This certainly aligns with Graeber's proposal of "Find some bright people, give them the resources they need to pursue whatever idea comes into their heads, and then leave them alone for a while."
The product of science is the result of a communal effort. So the incentives that affect individual scientists don't necessarily affect in the same direction the product over time as a whole. In particular, even though it doesn't pay to go against the tide, some scientists do it, and a few produce major new findings (and this is also sometimes rewarded with major prizes). Their impact on science may be larger than the many others who do follow the fashion of the day.
Part of the issue is that there are a lot of scientists, probably more than the optimal number from the point of view of optimizing the allocation of resources to advance science. This is due to the proliferation of higher education and the connection between teaching and research. A lot of science has to be exploration of relatively low importance, but perhaps there's too much of that, beyond the level that is necessary to facilitate great discoveries.
I have a piece coming out in a few weeks that argues lots of incremental results might be the natural end state of science, anyway. But I worry about the associated human cost, and the publish-or-perish pressure is bad for our mental health.
Well, I don't think that the scientific method was ever promoted as a wellness program, except perhaps when attempting to influence budget allocations.
1. ML research clearly has a very serious problem with treating research as marketing and idea promotion.
2. I agree that perverse incentives are almost certainly the underlying cause.
3. I would argue that there are other important incentives, in addition to funding, that influence the structural bias towards positive results.
To flesh these points out a bit more:
Prior work has used "percentage of papers reporting positive results" as a proxy for the amount of bias within a research field (see “Positive” Results Increase Down the Hierarchy of the Science, Daniele Fanelli, 2010). Fanelli looks at 2000+ papers in 20+ fields, and finds that the percentage of papers increases for the "softer" sciences; psychology has the highest percentage of positive results with about 90%. I then took a random sample of 400 papers in my subfield (using ML to solve PDEs) and found that of the 400 papers in the random sample whose abstracts mention positive and/or negative experimental results, 95% (220/232) mention only positive results, 5% (12/232) mention both positive and negative results, and 0% (0/232) mention only negative results.
The percentage of positive results in (my subfield of) ML appears to be much higher than in any other field of science! This of course doesn't prove anything (not the same experimental design, not a perfect proxy, etc), but it does suggest that ML has a really serious issue with researchers interpreting and presenting their results in biased ways so that they can market a paper with a "positive" result.
I wrote about incentives in a paper (under review):
"We emphasize both good intentions and perverse incentives as explanations for the apparent bias towards positive results. The culture of scientific ML is one in which well-intentioned researchers try to figure out ways that ML might be useful for science. In the process of doing so, they tend to be less interested in reporting ways that ML isn’t useful. Perverse incentives also contribute. Because ML research rewards novel ideas and positive experimental results, all else being equal articles with weak baselines and/or reporting biases are more likely to get accepted to prestigious venues and more likely to be widely cited. Incentives against negative results are particularly strong in scientific ML, because career advancement (in academia) and lucrative jobs (in industry) depend on the presumption that ML will be a useful tool for scientific problems. Negative results could cast doubt on that presumption, thereby undermining justification for one’s research area."
Interesting. I'd only caution that I don't know what the right number of positive results should be! I always worry about metascientific metrics and their construct validity. This comes up in Meehl's later lectures, so I'll return to it soon.
I mean federal funding for research has stayed flat (or shall I even say sometimes decreased?) in the past 20 years while inflation has only been going up. Congress has recently cut NASA and the NSF budget by like 2-6% and NIH is expecting another ~4% budget cut, while inflation this past year was like 9%. Unfortunately, the public in the US and the govt has been convinced that research funding is inefficient and a waste of money (while the US spends ~23% and ~15% on a horribly inefficient, ineffective, and completely overbloated healthcare/defense system).
Now, this leads to a problem. With decreased federal funding, there comes the issue of academics being less likely to take risks as they need to yield results to fight for the increasingly scarce funding. With decreased federal fundings, there comes the issue that academics need to turn to big companies for funding that has no other reason to be in research other than to fully commercialize science and turn it into a paper burgerfest about the newest, shiniest, newest algorithm/interface (e.g., the WHALES like Google, Meta, and Amazon). Unlike the 60s and 70s, being able to just do pure fundamental science is just not possible anymore because the infrastructure we’ve built for science in the US makes what you talk about disliking in the blog the only feasible route for most scientists. When you compare the US to countries like Norway and China and then hear politicians incessantly talk about how there are rivals in other parts of the world who are beating us in alternative energy and modern technology, just note they’re the ones keeping research funding flat (or even cutting it) while doing absolutely nothing to address the issue. In the overall scheme of things, $11 billion, $47 billion, and $25 billion for NSF, NIH, and NASA is not that much money to the federal govt, but congress keeps looking to slash funding in these agencies while propping up the ridiculously bloated, ineffective, and overpriced healthcare industry complex and military industry complex.
This was painful to read Ben, but so true. The only useful thing I can add to this post is that the postumous book by David Graeber (along with David Wengrow) "The Dawn of Everything" was a delightful inquiry into the long storied propensity of humans to construct institutions that become ends to justify means, rather than means to pursue ends.
I agree with virtually everything in this and the previous post on indirect costs. However, part of intellectual marketing is recruiting members of the research community to assemble a research road map of topics that we believe are fundamentally important. Then selling this to the funding agencies and to our colleagues who will be reviewing our proposals. In this case, university researchers can set the agenda and direct research in the right directions.
That said, I think we should explore alternative funding models. For example, could we give every reasonably good faculty member a small annual grant (exempt from F&A charges) to spend as they see fit without requiring a proposal? Enough to fund one student and one month of summer salary. Call these "Innovation Grants". Under the current system, every reasonably good faculty member eventually gets such a grant (often as part of a larger team), but at great expense in time spent grant writing, selling, and reviewing. Faculty could re-qualify every 5 years based on measures of creativity rather than "impact" (i.e., publications). Explicitly reward people pursuing novel directions. Removing the F&A charges would remove the incentive for universities to add faculty just to get the F&A money (which was a big problem when NIH expanded its grant programs).
We could incentivize larger collaborations by telling faculty that if they combined their individual grants into a consortium to address some bigger question, the consortium could be eligible for additional funding for infrastructure, technicians/software engineers, etc. This would be a purely bottom-up system. (There may be some bugs to be worked out in this idea.)
I like it. This certainly aligns with Graeber's proposal of "Find some bright people, give them the resources they need to pursue whatever idea comes into their heads, and then leave them alone for a while."
The product of science is the result of a communal effort. So the incentives that affect individual scientists don't necessarily affect in the same direction the product over time as a whole. In particular, even though it doesn't pay to go against the tide, some scientists do it, and a few produce major new findings (and this is also sometimes rewarded with major prizes). Their impact on science may be larger than the many others who do follow the fashion of the day.
Part of the issue is that there are a lot of scientists, probably more than the optimal number from the point of view of optimizing the allocation of resources to advance science. This is due to the proliferation of higher education and the connection between teaching and research. A lot of science has to be exploration of relatively low importance, but perhaps there's too much of that, beyond the level that is necessary to facilitate great discoveries.
I have a piece coming out in a few weeks that argues lots of incremental results might be the natural end state of science, anyway. But I worry about the associated human cost, and the publish-or-perish pressure is bad for our mental health.
Well, I don't think that the scientific method was ever promoted as a wellness program, except perhaps when attempting to influence budget allocations.
Three points:
1. ML research clearly has a very serious problem with treating research as marketing and idea promotion.
2. I agree that perverse incentives are almost certainly the underlying cause.
3. I would argue that there are other important incentives, in addition to funding, that influence the structural bias towards positive results.
To flesh these points out a bit more:
Prior work has used "percentage of papers reporting positive results" as a proxy for the amount of bias within a research field (see “Positive” Results Increase Down the Hierarchy of the Science, Daniele Fanelli, 2010). Fanelli looks at 2000+ papers in 20+ fields, and finds that the percentage of papers increases for the "softer" sciences; psychology has the highest percentage of positive results with about 90%. I then took a random sample of 400 papers in my subfield (using ML to solve PDEs) and found that of the 400 papers in the random sample whose abstracts mention positive and/or negative experimental results, 95% (220/232) mention only positive results, 5% (12/232) mention both positive and negative results, and 0% (0/232) mention only negative results.
The percentage of positive results in (my subfield of) ML appears to be much higher than in any other field of science! This of course doesn't prove anything (not the same experimental design, not a perfect proxy, etc), but it does suggest that ML has a really serious issue with researchers interpreting and presenting their results in biased ways so that they can market a paper with a "positive" result.
I wrote about incentives in a paper (under review):
"We emphasize both good intentions and perverse incentives as explanations for the apparent bias towards positive results. The culture of scientific ML is one in which well-intentioned researchers try to figure out ways that ML might be useful for science. In the process of doing so, they tend to be less interested in reporting ways that ML isn’t useful. Perverse incentives also contribute. Because ML research rewards novel ideas and positive experimental results, all else being equal articles with weak baselines and/or reporting biases are more likely to get accepted to prestigious venues and more likely to be widely cited. Incentives against negative results are particularly strong in scientific ML, because career advancement (in academia) and lucrative jobs (in industry) depend on the presumption that ML will be a useful tool for scientific problems. Negative results could cast doubt on that presumption, thereby undermining justification for one’s research area."
Interesting. I'd only caution that I don't know what the right number of positive results should be! I always worry about metascientific metrics and their construct validity. This comes up in Meehl's later lectures, so I'll return to it soon.
I mean federal funding for research has stayed flat (or shall I even say sometimes decreased?) in the past 20 years while inflation has only been going up. Congress has recently cut NASA and the NSF budget by like 2-6% and NIH is expecting another ~4% budget cut, while inflation this past year was like 9%. Unfortunately, the public in the US and the govt has been convinced that research funding is inefficient and a waste of money (while the US spends ~23% and ~15% on a horribly inefficient, ineffective, and completely overbloated healthcare/defense system).
Now, this leads to a problem. With decreased federal funding, there comes the issue of academics being less likely to take risks as they need to yield results to fight for the increasingly scarce funding. With decreased federal fundings, there comes the issue that academics need to turn to big companies for funding that has no other reason to be in research other than to fully commercialize science and turn it into a paper burgerfest about the newest, shiniest, newest algorithm/interface (e.g., the WHALES like Google, Meta, and Amazon). Unlike the 60s and 70s, being able to just do pure fundamental science is just not possible anymore because the infrastructure we’ve built for science in the US makes what you talk about disliking in the blog the only feasible route for most scientists. When you compare the US to countries like Norway and China and then hear politicians incessantly talk about how there are rivals in other parts of the world who are beating us in alternative energy and modern technology, just note they’re the ones keeping research funding flat (or even cutting it) while doing absolutely nothing to address the issue. In the overall scheme of things, $11 billion, $47 billion, and $25 billion for NSF, NIH, and NASA is not that much money to the federal govt, but congress keeps looking to slash funding in these agencies while propping up the ridiculously bloated, ineffective, and overpriced healthcare industry complex and military industry complex.
RE: "not towing party lines”. Loud Buzzer Sounding ….. “not TOEING party lines"
This was painful to read Ben, but so true. The only useful thing I can add to this post is that the postumous book by David Graeber (along with David Wengrow) "The Dawn of Everything" was a delightful inquiry into the long storied propensity of humans to construct institutions that become ends to justify means, rather than means to pursue ends.