In gender medicine, it often doesn't matter if the letters to the editor are published, they are simply ignored.
E.g., the Cass Review of pediatric gender dysphoria had a few critiques which were quickly rebutted, in the peer reviewed literature, but the incorrect critiques continue to be quoted (and the HHS report received similar unsupported accusations by some of the same people). The whole field has several widely quoted "facts" unsupported by, or even in conflict with, the evidence. [E.g. these rebuttals: https://adc.bmj.com/content/110/4/251.long , https://www.tandfonline.com/doi/10.1080/0092623X.2025.2455133 are simply ignored and the people who wrote the error-ridden paper(s) to which they are responding continue to have high authority at major US institutions.]
That's for the responses which are published. The others get rejected or held up for so many months that the incorrect information has been integrated into "common knowledge" by the time they appear.
It is unbelievable.
I do want to say that a peer reviewed letter to the editor will carry more weight than an editor reviewed one, if accepted.For instance, a non peer reviewed letter to the editor cannot be used to correct an incorrect statement (from the incorrect target article) in Wikipedia last I checked. But the time delay is a big problem, indeed.
This was great concise and focused arrow into the heart (of darkness) letter-review process, and while I knew it existed, it is more or less astonishing to me in 2025 how this exists in this form. I'm in an amused mood.
I decided to avoid academia in the late 80's because the entire system seemed hopelessly embedded in social prestige networks, and the thought nauseated me. I enjoy applause as much as anyone - I'm human - but I think my life depending on it is risky.
However around 2020 when COVID came to the fore, I felt quite nervous unless I understood what was happening at an epidemiological, immune system, and systemic approach level, and that's when the Academic Charlatan Parade became vivid.
I remember reading Jay Bhattacharya's quite surprising research on COVID prevalence around April that year, which pointed to 150% infection rate in NYC (LOL). Naturally, when confronted, he doubled-down. The damn paper is _still around_! Elementary statistical failure (arithmetic really). That's when I looked up good old Dr. Money's Gender Reimer work - and that medical atrocity is still around too. There is no quality feedback, if I use a tricky word - a sort of complete fiction (gender control!), and the volume of non-replicable research in soft sciences (much less epidemiology) has assumed catastrophic proportions. I am not alone in the observation.
It really has simply become gatekeeping and grand poo-bah management, in a Japanese-tea-ceremony flow of signals which are about delaying or avoiding quality. I love the idea of social rituals - Catholic Mass, Japanese Tea Ceremony, the never-ending end to a dinner in the summer in France... but this seems to be a social ritual pretending to be a quality control system.
My metier is complex systems, let me pencil on an envelope what I found actual research should resemble ( Iwas asked years ago by an Academic consortium in an unnamed Country which was looking for technological support to reduce cycle time and improve delivery (not precisely publication) of research.)
I broke up the information production process (Observe → Explore → Model → Test → Revise → Share → Correct) into a pipeline:
Plan: research question, hypothesis statement, success criteria, uncertainty bounds, intended use cases. These objects define demand, scope, and acceptance conditions for the information product, and the control structure for results.
Source: prior publications, datasets, experimental protocols, code libraries, instruments, assumptions, domain constraints. These are treated as input materials with varying quality and provenance, but all asserted, including exclusions.
Make: intermediate analyses, models, simulations, experimental results, figures, tables, narrative drafts. These objects transform sourced inputs into candidate knowledge claims.
Quality (Make sub-step): validation reports, robustness tests, sensitivity analyses, reproducibility artifacts, error logs, reviewer annotations. These objects explicitly measure fitness-for-use rather than subjective approval.
Deliver: published versions, metadata, confidence labels, version history, access links, machine-readable summaries. These objects enable distribution, consumption, and downstream reuse.
Return: corrections, revisions, replication results, retraction notices, root-cause analyses, updated versions. These objects close the loop by feeding defects and improvements back into planning and sourcing.
This externalizes visibility to each step of knowledge production, which then allows direct measurement of interior processes, likewise it makes the production process itself hard to falsify (seriously folks, we can trace a semiconductor chip to manufacturer/lot/day/machine but we can't tell if a researcher falsified images? Really?) People have letter debates in the letter-loop if someone didn't perform elementary statistics correctly. Letter-loop are Indirect, diffuse, with unmeasurable feedback with post-hoc alteration of knowledge trajectories - what could go wrong!
Software engineering and research in AI is, by contrast, shocking. Often goals published, research done, code and datasets published, replication is a "GitHub" command. The volume of valuable research is escalating, and always, always replicable as it is in engineering, chemistry, and other hard sciences. Money will flow to replicable research, with defined outcomes that are valuable to a community more and more, and unless soft sciences get it together, I predict in 30 years or less funding will descend to nil. It's an old debate STEM vs HUMANITIES on the budget showdown, but this isn't even debatable anymore, as you have shown.
Some of my recommendations were taken, some were a bridge too far because the system itself is broken (the letter loop), but diffuse ownership blocks systemic change.
Peer review is applicable at a tiny remote step in the system. It's present, like an appendix, in so many ways.
But change is happening nonetheless because engineering practices using modern technology are galloping. And some groups are debating categories.
[I read "toxic masculinity" link, which made my afternoon. Someone elsewhere on substack posted a critique of sociology, which like the "toxic masculinity" survey, or Canadian parents offering children to trans surveys - well, they are are all systems of surveys (if a survey is research) of adherence to a belief system, not a predictive test of reality. How do you tell them that? Why are adherence reports being funded (to your point).]
Here are metrics I would use for actual research, to initialize review by impartial jurors of actual measurements, not belief adherence. Again, good read.
Reproducibility rate (% of results independently reproduced)
Post-publication correction rate (defects per paper-year)
Time-to-error-detection (mean time until first substantive correction)
Method completeness score (presence of data, code, assumptions)
Reviewer accuracy score (alignment with later replication outcomes)
Claim survival half-life (time until material revision or invalidation)
Time-to-first-public-release
Review cycle time (submission → public availability)
Time-to-correction (error identified → corrected version released)
Iteration velocity (versions per year)
Latency to replication signal (publication → first replication attempt)
Thank yo for the post. Given to both sides of the issue:
-s***w the idiots, let them rot
-don your whacko-hunting gear
It would seem in this time/place, the 2nd is far better. Too often such lunacy proffers 'proof-of-publication' as certification of validity. The slightest scraping, removes the paint-on-the-pig to discover the horrendous toxins just below the surface.
Let me also offer that exposing the content, means it can be searched/indexed/vectorized and then used as response fodder by LLM (large language models) generating summarization and direction to such works and their sources.
The single greatest value of AI (assistive intellect) chat interfaces may be the surface exposure of such cretinous bile, with footnotes attached. Add a bit of competing hard science content, and you have a weak but usable arguing-bot to engage and hammer the lunacy.
Just a thought for dealing with those bent upon wedding shrimp or mating with the continent.
I've been unmistakeably chastized by AI for even asking certain questions & then am condescendingly directed toward the "correct" stance--ie, answers to questions I did not ask. You can just feel the disdain.
Thank you for your hard work and persistence to root out the rot in academia.
Exposing BS is ALWAYS good so keep up the excellent - and hard - work.
Because, contrary to the old cliché, NO publicity is always better than BAD publicity.
Thank you!
In gender medicine, it often doesn't matter if the letters to the editor are published, they are simply ignored.
E.g., the Cass Review of pediatric gender dysphoria had a few critiques which were quickly rebutted, in the peer reviewed literature, but the incorrect critiques continue to be quoted (and the HHS report received similar unsupported accusations by some of the same people). The whole field has several widely quoted "facts" unsupported by, or even in conflict with, the evidence. [E.g. these rebuttals: https://adc.bmj.com/content/110/4/251.long , https://www.tandfonline.com/doi/10.1080/0092623X.2025.2455133 are simply ignored and the people who wrote the error-ridden paper(s) to which they are responding continue to have high authority at major US institutions.]
That's for the responses which are published. The others get rejected or held up for so many months that the incorrect information has been integrated into "common knowledge" by the time they appear.
It is unbelievable.
I do want to say that a peer reviewed letter to the editor will carry more weight than an editor reviewed one, if accepted.For instance, a non peer reviewed letter to the editor cannot be used to correct an incorrect statement (from the incorrect target article) in Wikipedia last I checked. But the time delay is a big problem, indeed.
This was great concise and focused arrow into the heart (of darkness) letter-review process, and while I knew it existed, it is more or less astonishing to me in 2025 how this exists in this form. I'm in an amused mood.
I decided to avoid academia in the late 80's because the entire system seemed hopelessly embedded in social prestige networks, and the thought nauseated me. I enjoy applause as much as anyone - I'm human - but I think my life depending on it is risky.
However around 2020 when COVID came to the fore, I felt quite nervous unless I understood what was happening at an epidemiological, immune system, and systemic approach level, and that's when the Academic Charlatan Parade became vivid.
I remember reading Jay Bhattacharya's quite surprising research on COVID prevalence around April that year, which pointed to 150% infection rate in NYC (LOL). Naturally, when confronted, he doubled-down. The damn paper is _still around_! Elementary statistical failure (arithmetic really). That's when I looked up good old Dr. Money's Gender Reimer work - and that medical atrocity is still around too. There is no quality feedback, if I use a tricky word - a sort of complete fiction (gender control!), and the volume of non-replicable research in soft sciences (much less epidemiology) has assumed catastrophic proportions. I am not alone in the observation.
It really has simply become gatekeeping and grand poo-bah management, in a Japanese-tea-ceremony flow of signals which are about delaying or avoiding quality. I love the idea of social rituals - Catholic Mass, Japanese Tea Ceremony, the never-ending end to a dinner in the summer in France... but this seems to be a social ritual pretending to be a quality control system.
My metier is complex systems, let me pencil on an envelope what I found actual research should resemble ( Iwas asked years ago by an Academic consortium in an unnamed Country which was looking for technological support to reduce cycle time and improve delivery (not precisely publication) of research.)
I broke up the information production process (Observe → Explore → Model → Test → Revise → Share → Correct) into a pipeline:
Plan: research question, hypothesis statement, success criteria, uncertainty bounds, intended use cases. These objects define demand, scope, and acceptance conditions for the information product, and the control structure for results.
Source: prior publications, datasets, experimental protocols, code libraries, instruments, assumptions, domain constraints. These are treated as input materials with varying quality and provenance, but all asserted, including exclusions.
Make: intermediate analyses, models, simulations, experimental results, figures, tables, narrative drafts. These objects transform sourced inputs into candidate knowledge claims.
Quality (Make sub-step): validation reports, robustness tests, sensitivity analyses, reproducibility artifacts, error logs, reviewer annotations. These objects explicitly measure fitness-for-use rather than subjective approval.
Deliver: published versions, metadata, confidence labels, version history, access links, machine-readable summaries. These objects enable distribution, consumption, and downstream reuse.
Return: corrections, revisions, replication results, retraction notices, root-cause analyses, updated versions. These objects close the loop by feeding defects and improvements back into planning and sourcing.
This externalizes visibility to each step of knowledge production, which then allows direct measurement of interior processes, likewise it makes the production process itself hard to falsify (seriously folks, we can trace a semiconductor chip to manufacturer/lot/day/machine but we can't tell if a researcher falsified images? Really?) People have letter debates in the letter-loop if someone didn't perform elementary statistics correctly. Letter-loop are Indirect, diffuse, with unmeasurable feedback with post-hoc alteration of knowledge trajectories - what could go wrong!
Software engineering and research in AI is, by contrast, shocking. Often goals published, research done, code and datasets published, replication is a "GitHub" command. The volume of valuable research is escalating, and always, always replicable as it is in engineering, chemistry, and other hard sciences. Money will flow to replicable research, with defined outcomes that are valuable to a community more and more, and unless soft sciences get it together, I predict in 30 years or less funding will descend to nil. It's an old debate STEM vs HUMANITIES on the budget showdown, but this isn't even debatable anymore, as you have shown.
Some of my recommendations were taken, some were a bridge too far because the system itself is broken (the letter loop), but diffuse ownership blocks systemic change.
Peer review is applicable at a tiny remote step in the system. It's present, like an appendix, in so many ways.
But change is happening nonetheless because engineering practices using modern technology are galloping. And some groups are debating categories.
[I read "toxic masculinity" link, which made my afternoon. Someone elsewhere on substack posted a critique of sociology, which like the "toxic masculinity" survey, or Canadian parents offering children to trans surveys - well, they are are all systems of surveys (if a survey is research) of adherence to a belief system, not a predictive test of reality. How do you tell them that? Why are adherence reports being funded (to your point).]
Here are metrics I would use for actual research, to initialize review by impartial jurors of actual measurements, not belief adherence. Again, good read.
Reproducibility rate (% of results independently reproduced)
Post-publication correction rate (defects per paper-year)
Time-to-error-detection (mean time until first substantive correction)
Method completeness score (presence of data, code, assumptions)
Reviewer accuracy score (alignment with later replication outcomes)
Claim survival half-life (time until material revision or invalidation)
Time-to-first-public-release
Review cycle time (submission → public availability)
Time-to-correction (error identified → corrected version released)
Iteration velocity (versions per year)
Latency to replication signal (publication → first replication attempt)
Cost per validated claim
Reviewer labor hours per paper
Rework cost (corrections, withdrawals, resubmissions)
Wasted review effort (reviews on rejected or duplicated work)
Opportunity cost of delay (time × downstream impact blocked)
Infrastructure cost per publication (platform, admin, editorial)
Downstream reuse rate (citations with substantive use, not mention)
Inclusion in applied systems (standards, policy, products)
Decision impact score (changes decisions or models)
Clarity score (independent restatement accuracy)
Machine-readability score (data/code usable without manual repair)
Cross-domain transfer rate (use outside original field)
Defect discovery rate over time
Root-cause closure time
Process improvement adoption rate
Repeat error frequency
Reviewer learning curve (quality improvement over time)
Thank yo for the post. Given to both sides of the issue:
-s***w the idiots, let them rot
-don your whacko-hunting gear
It would seem in this time/place, the 2nd is far better. Too often such lunacy proffers 'proof-of-publication' as certification of validity. The slightest scraping, removes the paint-on-the-pig to discover the horrendous toxins just below the surface.
Let me also offer that exposing the content, means it can be searched/indexed/vectorized and then used as response fodder by LLM (large language models) generating summarization and direction to such works and their sources.
The single greatest value of AI (assistive intellect) chat interfaces may be the surface exposure of such cretinous bile, with footnotes attached. Add a bit of competing hard science content, and you have a weak but usable arguing-bot to engage and hammer the lunacy.
Just a thought for dealing with those bent upon wedding shrimp or mating with the continent.
I've been unmistakeably chastized by AI for even asking certain questions & then am condescendingly directed toward the "correct" stance--ie, answers to questions I did not ask. You can just feel the disdain.