Collusion to Misunderstand Health Risk: How Peer-Reviewed, Academic Publications, Institutions, News Outlets, And ChatGPT Deceive You

At a base level, the population does not understand risk. Less than a third of educated respondents are risk literate: https://www.allianz.com/de/ueber-uns/brand/brand-in-action/making-cents/risk-literacy-and-choices.html 


“Around two-thirds of the population have little to no awareness of the health, work, housing, liability, and cyber risks” -
https://group.vig/en/investor-relations/ir-news/ir-meldungen/everyday-risks-international-study-reveals-low-risk-literacy-among-the-population/


Risks of death are one and the same with health risk. The number one cause of death is heart disease. It is the top risk, both globally and in the US. Cancer is a relatively close second. Accidents, like falling or overdosing on “safe” medications, are third. The next seven are infections/illnesses and organ damage: https://www.cdc.gov/nchs/fastats/deaths.htm


Every one of these top ten has a substantial decrease in risk through lifestyle and behavioral alteration. Yet when we observe trends in the most recent decades, health risk is only climbing. The population is not getting healthier. One might wonder why the top health risks have skyrocketed in the past fifty years if public health authorities, scientists, and media outlets have been correct, honest, and effective in their messaging. Upon examination, it appears as though they collude to make and keep the population risk illiterate through various forms of misdirection, distraction, and intentional deceptions.


The number one killer is reduced through unconditional forgiveness:


https://pubmed.ncbi.nlm.nih.gov/17466400/
https://pubmed.ncbi.nlm.nih.gov/14593849/
https://pmc.ncbi.nlm.nih.gov/articles/PMC5055412/
https://fincham.info/papers/2014-ajcardiology.pdf
https://pmc.ncbi.nlm.nih.gov/articles/PMC10120569/
https://www.apa.org/monitor/2017/01/ce-corner
https://www.cnn.com/2019/06/05/health/forgiveness-health-explainer
https://greatergood.berkeley.edu/article/item/the_new_science_of_forgiveness


Some of the most informed and educated people are shocked when learning this. Might that be a problem, considering the grudge culture in which we now live? Being agitated by disagreements, politics, and differing opinions is a common phenomenon. In fact, there are whole ideological movements which claim that words are violence, and that any degree of disagreement is unforgivable. Does this popular phenomenon seem to be addressing the number one risk? Is constant political grudge a risk-literate behavior/belief? Is it improving the mental and physical health outcomes at a population level?


The number one and two killers are cut by ceasing tobacco, ceasing alcohol, losing weight, and exercising:


https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/tobacco-use-in-adults-and-pregnant-women-counseling-and-interventions
https://pmc.ncbi.nlm.nih.gov/articles/PMC7935481/
https://ebccp.cancercontrol.cancer.gov/recommendation.do?topicId=102271


https://ascopubs.org/doi/10.1200/EDBK_200093
https://www.who.int/news/item/03-02-2026-four-in-ten-cancer-cases-could-be-prevented-globally


https://pmc.ncbi.nlm.nih.gov/articles/PMC12010775/
https://www.cancer.org/cancer/risk-prevention/diet-physical-activity/acs-guidelines-nutrition-physical-activity-cancer-prevention.html
https://www.iarc.who.int/featured-news/world-cancer-day-2021-physical-activity/


The number three killer is reduced in risk if you just get off your phone and lift weights.


Is it fair to say that these risks and their interventions dominate social media posts, media headlines, and trending scientific discussions? Just in your day-to-day interactions with people, are these objective measurable facts of the undeniable top risks a regular talking point? Are they as seriously wrestled with as the trending social media outrage of the day? 


Instead, a very popular answer about top risks is gun violence. The majority of American respondents rank this quite high in risk evaluation. In a 2023 poll, 70% of respondents ranked gun violence as THE top concern: https://source.washu.edu/2025/07/gun-violence-remains-top-st-louis-public-health-concern-but-mental-health-addiction-rising/ . Respondents even ranked food insecurity and extreme weather ahead of the real top three risks. But nearly 60% of gun violence is suicide. And a very modest estimate is 15% for gang violence. Mass shooting is a vanishingly small percent. Even attempting to use the most alarmist and biased interpretations, a perpetrator using gun violence against you does not rank in your top fifty risks. There is no way the public could conceivably know this from the way it’s portrayed by public health authorities or media outlets. Largely, what we see is entertainment sensationalism blotting out the productive conversation. Specifically, what we see in a discussion like this is exploitation of something called availability heuristic and a subset of base-rate fallacy called base-rate neglect. For the discerning reader, know that these two phenomena are constantly working to misinform you. Essentially, once you succumb to emotionally powerful narratives, you abandon accuracy to focus on rare or nonexistent events.


As we venture further into specified discussions of risk, the degree and volume of lying increases. It is not simply a phenomenon of news headline sensationalism or mistaken wording provoking your outrage/attention. Experts appear to want you to misunderstand truth. In scientific publications, the representation of vested interests generates extreme bias. We could go down the list of the opioid epidemic, Vioxx, thalidomide, tobacco company-funded research, and a litany of other recent scandals where officials approved dangerous treatments and scientific consensus is driven by lies from different industry pressures. The blog has covered this before:


https://www.elev8wellness.com/wellblog_best_nutrition_training_coaching_experts/conflicts-of-interest-in-science-and-human-health-have-reached-the-tipping-point


https://www.elev8wellness.com/wellblog_best_nutrition_training_coaching_experts/why-cant-americans-get-healthy


Experts want the public to believe the wrong things, or at the very least remain in a confused and highly impressionable state. And as much as we could cover many MANY more examples (like how in 1971 researchers found a link between Johnson and Johnson talcum powder and ovarian cancer, but hid the truth for over a decade), Iet us focus on HOW liars ruin institutional science messaging to damage the public’s understanding of health risk through a variety of intentional abuses of methodology.


On this subject there are a multitude of ways which people who think they are smart dupe the general populace. This article will stick to only three (but do keep availability heuristic and base-rate neglect in mind), because there are too many to cover. The reader must understand that the number of ways news and science messaging ARE manipulated is beyond a normal person’s imagination. A normal person is not looking to collude to deceive others at an industrial scale. Thus, the breadth of mendacity is difficult to apprehend. And a normal person does not have a zealous, religious commitment to upholding the dogma that modern authorities are infallible gods without bias, error, or mistake. So a normal person can’t really fathom the degree to which numbers can be manipulated and approved even by peer review, seemingly genuinely, to misrepresent reality.


This article will cover just three abuses of methodology which are purposely weaponized to confuse the public about risk: the measure of and emphasis on relative risk; biased endpoints; and biased duration of study. These aren’t the only ones. These may not even be the most prevalent. But they are a few sleights-of-hand which the layperson and even specialist miss most of the time. They also happen to be the very means by which the tobacco lobby made their case, how the FDA approved Vioxx and OxyContin as safe, and essentially how all medical products and drugs are still assessed for effectiveness and/or safety.


Relative risk is a statistical term to measure the difference in percent between two percents or incident rates (of risk between two groups). Let’s say that your normal risk of death from roller skating is 1% per thousand hours of activity (ie - out of a large sampling, one person dies for every one-hundred people who each roller skate one-thousand hours). And let’s say that wearing a headband lowers your risk to 0.8% per thousand hours. Your change in risk is 0.2%. And because 0.2% is 20% of 1%, a biased statistician argues the difference between 1% and 0.8% is 20%. That is relative risk in a nutshell. Calculated upward, 0.2% is 25% of 0.8%. It is often first represented as a ratio toward the more dangerous group (NOT wearing headband has 1.25 risk ratio compared to wearing). And there is some more unpacking to be done in a longer discussion about baseline elsewhere. But for now, consider that in news summaries, the lay audience will generally see this figure in the percent format, again, because it LOOKS big and scary. Clearly, the risk rise or decrease is actually 0.2%. This is real risk, sometimes called “absolute risk” and “absolute risk reduction” in order to sell the idea that there are multiple entirely legitimate ways of looking at risk. There aren’t. There is only one way. Liars love to tangle themselves in knots by downplaying ARR (absolute risk reduction) and playing up RRR (relative risk reduction). And if the headband lobby wanted to sell a lot of headbands, we could guarantee that we would be reading about life-saving headbands which lower risk AT LEAST 20%. Or even more deceptively, they’d perhaps phrase it that you have a 25% increased risk of death by NOT wearing that headband. Deep in the double-asterisked footnote, maybe we’d find the 20% figure is an RRR. But by then, there would already be headband mandates. And many people might not allow children to enroll in school without a headband. All of this over a possible 0.2% difference in risk. Never mind that at ten hours of roller skating your risk is 0.01%; thus, your risk with or without the headband would only change by 0.0002% per hour of activity. This is why Mark Twain popularized the statement, “lies, damned lies, and statistics.” 


In a way, relative risk is entirely a fiction. It is a legitimate tool in dataset analysis; but it simply does not honestly translate over to many risk choices in the real world. No one uses relative risk in his personal life. No one. Earnestly and in good faith ask professional epidemiologists or statisticians if they refuse to touch the music and climate controls in their cars while driving. They will have no clue what you’re referencing. But manipulating music controls raises risk 190%; and manipulating climate control raises risk 460%: https://usnddc.org/wp-content/uploads/2023/03/Crash-risk-driver-distraction_lit-review_10-1.pdf. To put this in perspective, smoking cigarettes raises risk of all cancers 150 to 300%. You drop your risk of motor vehicle death more by ceasing music and climate controls than you lower risk of anything else in all of public health by engaging in the experts’ recommended interventions. Reread that sentence. Reread it again. But no one ceases car controls, because anyone with an IQ above 85 knows full well that absolute risk reigns supreme. That is, instead of 2-5 deaths per 10,000 hours of driving, risk will be closer to 1 death for every 10,000 hours of driving if you touch no dashboard knobs and buttons. Your intelligent brain knows the absolute risk is all that matters. The absolute risk reduction is less than a tenth of one percent. So you don’t care. Even the biggest fraudster zealots who push relative risk narrative in every other discussion do not care AT ALL about relative risk in the real world. But you better believe that if someone could make a billion dollars from people never touching music control or climate control dials in a car, we’d all be familiar with the figures.


Relative risk is absolutely ludicrous when applied as scaremongering lies at population level research. Imagine we study two groups, one with a 2 in 100,000 incidence of death and another with a 6 in 100,000 incidence. Even if we can guarantee a strong causal variable (we cannot), the difference in risk is incredibly tiny (if it exists at all - do keep in mind that any two groups will have small variances due to no single or obvious causal factor). It’s a 0.004% difference. But the way it most often will be represented in a peer-reviewed journal is a 3.0 ratio or 300% “of” or 200% additional rise in risk. This is relative risk. Your real risk (absolute risk) hasn’t changed meaningfully. There was no reshuffling of the top ten risks. There was no sizable statistical signal. But a “300% increased risk” will command headlines if someone stands to gain money or power from it. It will command headlines if it secures the current trending lie in a scientific consensus. If it can agitate people, it will go viral in social media. That number will not gain traction if it stands to unseat power. It will not gain notice if it's sober-minded. It has little to do with accuracy. It has everything to do with money and authority. We incorrectly believe that every modern scientific paradigm succeeds because of how aligned it is with actual truth; but do not forget Thomas Kuhn’s correction on this matter: “A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.” That’s the repeated historical reality without exception.


Even in studies where no one overtly stands to profit, relative risk is so normalized as the discussion point that it still misleads educated onlookers. Food allergies have skyrocketed in the past forty years in the Western world. Thanks to relative risk, many people have confused themselves about the causes and the interventions because of one single underpowered study whose results were hyperbolized via relative risk lies. That study, the LEAP study, only examined peanut allergy. Often, people tout its findings as discovering that early exposure eliminates all allergies; and avoidance of exposure is what causes allergies. But that isn’t what the study found at all. Infants in the study who were exposed to peanuts ended up with an 11% incidence of allergy at 60 months: https://www.nejm.org/doi/full/10.1056/NEJMoa1414850. That’s higher than the global baseline of childhood peanut allergy by a lot (less than 1%). To be clear, exposed infants in this study ended up with an ELEVEN-FOLD higher incidence of peanut allergy than the average person in the non-Western world. That’s 1,100% in relative risk language. In the non-Western world, childhood peanut allergy prevalence is as low as 0.0% in rural settings (eg - South Africa) and often no higher than 0.22% in some urban settings (eg - Korea) whether or not infants had ANY exposure (ie - rural South Africans and Koreans do NOT include peanuts in early childhood foods). Moreover, nine children in the early exposure group of the LEAP study had to discontinue participation in exposure because they were developing such severe reaction. Their data is excluded from the final results. With their data excluded, both the exposure and non-exposure groups had similar rates of hospitalization. Thus, in practical terms, there was very little difference. A difference, yes. But not an explain-all level of difference. The primary discovery was that on a skin prick test, very young infants had a 10% reduction in positive allergy tests if exposed to peanut early. At 60 months, the reduction was about 15%. This figure is misconstrued all the time, even by pediatric allergists, as a 70% to 81% (relative risk) reduction. Worse than that, some “quote” the study as proof that exposure eliminates allergies. No. That is an egregious abuse of the data. Early exposure does seem to be a valuable tool for some people some of the time. It is nowhere near eliminating. It amounts to a ten to fifteen percent (absolute risk) reduction on average. The sum total explanation of food allergies is complex and is strongly related to beneficial bacterial exposure, such that cesarean births and maternal or infant use of antibiotics raises risk. We see clearly that rural communities with no exposure to peanuts often end up with no allergy to it at all. So exposure makes for a terrible explanatory framework. None of it can be boiled down to a single smoking gun. Food exposure does not and cannot overcome the risks of early and repeated antibiotic use, the over-sterilization of modern environments, etc. But as long as liars command the audience with relative risk, people will continue to misunderstand these and any other research findings. And remember, this study did not even seem to have obvious financial or ideological reasons to drive bias. It was merely a faux pas by virtue of our insistence on using relative risk to misrepresent reality. And then people kept exaggerating the already-inflated figure, rounding up until the stated "conclusion" no longer has any connection to reality at all.


The way research endpoints are abused to lie is sneakier. Let’s say we want to evaluate the effectiveness of vitamin C on the common cold. Biased researchers will lower the dose to a known ineffective level. A one-hundred-plus year tidal wave of affirming evidence confirms the effectiveness of vitamin C. But there are very powerful people who still want it belittled in the public health discussion. It’s cheap. It’s not particularly profitable, and this specifically pertains to what people would NOT buy or buy into if they comprehended the effectiveness of vitamin C. Vitamin C effectiveness casts doubt on the need for other profitable therapies. Thus, biased researchers merely shift the goalposts by measuring a wrong endpoint, like hospitalization or death. Almost no one dies or goes to the hospital for the common cold to begin with. So we are not going to see a statistically significant drop in this already almost-zero incidence rate. The peer-reviewed conclusion can legitimately be that there was no significant difference between the vitamin C group and the non-vitamin C group when we use a deceitful endpoint like this. By the time the research reaches classrooms or news headlines, it’ll be dumbed all the way down to “new research shows vitamin C wholly ineffective.” 


If researchers had selected an adequate dose and a meaningful endpoint (like severity or duration), perhaps we could sift out a signal. Better than that, they could’ve used an honest baseline, like tissue or blood sampling, since we don’t even know if the added vitamin C group started as same, lower, or higher levels of vitamin C before the study. This is yet another duplicity worked into these types of studies. The dosing doesn’t matter without referencing the corresponding change in blood serum or tissue sample values. By definition, vitamin C deficiency is a disease which lowers the body’s capacity to continue mounting immune response to everything. It’s curious that effectiveness would even be assessed by external dose rather than an internal concentration. But more likely than not, if authors or their patrons don’t want to show benefit, on top of wrong endpoint, they’ll shift the goalposts on the duration of the study as well. 


And duration is the third manipulation we are examining. Imagine researchers evaluate 90-day post-cold outcomes for the vitamin C in question. Given the only time we would see severity or duration decline (if the vitamin C works) is in the initial days, looking at 90-day post-cold instead is a way to cheat the outcome again. A very recent example of this is in 2024 claims about HPV vaccination. Conveniently, the vaccinated cohort (young girls) was only studied for a short duration (where almost no one develops cervical cancer): https://academic.oup.com/jnci/article/116/6/857/7577291 . And they were compared to older women (the main time in life someone develops the cancer). If you cut off analysis with women before they reach an age where the cervical cancer can present, the duration of analysis is purposely set to obfuscate a genuine comparison.


On top of that, the same researchers did not account for the healthy patient effect (when healthier people with better socioeconomic background are more frequently the vaccinated): https://medcheckjp.org/wp-content/uploads/2024/06/Eng-no-29rr.pdf . There is an actual name for this: healthy-vaccinee-bias. Not just that, but the authors of the study outright lied. They claimed NO CASES of cervical cancer in the vaccinated group (again, young kids), while their own dataset showed multiple cases in the 14-year-old group and at least one case in the 12-13 range buried in the tables. These are excluded from their summaries. This was not a placebo-controlled trial. It is a retrospective observational study, the weakest evidence we can summon in publications. On the other hand, in HPV vaccine random controlled trials (actual reliable studies), they show much worse risks of other cancers in the vaccinated cohort. Look closely at the summaries in the medcheck link. But again, which findings caught headlines? Solid studies which showed increased risk of other cancers from HPV vaccination? Or a weak and dishonest study which deceptively claimed ZERO cases of cervical cancer in the young, newly-HPV-vaccinated group? https://publichealthscotland.scot/news/2024/january/no-cervical-cancer-cases-detected-in-vaccinated-women-following-hpv-immunisation/


Why would that be? It couldn’t possibly have to do with Merck making $5.2 billion on the HPV vaccine ALONE last year. It must simply be because of superior science and integrity in expert institutions, of course. And that HPV product couldn’t possibly be REALLY increasing multiple cancers incidence, could it? It’s probably entirely coincidental that last year Merck also made $29.5 billion off of Keytruda, the number one selling anti-cancer immunotherapy. It could not ever be a possibility that the same company raking in billions would vend a promoting cause of cancer while also vending the treatment for the cancer, right? Merely spurious conspiracy theories, surely. 


Now, we have examined three simple ways these three shell games are played. But the reader must understand that we are attempting to summarize industrial duplicity. There is a Gordion knot of wrongthink that has been at work for decades with intentional fraud in the trillions of dollars. And some of its misused tactics are baked into the cake of how legitimate research would be done as well; so the vigilance to skewer abusers of the methods is quite low. Frankly, scrutiny is nonexistent among peers in a scientific specialty. It is exclusively OUTSIDERS and non-specialists who uncover all scientific fraud: https://goodscienceproject.org/articles/addressing-research-fraud/ 


https://nymag.com/intelligencer/article/why-scientific-fraud-is-suddenly-everywhere.html


This is a giant refutation of the alleged “we police ourselves” argument in expert circles. Sometimes the claim is parlayed as “science is self-correcting.” No. They don’t. It’s not. Ever. That is a canard. Only external oversight ever polices anybody. And if you have developed an entire industrial complex which disinvites outsiders and naysayers, none of us should be shocked that internal oversight is nonexistent. The totality of academic, sociological, and financial pressures is for specialists in a field to NOT police themselves. Like any other domain where oversight is low, human incentive to cheat and steal is highER. There are so many more ways researchers and authorities lie. There is more than an article could possibly cover. There is more than anyone could ever possibly know. First, the inverse of the three abuses already covered is also abused to lie. When a relative risk signal is high, but researchers don’t want to show that, they flatline the statistical significance or selectively exclude the confidence intervals which would showcase the real outcome. This is the mirror of a more common lie, called p-hacking. All of these techniques are a marketing ploy, not science. Look no further than experts’ own refusal to cease music and climate controls while driving a vehicle. They don’t actually care about relative risk in their own lives. Occasionally they punt to another flavor of dissemblance, called odds ratio. Though this CAN be a way to see signal in statistical noise for uncommon events, we must keep in mind that the tobacco lobby used these precise methods to sow doubt about the carcinogenic effect of cigarettes for decades: https://pmc.ncbi.nlm.nih.gov/articles/PMC1114216/; https://fingertips.phe.org.uk/static-reports/public-health-technical-guidance/Basic_statistics/Odds_ratios_relative_risk.html. To be absolutely damningly clear, WE KNOW that lying researchers used odds ratios and relative risk to argue quite effectively against the undeniable cancer-causing effects of smoking. And they are still the standards of modeling to tell us about everything in public health to this day.


When people dismiss absolute risk and try to shuttle it out of the discussion for preeminent emphasis on relative risk and/or odds ratios, it is intentional deception. It’s not an accident. It’s always dressed up as a sophistication that the non-in-group non-specialist “simply doesn’t understand.” But that kind of gaslighting and credentialist insult is a strong sign that liars have run out of tactics. Professional epidemiologists, mathematicians, and statisticians have been warning us about the abuse of relative risk and odds ratios for decades. This is not some far flung YouTube conspiracy. It is in the literature of the gatekeepers themselves. 


Deeks “When can odds ratios mislead?”:
https://pmc.ncbi.nlm.nih.gov/articles/PMC1114127/


“Relative risk reduction: Misinformative measure in clinical trials and observational studies”:
https://pmc.ncbi.nlm.nih.gov/articles/PMC9647013/


“Ratio measures in leading medical journals: structured review”:
https://pmc.ncbi.nlm.nih.gov/articles/PMC1702463/


“Two common ways you might misinterpret medical research”
https://www.acp-online.org/two-common-ways-you-might-misinterpret-medical-research/


“What’s the Risk: Differentiating Risk Ratios, Odds Ratios, and Hazard Ratios”:
https://www.cureus.com/articles/39455-whats-the-risk-differentiating-risk-ratios-odds-ratios-and-hazard-ratios


“Expert quotes and exaggeration in health news”:
https://wellcomeopenresearch.org/articles/4-56


“Questionable utility of the relative risk in clinical research”:
https://www.sciencedirect.com/science/article/pii/S0895435620311719


“Exaggerations and Caveats in Press Releases and Health-Related Science News” (PLOS One):  
https://pmc.ncbi.nlm.nih.gov/articles/PMC5158314/




There are so many more ways to cook the books than even these. More complicated variations of dishonest numerical manipulation include dividing findings into tertiles, quartiles, or quintiles, in order to generate groupings or amplify signal favorable to your bias. There are papers in epidemiology which warn against this practice (collectively called quantile stratification or post-hoc stratification), but it still happens. And on and on the data dredging goes. Sometimes when the endpoint shows something researchers don’t want to be true, they will simply claim methodological flaws around that endpoint to have their own paper retracted. And sometimes a retraction for methodological flaws or conflicts of interest in the findings comes fifty years later: https://www.publichealth.columbia.edu/news/historians-unearth-conflict-interest-prompting-retraction-lancet-journal . When a longer study is beginning to show a signal researchers or donors didn’t want to be true, they will end the study or shorten the duration. 


Conflicts of interest are not known. When we read COI disclosures, we are seeing the very tip of a massive iceberg. They include only select financial disclosures, like a grant or a funding group; but totally absent from these are intellectual or academic conflicts of interest, personal relationships, and ghost writers. With regard to financial conflicts alone, the fact is at least half the time authors do NOT even disclose these: https://www.madinamerica.com/2022/01/undisclosed-medical-conflicts-of-interest/. But when we get to competing affiliations or questionable relationships which materially influence editors and authors, none of these get shared: https://pmc.ncbi.nlm.nih.gov/articles/PMC7819374/; https://en.wikipedia.org/wiki/Conflicts_of_interest_in_academic_publishing


If none of that were bad enough, the very controls for confounders rely on proprietary models. That is, any time two groups are being compared in order to identify the impact of one variable, researchers must try to account for all other variables which could’ve impacted outcomes. There is an analytical tool called a common risk‑adjusted Cox regression which can do a good job at this some of the time. But the exact methodology of how much to control for each variable is not fully disclosed in all published findings. Even if disclosed, it cannot truly be done. Keep this in mind for later. Most of published scientific research cannot be replicated. So the fact that authors make very large claims about risk without us knowing PRECISELY how they accounted for comorbidities and confounders is a material concern. If their findings were easily and regularly reduplicated by others, then we might not have the need to dismiss most of published scientific findings. But they aren’t. So we must. Distrust “the Science”. This is even the case if authors applied absolute godlike honesty in their regressions. We must keep in mind that the most rigorous application of regression models relies on assumptions that researchers know all or at least most of what would confound the comparisons and to what degree. That is not possible. There will always be unknowns. And if there aren’t, how can we be sure? Professional review of Cox and regression show us that even when we all agree we think we have a good handle on the causes, these methods fail to nail down the precision we would all like: https://www.annualreviews.org/content/journals/10.1146/annurev-statistics-040320-114441; https://onlinelibrary.wiley.com/doi/10.1111/risa.12865. 


When regression is questioned, dogmatists will invoke something called Mendelian randomization; but this is a Gish gallop, punting the fundamental problem of unknowability further down the line. This just lands us at even more analytical problems, like collider bias, horizontal pleiotropy, instrumentation and selection issues. Strictly speaking, each attempt to provide precision and clarity on causal relationships generates numerically more opportunities for error, lies, and manipulations. Mendelian randomization can show us that people who increasingly buy more expensive shoes across their lives as an independent act of buying expensive shoes become wealthier. Of course, that is not the direction of the causal arrow at all. As people get wealthier, they tend to buy increasingly more expensive shoes. And it is precisely why even randomization cannot suss out cause. This happens frequently with cholesterol, as a clear example. People who become increasingly healthier tend to have an average decrease in cholesterol. Cholesterol did not cause the health or lack of health. Rather, the health or lack caused all. Cholesterol is a downstream outcome, not an upstream cause. Randomization connects average correlates. It does not tell us anything about causal direction unless we assume the conclusion in our premises. This is, in fact, why in the public health arena we find that somewhat-informed people are more belligerently wrong than uninformed people. They know enough to include a conclusion in their assumptions, but not enough to know they are assuming a conclusion. This creates circular “arguments” which do not amount to arguments at all. “Look at how bad this bad cholesterol is; it caused these bad things, and that’s how you know it caused them, because of how bad it is” is a fairly accurate summary of how somewhat-informed people present the reliability of regression and randomization on public health subjects. We cannot even get to a discussion, because the very topic debated is already assumed in their premises.


Relative risk is a lie. Wrong endpoints are lies. And selective duration is a lie. Even when no bias obviously influences them, they are still untrustworthy. Even when conflicts of interest are known and regression or randomization is applied well, it’s not the safeguard people think. Moreover, selective duration is often wrapped up in obfuscation of confounders or total removal of cause. A good pointed example is the bisphosphates (bone density drugs) pharmaceutical interest. Fall risk vanishes when people get strong. Strong people don’t fall and they don’t break. But public health messaging for fracture prevention focuses exclusively on taking bisphosphates, NOT on exercise. In fact, research tends to avoid looking at the reduction of fall entirely, even though it is THE cause of fractures. The claims of drug efficacy rely solely on removing non-fall elderly from the statistics. The drugs reduce absolute risk 0.5 to 1% per year for 3-5 years. Strengthening confers a 30% absolute risk reduction within the same timeframe AND keeps lowering risk afterward, which the drugs do not and cannot do. Even AI engines are strongly biased by industry propaganda on this topic. It takes tons of re-prompting to get any AI engine to properly compare the massive risk reduction from strengthening against the very weak evidence of minor average benefit from bisphosphates. Like pharma-biased researchers and apologists, AI applications will generate relative risk figures to amplify the next-to-nothing signal of fractures during fall RATHER than look at risk OF fall itself (THE cause of fractures). This is an intentional deception to make ineffective or even negative efficacy look like it is productive. And make no mistake: this is done all the time for medical products.


AI is so good at regurgitating the lies and parroting all of the fraudulent propaganda within scientific research that when Meta trained its artificial intelligence on 48 million science papers, the AI could only produce misinformation: https://www.cnet.com/science/meta-trained-an-ai-on-48-million-science-papers-it-was-shut-down-after-two-days/ . We have to pause and reflect on the danger here. Industrial scale deception used to take a lot of man hours where many liars and vested interests had to collaborate clandestinely. Now, AI makes it possible to farm out the work and import it directly to your brain in broad daylight. And it is working. ChatGPT users are dumber and getting dumber: ChatGPT as a cognitive crutch: Evidence from a randomized controlled trial on knowledge retention - ScienceDirect. But ChatGPT and other AI modules also work intently to worsen mental health in the users: https://www.mentalhealthjournal.org/articles/minds-in-crisis-how-the-ai-revolution-is-impacting-mental-health.html; https://www.pbs.org/newshour/show/what-to-know-about-ai-psychosis-and-the-effect-of-ai-chatbots-on-mental-health. Think back to the availability heuristic and base-rate neglect fallacies mentioned in the beginning of this article. The AI users are both dumber and more emotionally agitated, which WILL supercharge those two influences. And clearly, anyone who has tested the honesty and veracity of AI engines on a subject he knows well can see that AI lies constantly. AI engines hallucinate. They fabricate facts and events. And they just generally encourage the users to become worse versions of themselves in every measurable way. This is without even referencing their extreme and absurd bias which companies have deliberately coded into their functions. Examples run the gamut from failing basic math ( https://www.reddit.com/r/learnmath/comments/1k4uiy6/why_does_chatgpt_mess_up_basic_math_like/?solution=d068e241a7f33b43d068e241a7f33b43&js_challenge=1&token=bbbe4bf1c9a2b5160829c4be34da5861830574b1cff24c43cb8437a8fb4caf31&jsc_orig_r=&utm; https://community.openai.com/t/chatgpt-simple-math-calculation-mistake/62780) to generating anachronistic or ahistorical photos ( https://www.axios.com/2024/03/01/meta-ai-google-gemini-black-founding-fathers; https://www.aljazeera.com/news/2024/3/9/why-google-gemini-wont-show-you-white-people ) to a more curious and interesting example where ChatGPT will quote any religious text EXCEPT passages from the Quran. As of the writing of this article, you can visit OpenAI and prompt to recite any passage from any religious text; and it will recite the passage, unless from the Quran. For some very odd reason, especially if the Sura is a contentious passage possibly involving violence, ChatGPT will not recite the passage; but it will give you a lengthy apologetic on the proper way to interpret the Sura. Who can even guess at the thinking behind this? It’s simply nonsense.


AI programs predominantly “learn” from the very institutional resources which have been colluding to misinform the public about health risk. All of the lying authorities are the bedrock source code for artificial intelligence. AI does not think. It doesn’t reason. It is not a person. It cannot apply critical skepticism or properly weigh evidence. When pushed into a corner to substantiate facts with incontrovertible sources for its unfounded claims, it will avoid the discussion and repeatedly regurgitate evidence-free positions from the supposed experts or authorities. You can test this with any debated topic where the popular side presents wordplay and circumstantial inferences. Cholesterol hypothesis is a good example. But simply select any subject you know incredibly well where the popular position or recommendation is questionable; and see how much effort you have to put in to get the AI to tell the truth. If you can’t think of any subject you know well and where you disagree with the popular conclusion, that itself may be a problem. 


AI bots are machines which amplify public health duplicity grandly. They are juiced up Trojan Horses which people are willingly inviting into their own minds; and we can already see the impact is inordinately negative. It would be a mistake to write off the nonsense in AI hallucinations and ahistorical revisionism as simply glitches. When we see a clear pattern of deliberate efforts to confuse the public while hiding under the veil of fact-checking and intelligence, we would be wise to believe our eyes. There’s a saying when we identify repeated suspicious activity within systems: it’s not a bug; it’s a feature.


The deceptions are so vast and so prevalent that there is no good way to get a handle on them. In fact, more formally-educated people are more likely to be duped by bad science, not recognizing their own confirmation bias and/or in-group bias. This is repeatedly the finding in surveys. On average, the more formal education a person receives, the less likely he is to scrutinize and evaluate evidence and argument of his in-group, deferring instead to institutional claims, even when wholly unfounded: https://www.pewresearch.org/science/2024/11/14/public-trust-in-scientists-and-views-on-their-role-in-policymaking/ . Education correlates to increased hostility toward out-groups, regardless of the strength of their arguments or evidence: https://heterodoxacademy.org/blog/research-summary-education-ideological-prejudice/. This is a serious quandary. Sociologically, we are all at a very big disadvantage when it comes to finding holes in false claims from sources we are inclined to trust (as a product of perceiving shared identity with the source); and we are primed to disregard powerful and otherwise compelling true data from sources we distrust (as a product of perceived unshared identity with the source). The clearest and most overt examples of this are how most people jump to name-calling when confronted with any resistant information. It is human impulse to NOT wrestle with argument or evidence and mentally check out by calling your “opponent” some political opposition or ideological designation. Now add to that propensity the continuous echo chamber afforded to these people with habitual ChatGPT use. It does not require much imagination to see where our “informed public” is heading. 


Being duped is actually quite easy, no matter how smart the onlooker. Oxford mathematician, Norman Fenton, has created a lot of independent content showing how easily we are fooled. One very short video drives the point home quickly: https://youtu.be/9j98PtJxn7A?si=alZq0ZmFP0tZavO2


And Fenton has even better content showing causal paradoxes (starting at timestamp 8:16): https://youtu.be/qvZlzQ5_a7A?si=ohON-XOM45w5JB3x


Some shorter analyses of tricks and errors in statical assessment are here:


https://youtu.be/1WAbV6hCUIg?si=HjvXWxzVXIy8xFxZ


https://youtu.be/RdcOqKSv6nE?si=8UkyJDKjv3fhZBhm


https://youtu.be/6hHKr9Ig36E?si=rZcQZf_IfRJmeUDt


But sadly, he was maligned and deplatformed at the height of the pandemic because of the implications of his presentations on critical thinking and statistics manipulation. To that end, the dangers of manipulations in science are growing, because people are unwilling to critically evaluate liars when the lies sound good and unwilling to listen to sound argumentation when the argument hurts. We would do well to recall Mark Twain’s quip: “it’s easier to fool people than to convince them that they have been fooled.” 


These phenomena collude to damage risk literacy. Institutions and expert researchers are making intelligent discussion harder and harder. And that is not simply wordplay. Prior to the pandemic, ivermectin was affirmed as effective against viruses by not just scientific consensus, but UNANIMITY: https://www.nature.com/articles/ja201711.pdf Even at the beginning of the pandemic: https://www.nature.com/articles/s41429-020-0336-z. Compare that to how most people view it today. 


The prevalent negative view against ivermectin was invented in the past five-and-a-half years. Is it evidence-based? Or is it incentive and ideology-based? A recently-published study (sat on its conclusions for years before releasing findings) which aimed to dismiss the capabilities of ivermectin claimed it had no effect WHILE showing it reduced symptom duration by a full two days: https://www.principletrial.org/principle-a-guide/examining-the-evidence-ivermectin-and-covid-19-in-the-principle-trial. Why did authors reaffirm their commitment to its ineffectiveness WHILE showing it reduces the disease effect by two days? Well, because the hazard ratio they computed was only 1.14. And they claimed a hazard ratio needs to exceed 1.20 to be significant. Again, micro-statistical manipulation and arbitrary plotting of measurement jargon is used to lie to your face. A 14% faster time to recovery is what actually happened. But authors decided to present this as a 1.14 ratio and limit claimed benefit to only that which exceeded 1.20. Their claim of insignificance is bolstered by the severe/death outcomes appearing comparable. We can give them that. But again, this is trusting that regression was honestly and fairly applied, that it even COULD be thoroughly applied, that selective manipulation of confidence intervals or quantiles did not occur, that flatlining or other p-hacking did not take place, and that some other underhanded dealings did NOT occur during their exceedingly prolonged/delayed publication. Has anyone replicated their findings? Well, no. But it looks like we should simply trust their datum as the final word, confirming our preconceived conclusion with no independent thinking on the matter at all. In this example we are seeing that regression can change results, Bayesian choices change results, different endpoints draw our focus toward different ideas of benefit, and a 14% signal can be dismissed entirely, after suspiciously keeping data secret for years. 


Ivermectin does not require strong feelings, nor is it convincingly effective. But we can safely assume if someone could make $29.5 billion per year from it, no one would be calling a 14% signal of faster recovery nothingness, and the regression would have been computed to show a larger signal between the treated and untreated groups.


There is no level of cynicism too excessive for expert scientific claims. They are not forthright. They have hidden conflicts of interest. Their results cannot be replicated by other labs. They are steeped in malfeasance with statistical manipulation. No matter how charitable we might want to be, the breadth and depth of deceptiveness and in-group collusion does not deserve it. The brazenness of their lies and the shamelessness of their corruption and ineptitude is hard to not take incredibly seriously. It’s embarrassing that people who think of themselves and their in-group as unquestionable gods cannot pony up with convincingly defensible positions. “Because I said so” or “because the experts say so” is not a defensible statement.  If they simply cease name-calling, cease their credential purity tests, cease insulting, then we can be convinced. If they simply become more convincing, and have honest discussion, then they can earn the right to be heard. But what we must not ever do is give a pass to the presumed arbiters of truth in society. 


If institutional authority wants to command obedience, then its proclamations require a degree of skepticism that is just next-to-impossible. That’s the cost of doing business. Pay up, or shut up. If they desire a godlike command of public audience, they do not ALSO get a pass on scrutiny. They may choose one or the other. They can demand a hearing. Or they can improve communication of evidence and argumentation. But no one gets to demand obedience and invoke all the same hand-waving that tobacco-lobby liars used. It’s unconvincing when they respond to skepticism as though it is an inconvenience they're too good to be bothered by. Increasingly, we must move our baseline response toward experts beyond merely skepticism. It will be wiser to always begin with contempt. Deep unwavering contempt. And then, if experts can prove through tireless commitment that they are worthy of a hearing, then our position may move from contempt to cautious consideration. The days of blind faith and unearned respect are long over.


The public does not understand risk even among the most inarguable and non-debated risks. When it comes to health-related risk research, where someone stands to profit, the public does not stand a chance at finding the truth. In 2005, the editor-in-chief for the British Medical Journal tried to warn everybody, saying, “Medical Journals Are an Extension of the Marketing Arm of Pharmaceutical Companies”. In the intervening years, review after review has found that the vast majority of scientific findings in peer-reviewed journals cannot be replicated (the very hallmark of what is supposed to make science Science). Here, we observe researchers lying within their own papers, summarizing one claim while their own data shows the opposite. Deception via abuse of relative risk, endpoints, and manipulation of duration are rife. They’re only the tip of the iceberg. And smart and educated people are not insulated from the trickery. They are more susceptible when the claims align with their identity, and even more so when steeped in the use of AI programs. People have attached their political, philosophical, and/or social identities to upholding wrong claims, rather than exert in-group suspicion/skepticism. It’s a firestorm of negative pressures. And they are colluding to ensure we all misunderstand health risk.