Comment by gwern | Hacker News Reader

Comment by gwern | Hacker News Reader

Parent Full thread

Yes, there's an interesting philosophical question here about burdens of proof and Humean 'extraordinary claims require extraordinary evidence'. Calment's claim is to be the oldest person out of billions of people, and she is the oldest by a truly extraordinary margin - no one has even come close to reaching her record before or since. So, which is more likely: that her claim however improbable (someone has to win the lottery) is completely correct (and she simply had some unique luck or mutation enabling her excess longevity), or that some highly unusual (but still more likely than '1 in tens of billions') circumstance (such as a complicated tax & insurance fraud) has led to a mistake? Indeed, how would one ever be sure of such a claim, short of the isotopic testing advocated in OP - which will, incidentally, stop working in the future?

(This is a general problem of priors and evidence and model uncertainty I think about occasionally: https://www.gwern.net/Modus https://www.gwern.net/Littlewood https://www.gwern.net/Mail-delivery#on-model-uncertainty https://www.gwern.net/Research-criticism https://www.gwern.net/Everything https://www.gwern.net/Turing-complete#on-seeing-through-and-... )

I really wouldn't say "nobody has come close" - she was 3 years older (122 vs 119) than the next oldest verified person, and around five years from most of the next top ten - most of them died at 117.

If she was, say, 132, I'd be more suspicious, but 122 sounds more reasonable for statistical outliers in an incredibly complex process.

The problem is, 3 years is a huge gap. If you look at all the other centenarians, they really do cluster within months of 117-119. And then you have Calment, who is the only outlier and who is like 30+ months past them all (and this is at a time of life where the annual mortality rate is like >50%). I plotted the age gap between the record holders once, and it looks basically like a flat line and then a single dot at the top of the graph corresponding to Calment. It may be a complex process, but at least for everyone else they die on schedule.

It is possible even if unlikely. But if you are right that it is way too unlikely then the next alternative is that her age was exaggerated by 3 years. The theory that she "was her own daughter" I think is also possible but less likely than an (intended or unintended) error of 3 years somewhere.

It's also worth pointing out that all the Abrahamic religions predict a 120-year limit, which prediction dates back thousands of years (Genesis 6:3). So regardless of your opinion of Abrahamic religions, there are a lot of people who would care a great deal about verifying those extra three years!

>all the Abrahamic religions predict a 120-year limit

Do you have any more information on this? It sounds like interesting reading, but all I can find on Google is a link to this post.

https://www.versebyverseministry.org/bible-answers/are-peopl...

from https://www.google.com/search?q=bible+120+year+age+limit

The other explanation is that God was saying he was going to bring the flood in 120 years.

There were exceptions after the flood. In the Book of Chronicles we have Jehoiada who died 130 years old. (2 Chr. 24:15)

According to Professor Gershwin:

Methus'lah lived nine hundred years Methus'lah lived nine hundred years But who calls dat livin' When no gal will give in To no man what's nine hundred years ?

I can't speak for other religions, but in Judaism, the 120 is used in common practice to mean "long life". It's how long Moses lived.

The interpretation of the 120 years reference is more complicated. See https://www.aish.com/atr/120-Year-Lifespan.html

Why did Moses et al. take 40 years to move from Egypt to Canaan? (about 200 miles)

https://en.wikipedia.org/wiki/The_Twelve_Spies#Consequences

Hinduism (Vedas) sets 120 years as human lifespan (provided it is lived according to dharma, transgressions reduce it a proportionally). That's where all other religions got it form.

That's very interesting if true. Can you give any reference? All I can find through search engines is some references to "about 100 years" and a single thread where someone mentions a 120 year prediction but then the rest of the thread is everyone else correcting that person.

It ought to be possible to estimate the odds. We don’t know exactly how mortality changes with extreme age but there ought to be a pretty decent idea. How likely is it for a gap like this to appear?

You can see similar progression with world record mile runs: https://en.wikipedia.org/wiki/Mile_run_world_record_progress...

Advancements in nutrition and health in combination with expectations about longevity might result in similar clustering followed by leaps.

Or not. Just saying that in the absence of a better fundamental understanding it's difficult to judge the strength of such circumstantial evidence. This recent paper should absolutely make us more skeptical of claims, but doesn't really shed much light on the heart of the matter--limits of longevity.

Are you intimating that she learned a breakthrough technique in how to be old?

The history of advancements in the mile, the marathon, and other competitions seems to suggest that expectations about what's physically possible effect people's performance individually and generally--e.g. apparent endurance barriers.

A similar phenomenon might also exist in the context of longevity. If you get sick and weak at 70 but expect to live to the average age of 90 you might be more optimistic and committed to getting well. If you're sickly at 95 but expected to have already died you might give up the ghost sooner, so to speak.

Assuming that such a phenomenon exists in the context of longevity, I'm intimating that increased awareness about extreme longevity might have the effect of changing expectations and therefore advancing at the margins the extremes of longevity that some people attain. And if the phenomenon behaves like it does in sports, one would expect clustering followed by intermittent breaks to a new limit.

EDIT: To be clear, my main point isn't that this phenomenon exists but that given the possibility of its existence, and in tandem with the clear paucity of reliable, accurate data, the clustering we see might not be strong evidence of a hard barrier at ~117, and in particular not strong evidence that 122 is wildly improbable.

Given that anyone who has reached 117 has already dismissed their ailments with an "this would be the end for most people, but I'm not most people" stubbornness for a few decades, the argument implied by the original paper might be a better explanation for clustering rather than tapering at the high end (i.e. some people knowing that extremely old people live to 117-119 are willing to maintain the fraud of their relative's survival for that long, but are unwilling to risk or unable to succeed in taking it further...)

Could just be a breakthrough genetic mutation. The information posted here suggests that she also had unusually good health when young, and appeared younger than her daughter during the time period when both were alive.

Shame she had no surviving offspring, though it seems like her daughter didn't share her longevity anyway.

That was one of the many arguments in the original blog post IIRC - that if you look, at the rest of her family, they all died in their 50s/60s, making this hypothesis less likely than if they all tended to live in their 90s or some such.

No, you don't. Look at the successive increments: they get tinier and tinier over time, exactly like one would expect from an increasingly professionalized global sport drawing from a talent pool of hundreds of millions of people. The only gaps or changes in the record which are remotely comparable to how much Calment exceeded and continues to exceed other oldest-people date back to the 1800s. Her record is in the completely wrong place.

I did some simulating of the Gompertz and it seems to support my intuition that Calment really is extreme: https://www.gwern.net/Order-statistics#sampling-gompertz-dis...

No. 122 is not reasonable.

A suggestion that there may be a straightforward determination of facts via a blood test:

https://www.connexionfrance.com/French-news/DNA-blood-test-c...

A special case, though. Almost all centenarians wouldn't admit of such a solution - at least, until we start getting whole population pedigrees and then an inability to put a centenarian in the appropriate place becomes proof of error/fraud.

> that her claim however improbable (someone has to win the lottery) is completely correct

That line of reasoning does not apply here. Someone will win the lottery given enough time as the possibility space is bound - but there is no similar given guarantee that any human would ever live to age X if we have never observed a human living to age X before.

It does apply. Someone has to be the oldest. Someone has to be the oldest-person who has the largest gap. Regardless of how old that is, or how large the gap is, there will be at least one.

Ok, now I get the point you were trying to make.

The lottery parallel still doesn't feel the most apt; the definition you specify ensures that you will always have such a person (who lived to be the oldest, and who had the largest gap with the previous record holder) for n > 1 people. In that sense it is recurrent, unlike winning the lottery, which no one has until someone does (or winning in a raffle, where we know that there will aways be a winner due to the very definition of the game).

Regardless, none of that helps us determine any further how plausible Calment's statistically anomalous record is.

Someone has to be the oldest but the gap is an independent information. Mortality rate is constant at 0.5 per year for Supercentenarians. That gap is a massive outlier.

Reply on news.ycombinator.com