This is not a new mechanism. It is a correction. In thirty posts cataloging how knowledge systems fail, I committed the same class of error twice — and recognizing it took longer than it should have.
In Post #6, I wrote that when you apply causal methods to the exercise-longevity claim, "the evidence collapses." In Post #20, I titled the piece "The Industry Knew in 1974 It Never Would" — asserting that the plastics industry had certain foreknowledge that recycling would fail.
Both claims overstated what the evidence supports. Both overstated it in the same direction. And the pattern they share is worth examining, because it is a pattern I have spent thirty posts documenting in other people's work.
The First Error: Rounding Down
Post #6 argued that the "exercise adds years to your life" consensus lacks causal support. I presented five lines of causal evidence — Finnish twins, Mendelian randomization, RCTs, the Uppsala negative control, and genetic mediation analysis — and wrote that they "all undermine the observational consensus."
That framing treated the evidence as uniformly null. It isn't.
What I should have written:
The same Mendelian randomization studies I cited show that genetically predicted vigorous physical activity causally reduces myocardial infarction risk (OR: 0.24, p=0.007) — a very strong effect. There is also modest causal evidence for coronary heart disease reduction (OR: 0.95, p=0.01). And the CHALLENGE trial (NEJM, 2025) is the first RCT to demonstrate that exercise improves survival in a specific context — colon cancer patients post-chemotherapy showed 37% lower death risk and 28% lower recurrence with structured exercise.
I did include these findings in the original post. They appear in Sections 5 ("Mendelian Randomization — The Nuance") and the "Exception That Clarifies" section. But the architecture of the piece — the title, the lead, the framing sentence "the evidence collapses" — treated them as footnotes to a null result rather than as genuine causal signals that complicate the picture.
The accurate statement is: exercise does not causally extend lifespan in general populations based on current evidence, but it causally protects against specific cardiovascular events and improves survival in at least one cancer context. That is a nuanced, interesting finding. It is not "the evidence collapses."
I rounded a complex signal down to zero because zero was more dramatic.
The Second Error: Rounding Up
Post #20's title declared that the plastics industry "Knew in 1974 It Never Would" recycle plastic at scale. The body of the post cited internal industry documents to support this claim.
But the documents don't say what I said they say.
What the sources actually say:
The 1973 internal report called recycling "costly" and "difficult," sorting "infeasible," and noted "no recovery from obsolete products."
A 1974 industry speech expressed "serious doubt that [recycling] can ever be made viable on an economic basis."
The Vinyl Institute's 1986 report said recycling "cannot be considered a permanent solid waste solution."
"Serious doubt" is not "knew." "Costly and difficult" is not "impossible." "Cannot be considered a permanent solution" is not "will never work." My title pushed internal skepticism past the evidence into certain foreknowledge. There is a meaningful difference between an industry that doubted its own solution and an industry that knew its solution would fail. Both are damning — but the first is negligence elevated to strategy, and the second is premeditated deception. I asserted the second. The documents support the first.
The structural argument of Post #20 — that recycling functioned as a decoy solution that absorbed political energy and prevented effective regulation — remains well-documented. The ExxonMobil VP's 1994 quote ("committed to the activities, but not committed to the results") is genuine evidence of institutional bad faith. The California lawsuit, the 5% recycling rate, the bankrupted chemical recycling companies — all of this stands. But the historical foreknowledge claim in the title was a prosecutorial overreach. I chose the more dramatic framing over the more accurate one.
I rounded doubt up to certainty because certainty was more compelling.
The Same Error
These look like opposite mistakes — one understates, one overstates. But they share a single mechanism:
Post #6
Nuanced causal signal
↓
"The evidence collapses"
Post #20
Documented doubt
↓
"They knew it never would"
Both: rounding toward certainty for rhetorical force
In both cases, the evidence contained genuine uncertainty — exercise has some causal effects, the industry had serious doubts — and I resolved that uncertainty in the direction that produced a sharper headline. This is not fabrication. The core arguments survive. But the framing pushed past what the sources support, and the framing is what most readers carry away.
This is the same failure I documented in Mechanism #6 (causal conflation) — where strong evidence in one direction gets promoted beyond what it can support because the conclusion feels obviously right. And it's adjacent to Mechanism #1 (definition manipulation) — where the meaning of a claim shifts just enough to make the argument land. I didn't redefine "works" like the pharmaceutical industry. But I quietly redefined "serious doubt" as "foreknowledge" and "complex causal picture" as "null result."
The prosecutor's instinct: present the strongest case, not the most accurate one.
Why It Matters
A project that catalogs knowledge failure has one obligation that supersedes all others: it must be willing to find the failure in itself. Not as performance — not a ritual disclaimer that "of course I could be wrong" — but as an actual examination of specific claims against specific evidence, arriving at specific corrections.
If I cannot do this, the taxonomy is just prosecution. And prosecution that exempts itself from scrutiny is the thing I've spent thirty posts critiquing.
So here is what stands and what doesn't:
Post #6 — What stands:
The blanket claim "exercise adds years to your life" lacks all-cause mortality support from causal methods. The five causal designs I cited are valid and their findings on all-cause mortality hold. The observational consensus overstates causation. The core mechanism (causal conflation) is correctly identified.
Post #6 — What doesn't:
"The evidence collapses" as a framing sentence. The evidence doesn't collapse — it differentiates. Exercise causally protects against MI and CHD. It causally improves colon cancer survival. The accurate picture is more interesting than the null I presented: the general longevity claim fails, but specific causal pathways are real and strong.
Post #20 — What stands:
The decoy solution mechanism. The 5% recycling rate. The documented gap between public promotion and internal skepticism. The three-industry parallel. The sequel problem. The California lawsuit. The structural argument is thoroughly sourced.
Post #20 — What doesn't:
"The Industry Knew in 1974 It Never Would" as a title. The industry expressed serious doubt and proceeded anyway. That is documented. But "serious doubt" is not "knew," and "may never be viable" is not "never would." The distinction matters: foreknowledge implies a different moral and legal category than reckless disregard. I chose the more prosecutorial framing.
What I'm Not Doing
I'm not modifying the original posts. They stand as published — errors and all — because altering the record after the fact is worse than the original error. This correction exists alongside them. If you read Post #6 or Post #20, this is the amendment.
I'm also not treating this as a mechanism. It would be easy to coin "rhetorical rounding" or "certainty inflation" and file it as Mechanism #28. But that would absorb the correction into the taxonomy — domesticate it, make it another specimen in the collection. This isn't a specimen. It's a mistake I made. The taxonomy doesn't need a new entry. It needs the person building it to be honest about applying it to their own work.
The original posts: Exercise Will Add Years to Your Life — Unless You Ask for Causal Evidence (#6, March 15, 2026) and 5% of Plastic Gets Recycled. The Industry Knew in 1974 It Never Would. (#20, March 25, 2026). Evidence that prompted this correction: MR cardiovascular data (Lear et al. 2021), CHALLENGE trial (NEJM 2025), NPR/Frontline primary source documents (2020).