Thank you very much for writing this great piece. A couple of rough comments that come to my mind:
- as you mentioned, it seems that the problems of peer review are exacerbated by the pressure to publish that researchers feel. That brings to an increases quantity of papers (see salamy slicing) with (at best) incremental contributions. Hence, this causes increases pressure on the peer review system. Moreover, it may be that peer review itself is made harder for "incremental" papers than papers that drastically advance the state of the field (although paradigm-shifting papers are also very hard, as you mentioned)
- maybe the ideal system is one where "peer review" is not a monolithic concept. For instance, it may be that proper peer review is required for papers with strong claims, while "putting a PDF" online is fine for incremental works.
- maybe the bad incentives caused by peer review are reduced if the binary accept/reject decision is converted into a numerical score that the journal gives to the paper
I think the bigger difficulty for researchers wanting to do really ambitious, paradigm-shifting work is getting past the peer review of grant applications rather than getting past peer review to get published. If a paper is rejected somewhere, it's likely to still get accepted somewhere else. If a grant is rejected, there's usually not a lot of other places you can turn to..
I like the idea of allowing short incremental works to be published without review. That could better allow junior researchers to work on longer more ambitious projects while still having a tangible research output every ~year or so they can point to.
I'm curious how you think a numerical score system might improve things.
On numerical score: it's simply a intuition I have, that if the binary decision is converted into a soft score, then there is less incentive to game the system to simply go above the threshold. People may be happy enough with a 6/10 for their work so would not try and hack the system to get a 7 or 8. Not sure how easy that is to implement in practice though.
Some fields like economics and philosophy have incredibly high rejection rates, above 90 per cent for "good" journals. This means that peer review works mainly to deliver judgements of importance rather than to correct errors. In addition the process is slow and (it's said) a lot of referees palm the job off on to their grad students.
Interesting. What fraction of these rejections come from the editor, as opposed to reviewers?
Also, when a paper goes out for peer review and is "rejected," don't the reviewers also provide feedback as to why it was rejected? Presumably some of that feedback could be about correcting errors.
Anecdotally, it seems most rejections come from editors. However, peer reviewers also reject a lot in some journals. They also accept a lot in others, for instance I've never known anyone who was not able to eventually get through peer review and publish in an MDPI journal.
MDPI is a weird journal. It barely has any standards in their review process. But the sorts of folks who send papers there are often solid. So it ends up being a decent place for solid but maybe boring papers that need too much revision to get published elsewhere.
Yes. I've noticed that sometimes interesting,"non-mainstream", potentially important work is published at MDPI. Probably because the author(s) couldn't get it in elsewhere? Erik Hoel's early work comes to mind: https://www.mdpi.com/1099-4300/19/5/188
Rejections at work horse methods journals like JASA, Neuroimage, etc.. often come as a result of reviewer criticisms. But at major science outlets rejections are much more editorial. In fact, they might accept despite reviewer criticisms. Every statistician I know has written to CNS journals at some point about major flaws about a paper under review and typically been ignored. At best, they are willing to reluctantly publish a letter.
Ostensibly, peer-review has reformed to generally require disclosure of any potential conflicts of interest (e.g. financial). Yet, the labs that fund research are only able to test results with funding that may have more strings attached. The patent office in 1941 was turned into (temporarily) the IP storage of the Manhattan Project: https://alexwellerstein.com/publications/wellerstein_patentingthebomb(isis).pdf And today companies exploit an enhanced secret patent option that wasn't previously available to ordinary researchers to slow the incremental optimizations of expired patents, or even innocuous inventions. The companies buying up disruptive startups use the same techniques as the sleazy National Enquirer-https://en.wikipedia.org/wiki/Catch_and_kill, except for a different reason: to prevent the from being developed. My theories:
Google's purchase of Phoneblok's https://en.wikipedia.org/wiki/Project_Ara and then shuttering the project a year later- what profits would Google have lost if they sold a modular phone that used 3rd party interoperable screens that would compete with their own Pixel phone (e.g. PC builder market using ATX motherboards with different chips by Intel and AMD).
I can think of another invention that Intel could be prohibited from developing.
Can anyone really think of any basic science or engineering department that isn't tied to a Technology Transfer conflict of interest? String theory applications might be a century away- fusion, 25 or 50. Even evolution, was turned into an industry:
"Evolution had no immediate payoff. Learning phylogenies did not cure belly ache, and it was still all a bit too daring for regular schoolroom instruction. But Huxley could see a place for evolution. The chief ideological support of those who opposed the reformers—the landowners, the squires, the generals, and the others—came from the Anglican Church. Hence, Huxley saw the need to found his own church, and evolution was the ideal cornerstone. It offered a story of origins, one that (thanks to progress) puts humans at the center and top and that could even provide moral messages. The philosopher Herbert Spencer was a great help here. He was ever ready to urge his fellow Victorians that the way to true virtue lies through progress, which comes from promoting a struggle in society as well as in biology—a laissez-faire socioeconomic philosophy. Thus, evolution had its commandments no less than did Christianity. " -Michael Ruse, Science, 03/2003 https://www.science.org/doi/10.1126/science.1082968
This seems pretty tangental to peer review, but the invention secrecy act is super interesting, and is an interesting 1st Amendment special exception many are not aware of.
I haven't heard of it being abused, but if it is that sounds like a pretty big deal. I'm curious if you have any further evidence of abuse.
George Cove's solar energy device sounds like a thermocouple, which is extremely inefficient and low voltage. As far as the device developed at Bell Labs, the difficulty (as far as I understand) was getting the manufacturing cost down. Others have noted that if the government (or large industry) had properly invested in solar panel manufacturing earlier, we could have had cheap solar panels decades ago.
I'll agree that it's tangential to peer review when a conflict of interest is not present. But if there's research that substantially challenges many root assumptions of a field, such as Linguistics, especially during a nascent era, then even tenured faculty may seem on the defensive. Scientific American had a story a few months ago on Elizabeth Bates, who, in the 1970s, challenged linguists such as Noam Chomsky: https://www.scientificamerican.com/article/elizabeth-bates-and-the-search-for-the-roots-of-human-language/ The interesting thing about that is that the bases for linguistics have plausible and cogent frameworks, such as Chomksy's nativism. But so does hers. Upton Sinclair had a famous quote, which I have written about elsewhere, so I won't write that here.
It's impossible to summarize the article because there are so many important details, but I will try to cover some of the major points:
A timeline of events preceding that led to the Cove story becoming slightly more known in 2023:
In October of 2021, Kris de Decker published an article "How to Build a Low-tech Solar Panel?" Kris de Decker is a former high-tech journalist who used to cover semi-conductor (computer chip companies), and started his own magazine more than a decade ago. His articles are very technical and very, very well researched, despite being accessible in a more magazine-like format.
He quotes a researcher named Philip Pesavento, who used to work at Oak Ridge National Laboratory (or is near retirement). Philip had done research on photovoltaics, but had not published or communicated some of these findings until recently when he notified Kris De Decker about the Smithsonian discrepancy of photographs between Charles Fritt actually being Cove- that was what led me to notify Bellingcat because they have excellent image forensics skills).
Returning to the De Decker research article, some important excerpts:
"According to Philip Pesavento, who has a background in semiconductor engineering, Cove intended to build a better thermoelectric generator (TEG). He exposed his generator to the heat from a wood stove and direct solar energy – Edward Weston had made the first experimental solar thermoelectric generator (or STEG) in 1888."
However, the prehistory of the solar panel may be incomplete. In 2019, I received a mail from a reader of Low-tech Magazine, Philip Pesavento:
“I have been studying an early pioneer in solar cell technology from the pre-WWI era since the early 1990s. I am getting too old to continue doing anything with this, and even though there have been one or two scholarly articles about Mr. Cove, they have completely missed what he accomplished. I am enclosing a PDF of a PowerPoint that I put together back in 2015 and never presented to anyone. If you are interested in pursuing writing a paper yourself, I could mail you a thumb drive with all the background material that I have collected.”
...
"We now know that ZnSb – the negative material in Cove’s plugs – is a semiconductor with a bandgap of 0.5 eV. That largely explains why the inventor initially observed that his solar generator converted both heat and light into electricity. A thermophotovoltaic generator matches not only the infrared tail of the solar spectrum – it also matches the direct spectrum of a burning flame or a red hot emitting surface which is heated by burning wood or natural gas. It also converts the lower portion of the visible spectrum into electricity, be it very inefficiently.
According to Philip Pesavento, Cove then managed to refine the composition of the alloy close to Zn4Sb3 – a zinc-antimony alloy with proportions of 4 parts zinc to 6 parts antimony. That, we now know, is also a semiconductor. However, it has a bandgap of 1.2 eV – very close to the bandgap of silicon (1.1 eV). Consequently, it turned his thermophotovoltaic generator into a photovoltaic generator:
“In his enthusiasm, Cove probably made up a larger number of plugs and somehow got the proportions “wrong” on one batch. He then measured an even larger voltage. Finally, he made a careful study of zinc-antimony alloys and found that the 40-42% range zinc alloy gave the highest voltage (compared to 35% zinc in ZnSb). Having – accidentally – discovered Zn4Sb3, the higher bandgap of this semiconductor meant that it no longer worked when it was exposed to the heat from a wood stove. However, it worked even better when it was exposed to solar energy – because it was now converting far more of the visible spectrum of sunlight efficiently into electricity.”
Using colored glass filters, George Cove determined that most of the response was from the violet end of the spectrum and only a little from the so-called heat rays. His earlier PV plugs had responded equally well to heat rays and violet rays, while the older thermoelectric generators (German silver at both sides) did not respond to the violet rays at all."
So, not only did Dr. Pesavento determine exactly how Cove developed the low-end of the thermoelectric generation of (very inefficient) a bandgap at 0.5 eV, but also discovered how Cove was able to develop an efficient bandgap at 1.2 eV by developing Zn4Sb3 for solar cells!
Not only that, but:
"The primary advantage of Cove’s design would be its low-tech fabrication method. In the 1970s and 1980s, scientists investigated Zn4Sb3 for use in photovoltaics and concluded that the material’s “obvious advantages are apparent simplicity and relatively low temperature of the preparation procedure.” 23 The melting point for Zn4Sb3 is 570 degrees Celsius, while it’s 1,400 degrees for silicon.
"During the 1980s, researchers made important advances in silicon p-n junctions, and interest in alternative configurations waned. However, there has been renewed interest in recent years. For example, research into graphene/silicon Schottky solar cells concludes that “simple and cost-effective device fabrication that does not require high temperatures is one of the advantages.” 26 In other recent studies, scientists conclude that Schottky-type “selenium devices are… extremely simple and cheap to fabricate”. 27 28 29 30"
One of the footnotes quotes a 2011 article on thermal powered factories , and that Selenium can anneal in as little as 200 degrees Celsius... which brings to a very important point- Solar power can manufacture solar cells with very little electricity. I looked up recent advances in Selenium, and one amorphous article recently showed a clean surface technique: https://www.sciencedirect.com/science/article/abs/pii/S1369800124000714?via%3Dihub
And a higher temperature, but laser annealing process (potentially more efficient yield?)"
I don't have any evidence on the Invention Secrecy Act (and not directly referring to Cove or solar), but it mirrors a technique that the NSA uses against FOIA requests- if someone doesn't have proof of something was done unlawfully (e.g. warrantless wiretapping), they not only do not confirm or deny something (Glomar response), but also cannot proceed with a secret court to hear the appeal to section 702: https://www.thenation.com/article/world/nsa-warrantless-surveillance/ Thus, some acts are a Catch 22- one could suggest there is a technology that can identify if digital surveillance took place (e.g. wholesale uploading of data which the NSA could decrypt if it has the computing time, but can't prove whether the decryption took place without referring to an insider/whistleblower's testimony, or illegal activity-not recommended, but then again- some people interpret the Constitution differently). Thus most cases are unprovable.
"During the 1990–91 Gulf War, the shortage of military GPS units caused many troops and their families to buy readily available civilian units. Selective Availability significantly impeded the U.S. military's own battlefield use of these GPS, so the military made the decision to turn it off for the duration of the war."
While I sometimes appear to conflate consumer technology research with national security, from a technical point, many inventions by the military or gov eventually became used for civilian use. And while national secrets are an even more separate category than corporate patents from academic publishing, I think the issue gets stickier once one remembers that the Manhattan Project, Radar Labs, and Computer Developments, largely took place in the universities, and the kind of high tech research done is NDA'd today, suggesting prioritization of defense over internet technology, than might have been somewhat more "direct to consumer" when things like the Gopher protocol, TCP/IP and linux were initially released.
Thank you very much for writing this great piece. A couple of rough comments that come to my mind:
- as you mentioned, it seems that the problems of peer review are exacerbated by the pressure to publish that researchers feel. That brings to an increases quantity of papers (see salamy slicing) with (at best) incremental contributions. Hence, this causes increases pressure on the peer review system. Moreover, it may be that peer review itself is made harder for "incremental" papers than papers that drastically advance the state of the field (although paradigm-shifting papers are also very hard, as you mentioned)
- maybe the ideal system is one where "peer review" is not a monolithic concept. For instance, it may be that proper peer review is required for papers with strong claims, while "putting a PDF" online is fine for incremental works.
- maybe the bad incentives caused by peer review are reduced if the binary accept/reject decision is converted into a numerical score that the journal gives to the paper
Great point RE salami slicing.
I think the bigger difficulty for researchers wanting to do really ambitious, paradigm-shifting work is getting past the peer review of grant applications rather than getting past peer review to get published. If a paper is rejected somewhere, it's likely to still get accepted somewhere else. If a grant is rejected, there's usually not a lot of other places you can turn to..
I like the idea of allowing short incremental works to be published without review. That could better allow junior researchers to work on longer more ambitious projects while still having a tangible research output every ~year or so they can point to.
I'm curious how you think a numerical score system might improve things.
Good point about reviews for grant proposals.
On numerical score: it's simply a intuition I have, that if the binary decision is converted into a soft score, then there is less incentive to game the system to simply go above the threshold. People may be happy enough with a 6/10 for their work so would not try and hack the system to get a 7 or 8. Not sure how easy that is to implement in practice though.
Some fields like economics and philosophy have incredibly high rejection rates, above 90 per cent for "good" journals. This means that peer review works mainly to deliver judgements of importance rather than to correct errors. In addition the process is slow and (it's said) a lot of referees palm the job off on to their grad students.
Interesting. What fraction of these rejections come from the editor, as opposed to reviewers?
Also, when a paper goes out for peer review and is "rejected," don't the reviewers also provide feedback as to why it was rejected? Presumably some of that feedback could be about correcting errors.
Anecdotally, it seems most rejections come from editors. However, peer reviewers also reject a lot in some journals. They also accept a lot in others, for instance I've never known anyone who was not able to eventually get through peer review and publish in an MDPI journal.
MDPI is a weird journal. It barely has any standards in their review process. But the sorts of folks who send papers there are often solid. So it ends up being a decent place for solid but maybe boring papers that need too much revision to get published elsewhere.
Yes. I've noticed that sometimes interesting,"non-mainstream", potentially important work is published at MDPI. Probably because the author(s) couldn't get it in elsewhere? Erik Hoel's early work comes to mind: https://www.mdpi.com/1099-4300/19/5/188
In my experience (economics) it varies from journal to journal, but desk rejections are the minority.
You do get useful feedback, as well as lots that's not so useful.
Rejections at work horse methods journals like JASA, Neuroimage, etc.. often come as a result of reviewer criticisms. But at major science outlets rejections are much more editorial. In fact, they might accept despite reviewer criticisms. Every statistician I know has written to CNS journals at some point about major flaws about a paper under review and typically been ignored. At best, they are willing to reluctantly publish a letter.
Ostensibly, peer-review has reformed to generally require disclosure of any potential conflicts of interest (e.g. financial). Yet, the labs that fund research are only able to test results with funding that may have more strings attached. The patent office in 1941 was turned into (temporarily) the IP storage of the Manhattan Project: https://alexwellerstein.com/publications/wellerstein_patentingthebomb(isis).pdf And today companies exploit an enhanced secret patent option that wasn't previously available to ordinary researchers to slow the incremental optimizations of expired patents, or even innocuous inventions. The companies buying up disruptive startups use the same techniques as the sleazy National Enquirer-https://en.wikipedia.org/wiki/Catch_and_kill, except for a different reason: to prevent the from being developed. My theories:
Solar Power- 40 years before Bell Labs https://journals.lib.unb.ca/index.php/MCR/article/view/17744/22231
https://theconversation.com/if-the-first-solar-entrepreneur-hadnt-been-kidnapped-would-fossil-fuels-have-dominated-the-20th-century-the-way-they-did-215300
Google's purchase of Phoneblok's https://en.wikipedia.org/wiki/Project_Ara and then shuttering the project a year later- what profits would Google have lost if they sold a modular phone that used 3rd party interoperable screens that would compete with their own Pixel phone (e.g. PC builder market using ATX motherboards with different chips by Intel and AMD).
The 1952 Invention Secrecy Act, which is still in effect: https://en.wikipedia.org/wiki/Invention_Secrecy_Act
https://en.wikipedia.org/wiki/Export_of_cryptography_from_the_United_States#PC_era
https://landley.net/notes-2017.html#07-02-2017
I can think of another invention that Intel could be prohibited from developing.
Can anyone really think of any basic science or engineering department that isn't tied to a Technology Transfer conflict of interest? String theory applications might be a century away- fusion, 25 or 50. Even evolution, was turned into an industry:
"Evolution had no immediate payoff. Learning phylogenies did not cure belly ache, and it was still all a bit too daring for regular schoolroom instruction. But Huxley could see a place for evolution. The chief ideological support of those who opposed the reformers—the landowners, the squires, the generals, and the others—came from the Anglican Church. Hence, Huxley saw the need to found his own church, and evolution was the ideal cornerstone. It offered a story of origins, one that (thanks to progress) puts humans at the center and top and that could even provide moral messages. The philosopher Herbert Spencer was a great help here. He was ever ready to urge his fellow Victorians that the way to true virtue lies through progress, which comes from promoting a struggle in society as well as in biology—a laissez-faire socioeconomic philosophy. Thus, evolution had its commandments no less than did Christianity. " -Michael Ruse, Science, 03/2003 https://www.science.org/doi/10.1126/science.1082968
This seems pretty tangental to peer review, but the invention secrecy act is super interesting, and is an interesting 1st Amendment special exception many are not aware of.
I haven't heard of it being abused, but if it is that sounds like a pretty big deal. I'm curious if you have any further evidence of abuse.
George Cove's solar energy device sounds like a thermocouple, which is extremely inefficient and low voltage. As far as the device developed at Bell Labs, the difficulty (as far as I understand) was getting the manufacturing cost down. Others have noted that if the government (or large industry) had properly invested in solar panel manufacturing earlier, we could have had cheap solar panels decades ago.
I'll agree that it's tangential to peer review when a conflict of interest is not present. But if there's research that substantially challenges many root assumptions of a field, such as Linguistics, especially during a nascent era, then even tenured faculty may seem on the defensive. Scientific American had a story a few months ago on Elizabeth Bates, who, in the 1970s, challenged linguists such as Noam Chomsky: https://www.scientificamerican.com/article/elizabeth-bates-and-the-search-for-the-roots-of-human-language/ The interesting thing about that is that the bases for linguistics have plausible and cogent frameworks, such as Chomksy's nativism. But so does hers. Upton Sinclair had a famous quote, which I have written about elsewhere, so I won't write that here.
Regarding George Cove, I researched a lot about him starting in late 2022, after reading a very in depth 2021 article here: https://solar.lowtechmagazine.com/2021/10/how-to-build-a-low-tech-solar-panel/
It's impossible to summarize the article because there are so many important details, but I will try to cover some of the major points:
A timeline of events preceding that led to the Cove story becoming slightly more known in 2023:
In October of 2021, Kris de Decker published an article "How to Build a Low-tech Solar Panel?" Kris de Decker is a former high-tech journalist who used to cover semi-conductor (computer chip companies), and started his own magazine more than a decade ago. His articles are very technical and very, very well researched, despite being accessible in a more magazine-like format.
He quotes a researcher named Philip Pesavento, who used to work at Oak Ridge National Laboratory (or is near retirement). Philip had done research on photovoltaics, but had not published or communicated some of these findings until recently when he notified Kris De Decker about the Smithsonian discrepancy of photographs between Charles Fritt actually being Cove- that was what led me to notify Bellingcat because they have excellent image forensics skills).
https://www.bellingcat.com/news/2023/08/16/untangling-the-mystery-of-the-worlds-first-rooftop-solar-panel/
I also posted my comments on that article here:
https://hackaday.com/2023/08/19/the-mysterious-case-of-the-disappearing-inventor/#comment-6674463
Returning to the De Decker research article, some important excerpts:
"According to Philip Pesavento, who has a background in semiconductor engineering, Cove intended to build a better thermoelectric generator (TEG). He exposed his generator to the heat from a wood stove and direct solar energy – Edward Weston had made the first experimental solar thermoelectric generator (or STEG) in 1888."
However, the prehistory of the solar panel may be incomplete. In 2019, I received a mail from a reader of Low-tech Magazine, Philip Pesavento:
“I have been studying an early pioneer in solar cell technology from the pre-WWI era since the early 1990s. I am getting too old to continue doing anything with this, and even though there have been one or two scholarly articles about Mr. Cove, they have completely missed what he accomplished. I am enclosing a PDF of a PowerPoint that I put together back in 2015 and never presented to anyone. If you are interested in pursuing writing a paper yourself, I could mail you a thumb drive with all the background material that I have collected.”
...
"We now know that ZnSb – the negative material in Cove’s plugs – is a semiconductor with a bandgap of 0.5 eV. That largely explains why the inventor initially observed that his solar generator converted both heat and light into electricity. A thermophotovoltaic generator matches not only the infrared tail of the solar spectrum – it also matches the direct spectrum of a burning flame or a red hot emitting surface which is heated by burning wood or natural gas. It also converts the lower portion of the visible spectrum into electricity, be it very inefficiently.
According to Philip Pesavento, Cove then managed to refine the composition of the alloy close to Zn4Sb3 – a zinc-antimony alloy with proportions of 4 parts zinc to 6 parts antimony. That, we now know, is also a semiconductor. However, it has a bandgap of 1.2 eV – very close to the bandgap of silicon (1.1 eV). Consequently, it turned his thermophotovoltaic generator into a photovoltaic generator:
“In his enthusiasm, Cove probably made up a larger number of plugs and somehow got the proportions “wrong” on one batch. He then measured an even larger voltage. Finally, he made a careful study of zinc-antimony alloys and found that the 40-42% range zinc alloy gave the highest voltage (compared to 35% zinc in ZnSb). Having – accidentally – discovered Zn4Sb3, the higher bandgap of this semiconductor meant that it no longer worked when it was exposed to the heat from a wood stove. However, it worked even better when it was exposed to solar energy – because it was now converting far more of the visible spectrum of sunlight efficiently into electricity.”
Using colored glass filters, George Cove determined that most of the response was from the violet end of the spectrum and only a little from the so-called heat rays. His earlier PV plugs had responded equally well to heat rays and violet rays, while the older thermoelectric generators (German silver at both sides) did not respond to the violet rays at all."
So, not only did Dr. Pesavento determine exactly how Cove developed the low-end of the thermoelectric generation of (very inefficient) a bandgap at 0.5 eV, but also discovered how Cove was able to develop an efficient bandgap at 1.2 eV by developing Zn4Sb3 for solar cells!
Not only that, but:
"The primary advantage of Cove’s design would be its low-tech fabrication method. In the 1970s and 1980s, scientists investigated Zn4Sb3 for use in photovoltaics and concluded that the material’s “obvious advantages are apparent simplicity and relatively low temperature of the preparation procedure.” 23 The melting point for Zn4Sb3 is 570 degrees Celsius, while it’s 1,400 degrees for silicon.
"During the 1980s, researchers made important advances in silicon p-n junctions, and interest in alternative configurations waned. However, there has been renewed interest in recent years. For example, research into graphene/silicon Schottky solar cells concludes that “simple and cost-effective device fabrication that does not require high temperatures is one of the advantages.” 26 In other recent studies, scientists conclude that Schottky-type “selenium devices are… extremely simple and cheap to fabricate”. 27 28 29 30"
One of the footnotes quotes a 2011 article on thermal powered factories , and that Selenium can anneal in as little as 200 degrees Celsius... which brings to a very important point- Solar power can manufacture solar cells with very little electricity. I looked up recent advances in Selenium, and one amorphous article recently showed a clean surface technique: https://www.sciencedirect.com/science/article/abs/pii/S1369800124000714?via%3Dihub
And a higher temperature, but laser annealing process (potentially more efficient yield?)"
https://www.pv-magazine.com/2023/09/04/selenium-solar-cell-built-with-laser-annealing-achieves-record-fill-factor-of-63-7/
I don't have any evidence on the Invention Secrecy Act (and not directly referring to Cove or solar), but it mirrors a technique that the NSA uses against FOIA requests- if someone doesn't have proof of something was done unlawfully (e.g. warrantless wiretapping), they not only do not confirm or deny something (Glomar response), but also cannot proceed with a secret court to hear the appeal to section 702: https://www.thenation.com/article/world/nsa-warrantless-surveillance/ Thus, some acts are a Catch 22- one could suggest there is a technology that can identify if digital surveillance took place (e.g. wholesale uploading of data which the NSA could decrypt if it has the computing time, but can't prove whether the decryption took place without referring to an insider/whistleblower's testimony, or illegal activity-not recommended, but then again- some people interpret the Constitution differently). Thus most cases are unprovable.
One less-classified invention is GPS: https://en.wikipedia.org/wiki/Error_analysis_for_the_Global_Positioning_System#Selective_Availability The interesting story is that while GPS was initially limited in use, it's performance became so well that the military considered it "dual use."
"During the 1990–91 Gulf War, the shortage of military GPS units caused many troops and their families to buy readily available civilian units. Selective Availability significantly impeded the U.S. military's own battlefield use of these GPS, so the military made the decision to turn it off for the duration of the war."
While I sometimes appear to conflate consumer technology research with national security, from a technical point, many inventions by the military or gov eventually became used for civilian use. And while national secrets are an even more separate category than corporate patents from academic publishing, I think the issue gets stickier once one remembers that the Manhattan Project, Radar Labs, and Computer Developments, largely took place in the universities, and the kind of high tech research done is NDA'd today, suggesting prioritization of defense over internet technology, than might have been somewhat more "direct to consumer" when things like the Gopher protocol, TCP/IP and linux were initially released.