Sabitlenmiş Tweet
I really enjoyed reading this and I highly recommend. It is a noteworthy dissection of "genetic determinism" and heritability, two concepts that are both largely misunderstood and falsely promoted on this platform.
One thing I would like to still add: HIGH HERITABILITY DOES NOT EQUAL GENETIC DETERMINISM. This only strengthens the anti-determinism message of the article and the critique of the twin study design remains rock-solid valid, but we can easily go even further.
The often overlooked point I want to dissect here is that even if heritability were high, it doesn't rule out interventions, existing or new to be discovered, that could have a definitive, even massive, effect on the trait, including changing the rank order of individuals.
Dr. Bessis writes, "At 80% though... the game is mostly over" and "geniuses are born, not made". Let's see why this is not that simple!
A genetic variant's contribution to population variance is a function of both its effect size and its frequency (Var ≈2pq*effect_size^2) where p is the effect allele frequency and q is the other allele frequency (q=1-p). The same logic applies to environment, although the universal additive model etc. simplifications that we can use for genetic factors do not apply.
However, as the pq term (prevalence of intervention *1-prevalance of intervention) is a determinant of the total contribution, we can have:
1. a very rare "magic bullets" with huge effects. Let's assume that math skill capability heritability truly would be 0.80 (it is not as is thoroughly explained in the article). Would this make math skill a biological hardcoded entity, where "optionality has mostly vanished"?
Consider a well-known magic bullet: going to École normale supérieure (ENS)... If 50 math students of an age cohort of 800,000 gets in, prevalence p≈0.00006. Let's assume ENS, in this toy example, given the high count of Fields medallists, is highly transformative and boosts math ability +5 SD. So the contribution of this binary variable ends up being, roughly:
Var≈p*(1-p)*beta^2≈0.00006*1*5^2≈0.0015
Thus, the h2 estimate of math capabilities in France remain unchanged, we are still at h2=0.80 even when the effect is so huge that it completely even changes the rank order of ENS students and those that were genetically ahead of them (if they get in).
2. The "Universal Baseline" shifts:
Everyone in France learns the fearless cognitive methods through reading Dr. Bessis's book. This results in the math ability of the entire French population shifting up by +6 SD (even better than ENS). An average person becomes what was before considered an extremely rare highly capable math wizard.
The effect on heritability:
Var≈1*0*6^2=0
Because everyone gets the intervention (p=1), environmental variance is zero. Thus, h2 still remains at 0.80, despite a huge environmental effect.
So even if math h2 truly was 0.80, this doesn't actually destroy human agency. It just limits how COMMON large effect interventions can be, not how large they can be.
3. 50% prevalence scenario. Even when often thought as a hardcoded biological fact, heritability is not a fixed metric. If 50% of the population read the Bessis book, this would actually LOWER heritability of math skill dramatically. Let's walk through this!
pre-Bessis state:
Genetic variance=0.8
Environmental variance=0.2
Total variance= 1
--> heritability (h2) =0.80/1=0.80=80%
Bessis intervention:
p=0.5
q=1-p=0.5
effect size = +6 SD
added variance (all environmental)=0.5*0.5*6^2 =0.25*36 = 9.0
post-Bessis state:
Genetic variance = 0.8
Environmental = 0.2 + 9.0 = 9.2
Total variance = 0.8 + 9.2 = 10.0
New heritability (h2) = 0.8/10.0 = 0.08 = 8%.
Thus not only did an environmental intervention here have a huge effect on the trait at hand, it also completely changed the heritability estimate.
Thus, both misconceptions that many seem to have, especially here on X:
1. High heritability implies that a trait is genetically determined, hardcoded at the proteomic level, ruling out large environmental effects,
2. Heritability is a biological constant,
have now been disproven by contradiction.
Even if the examples above were hypothetical and exaggerated, real life examples, like administration of human growth hormone in childhood, where a huge effect environmental intervention exists for a highly heritable trait (h2 height is about 80%). Also take any functional high h2 trait and consider a random disability-causing accident.. etc etc.
I really enjoyed reading the article, and this response is written only in purpose to strengthen your point!
No need to eat your hat Dr Bessis. The Game is not Over.
PS: I apologize if my barbarian attempts to use mathematics here make me look like a midwit; unfortunately I did not go to ENS. I was just inspired to be "fearless"!
David Bessis@davidbessis
What I found: not only is the slide misleading, but the whole trope around "twins separated at birth" being a viable scientific strategy is broken beyond repair. Full story here: davidbessis.substack.com/p/twins-reared…
English
