25 Questions for Alexander Strudwick Young
LLM-generated interview questions for Alexander Strudwick Young on family-based GWAS, direct vs. indirect genetic effects, missing heritability, polygenic embryo screening, and data-access gatekeeping in human genetics.
In April 2026, I learned that noted human geneticist Alexander Strudwick Young (podcasts, Twitter highlights) of UCLA and Herasight, known for his work on heritability and GWASes and embryo selection was struggling with cancer at 35. He had limited time left for interviews such as his May 2026 podcast with Steve Hsu.
So I used Deep Research across GPT-5.5 Pro, Claude-4.7-opus, and Gemini-3.1-pro-preview to develop a biographical overview, and then my “interview prompt” for a comprehensive set of last-minute questions.
GPT-5.5 Pro’s consolidated biographical brief:
Alexander Young—the statistical geneticist whose papers appear both as Alexander I. Young and Alexander Strudwick Young—is one of the clearest witnesses to the biggest interpretive shift in human complex-trait genetics since the rise of GWAS. He is easy to miscast as merely an “IQ/heritability” guest, but that undersells him. His real importance is methodological: he sits at the hinge between the old population-GWAS era and the newer family-based attempt to separate direct genetic effects from parental environment, assortative mating, and population structure. (Young et al 2019)
His trajectory is unusually clean. He was trained first in mathematics and statistics, then in computational biology, completed a doctorate in genomic medicine and statistics at Oxford, worked at deCODE in Reykjavik, collaborated with Augustine Kong at Oxford’s Big Data Institute, moved through the USC/SSGAC orbit, and is now based in UCLA Human Genetics. (Young et al 2022, Bristol 2019)
The key intellectual inheritance runs from Fisherian quantitative genetics through Peter Donnelly’s statistical genetics and Kong’s use of Icelandic pedigrees. What Young seems to have taken from that lineage was not just a taste for big data, but the conviction that Mendelian segregation within families is the decisive natural experiment in human genetics. That theme is explicit in his own description of his research: using the randomization of genetic material within families to disentangle nature and nurture, and then tracing how genetic and cultural transmission interact with mating patterns to shape population differences.
His rise maps onto a sequence of papers that rewired the field’s interpretation of genotype-phenotype association. As a coauthor on Kong et al’s “The nature of nurture”, he helped show that even non-transmitted parental alleles can matter because parents transmit environments as well as DNA. In RDR, he helped build a heritability estimator that uses random variation in relatedness to strip away environmental bias. In his 2019 Science review, he argued that the real question is not merely whether associations replicate, but what mixture of direct effects, indirect effects, and structure they contain. (Kong et al 2018, Young et al 2018, Young et al 2019)
That program continued in snipar, the educational-attainment within-family work, and the newer family-GWAS papers. The point was not merely to say that standard GWAS are “confounded”, but to make family-based estimation scalable enough to compete with population GWAS in real biobanks, and to show empirically that many behavioral and social-trait associations shrink, split, or change meaning once one isolates direct effects. That is why Young matters historically: he is one of the people who turned “within-family” from a pious methodological caveat into a serious alternative research program. (Young et al 2022, Okbay et al 2022, Guan et al 2025, Tan et al 2024)
This is why he is interesting far beyond behavior-genetics gossip. He can explain, in one interview, why educational attainment became such a revealing trait for the whole field; why assortative mating can look like shared biology; why “heritability” is not one number but a family of targets; why additive models work as well as they do despite nonlinear biology; and why the same forces that make population GWAS causally impure can also help make them predictively useful. (Okbay et al 2022, Border et al 2022, Benjamin et al 2024, Young et al 2018)
Young is also unusual in being simultaneously insider, critic, and builder. He sits inside the SSGAC/social-science-genomics world whose summary statistics transformed the field, has spent years pointing out where its headline estimates overreach, and is now an advisor to Herasight, where the question becomes what cleaned-up within-family signal can honestly support in IVF and embryo screening. That combination makes him much more interesting than either a pure academic methodologist or a pure startup founder. (Li et al 2025)
He also appears to be moving, at least in part, from pure methodology toward the governance of science itself. In a recent unpublished draft on NIH-controlled data access, he argues that committees can quietly decide which genetics questions are even licit to ask, especially around education and intelligence. That suggests an interviewer would not only get a tour of modern statistical genetics, but also a first-person account of how controversial science is administratively governed. (Young 2026, unpublished draft)
For an interviewer, then, the attraction is not simply that Young has strong views. It is that he can narrate a whole scientific transition from the inside: candidate genes to GWAS, GWAS to family-based designs, direct effects to indirect effects, academic method papers to reproductive technology, and open inquiry to bureaucratic gatekeeping. The papers will remain, but a long synthetic first-person account of how this shift looked from inside the field still does not really exist; what exists publicly are a few podcast appearances, scattered blog posts, and the papers themselves. (CSPI 2022)
Then I generated several sets of questions, and had GPT-5.5 Pro curate them into a final set of AI questions:
Origins and formation:
When you look back, what was the original puzzle that made human genetics feel worth a life’s work?
You came in through mathematics and statistics. Which habits from that training helped you most, and which had to be unlearned once you were dealing with biological and social traits?
How did working first with Icelandic pedigrees and then with large, admixed American biobanks reshape your sense of what family structure and population structure can reveal?
Which mentors, collaborators, or critics most shaped your thinking, and what exactly did you take from each?
The scientific crux:
Candidate-gene studies gave way to GWAS. Was the move from population GWAS to family-based designs another paradigm shift of the same order, or a quieter but deeper change in what the field thinks it is estimating?
If you had to explain to a policymaker why they should not read a GWAS effect size for educational attainment at face value, what would you say, and what should they trust instead?
Why did family-based designs stay peripheral for so long, given that they now look so close to the causal heart of the problem?
Where do you think the deepest disagreement lies between your work and the classical GREML / twin / heritability tradition: the target, the assumptions, or the interpretation?
When genetic nurture first came into focus, did it feel like a biological discovery, or a mathematical formalization of something sociologists already knew?
Are the same indirect effects, assortative patterns, and environmental correlations that make population GWAS causally impure also part of what makes them predictively useful?
What did educational attainment teach the field that height or disease traits could not?
Do you think of assortative mating mainly as a nuisance term to be corrected away, or as a substantive social and historical process that helps generate durable inequality?
Deep biology, evolution, and technology:
Inside cells, biology is profoundly nonlinear. Why, at the population level, do additive models work as well as they do for complex traits?
Where do you expect additivity to fail first—extreme tails, gene-environment interaction, rare variation, or embryo selection—and how much should we care?
What does the contrast between highly polygenic traits like educational attainment and more concentrated disease architectures tell us about selection, developmental constraint, and what the phenotype really is?
What, if anything, do you think we can now say responsibly about ongoing selection in humans once we apply the lessons of family-based de-confounding?
If future models absorb genomes, pedigrees, medical records, and phenotypes at enormous scale, what parts of causal genetics become easier, and what still require explicit family-based reasoning?
Governance, institutions, and translation:
Can you tell the Framingham/dbGaP story from the inside, and what it changed in your understanding of how science is actually governed?
To what extent has the limiting reagent become permission rather than data or mathematics? Do topic-based restrictions mostly block analyses after the fact, or do they start upstream by changing what cohorts dare to measure?
What led you from methodological critique into embryo screening and private-sector translation, and what would scientifically honest and ethically defensible use of that technology look like?
What do serious advocates of embryo screening get wrong, and what do serious critics get wrong?
Character, revision, and legacy:
Your career has often put you in the role of methodological critic—the person explaining why the headline number is too high or too impure. What is that role like inside a field?
Which criticisms of your work have genuinely improved it or changed your mind?
Which findings from the GWAS era do you expect to survive essentially intact once we have much richer trio and pedigree data, and which literatures do you think will look weakest in hindsight?
If we eventually sequence millions of pedigrees and cleanly map the obvious confounds, what fundamental mystery about human variation will still remain—and when historians write the GWAS era, will standard unrelated GWAS look more like a necessary stepping stone or a costly detour?
Further questions from Anonymous:
How did a political faction of professors at UCLA prevent you from being offered a professorship?
Describe in detail how research on genetics of intelligence has and is being suppressed in academic science. How would the field advance if this were not the case?
What would you do differently in life if you had another chance?
Share some juicy quotes from private conversations, without revealing any identities.
How heritable is g? How well can it be predicted, given the genomes and g scores of millions of individuals?
Machine learning underwent this interesting phase transition in the early 2010s from architectures that made heavy use of feature engineering, to a deep learning paradigm where features were learned directly from human data with no human involvement. It seems like genetics underwent a somewhat similar transition from candidate gene style studies to hypothesis free designs like genome-wide association studies.
Can you talk to us about what drove that transition and what new insights it revealed that would have been impossible before?
We know that within a cell, there’s a lot of non-linear processes happening, many of which involve proteins that are made by the genome. Yet one of the most widely replicated findings in the field is that genetic variants have mostly linear effects on disease risk or trait values.
How can this be?
There seems to be a huge number of children every year born from IVF that could benefit from polygenic embryo screening, yet less than 1% of the children born via IVF take advantage of this tech.
Can you talk about what’s holding back clinicians and regulatory bodies like the American Society for Reproductive Medicine (ASRM) from endorsing this technology?
If you look at the size of the impact different variants have on traits, there seems to be a huge amount of variation in the effect sizes. For something like Type 1 diabetes or Alzheimer’s, you find these genetic variants with huge impacts that are actually quite common. Whereas for a disease like schizophrenia, variants with impacts seem to be quite rare.
What’s going on here? Was there some evolutionary force that created this equilibrium?
Why do some traits like type 1 diabetes have a hundred genetic variants involved in determining someone’s genetic risk while traits like IQ have on the order of 10,000? Why does evolution seem to favor such massive polygenicity for some traits and not others?
There’s this common misconception that, because of modern medicine, human evolution has stopped. But obviously if some people are having kids and some people aren’t, there must still be selection going on.
What do we know about modern human selection? What traits are being selected for in humans right now?
There’s this somewhat popular notion on some corners of the internet that an “average” genome is better, in the sense that if someone just had all the most common alleles at every place in their genome, they’d be better off because they would have lower mutational load.
Is this true? Would someone with a modal genome be smarter or better looking than average?