Originally posted on the EA subreddit.
First, I will present a rough sketch for why genetic enhancement could be a plausible cause X. Then I will list some specific proposals for genetic interventions. I will conclude by responding to objections. If there is interest, I may write more posts on this topic.
Basic argument
There are two main ways to think about genetic enhancement as a potential EA cause area. One perspective is focused on improving short-term human welfare. While reducing “defects” (disabilities, depression, etc.) would be a major focus, this perspective could additionally encompass increasing the frequency of beneficial traits, such as longevity-promoting alleles of the FOXO3 gene. The key idea is that enhancement is performed to increase individual well-being.
The other way of thinking about genetic enhancement, and the one I prefer, takes a long-term view. Changes are made with the far-future of our society in mind. For instance, by drastically increasing IQs, we could put our civilization in a better position to solve complex challenges that exist today or will arise down the line. Other interventions such as increasing empathy and decreasing Dark Triad personality traits could be used to influence the values of our descendents and avert future moral catastrophes.
The short-term case for genetic enhancement
If we want to improve short-term human well-being, we can group most possible interventions into two broad categories:
- Improve the quality and duration of life for presently-existing humans.
- Change which (and how many) humans will be born in the first place.
The first method is far less philosophically controversial. Everyone agrees that once you’re born, it’s better to live a happy life than a miserable one. On the other hand, when it comes to changing the number and identity of people in existence, we enter the muddy waters of population ethics, fraught with paradoxes and impossibility theorems. For the most basic version of my argument to work, it suffices to assume that when selecting among a fixed number of “potential children” who could be born, we should choose the ones with the highest expected well-being.
I have not yet said anything about genes specifically! Of course, both environment and genetics could be factors when evaluating the expected well-being of potential future humans. However, living environment in general is already subject to more optimization (see next paragraph). It also appears that many traits that contribute to a happy, successful life are highly heritable, including psychological traits (see appendix 1). Further, taking a slightly longer-term view, genetic information is directly passed on for multiple successive generations, while environment is ephemeral. I think a focus on genes is justified.
Going back to the dichotomy of altruistic interventions, the first option is the tack that most effective altruists interested in short-term human welfare have taken. It is also the strategy most popular among do-gooders in general. In comparison, little philanthropic effort is devoted to efforts to (e.g.) decrease the number of children born with severe congenital diseases or to increase the frequency of welfare-promoting alleles in the population. Aside from philanthropy, people in general seek out the best living environment for themselves and their children. On the other hand, although sexual selection does exist, people generally do not explicitly optimize the genetics of their progeny. So it’s clear that improving population genetics is comparatively neglected.
Further, I would argue that genetic enhancement is quite important. A meta-analysis of twin studies found that genetic factors explain 36% of variation in subjective well-being (appendix 1). While environmental conditions obviously play a huge role in well-being as well, the role of genes cannot be understated. If all EA optimization to date has only been targeting 64% of the problem, there might be a lot of low-hanging fruit in that remaining 36%.
It is harder to tell how tractable genetic enhancement is, which depends on the specific type of intervention. I may write a follow-up post exploring this issue in depth if there is sufficient interest.
The long-term case for genetic enhancement
The long-term trajectory of our civilization depends largely on how we navigate the minefield of emerging technologies that will be developed over the next couple of centuries. It likewise depends on the values that our descendents hold.
It seems that increasing IQ and other traits such as affective empathy (see appendix 1) could help humanity to reach a prosperous future. Imagine a world full of people as bright as John von Neumann and as ethical as Gandhi. Wouldn’t such a world be in a vastly better position to avoid existential risks and risks of future suffering than our present Earth?
Bostrom, and others, are fond of saying that we are the “stupidest possible biological species capable of starting a technological civilization.” In many ways, genetic enhancement would change this outlook. If the brightest humans on Earth one day become much wiser than we currently are, then I believe it is highly likely that we could improve the chances of successfully mitigating existential risk.
Consider the case of aligning an artificial intelligence. One model of AI development is that it is a constant race between capabilities gained by factors which provide virtually no safety guarantees, such as increasing hardware capacity, and capabilities gained by a principled understanding of AI design, such as the presentation of causality provided by Judea Pearl. MIRI believes that if we have a principled understanding of advanced AI design before we can build it via stupid means, like simulating evolution, then our chances of aligning such AI are increased dramatically.
In order to gain the upper hand in this race for technological capabilities, we could leverage genetic enhancement to maintain a favorable balance in differential technological development. For instance, progress in hardware is arguably bottlenecked by economic demand, and would not be significantly accelerated by the advent of a hundred John von Neumann level scientists. However, deep insights into the nature of intelligence are the type of thing we should expect if we have a highly competent core group of humans working on the problem.
Ensuring that this highly skilled group of people comes into existence would be much easier to accomplish than the widespread adoption of genetic enhancement implied by the short-term view I presented above. Therefore, it is likely more tractable. Still, selecting embryos for intelligence is highly controversial, and would not be something that most ethics panels would currently approve of.
In the long term, I believe selecting embryos for favorable traits will happen anyway, regardless of ethical qualms, because once the technology has been demonstrated, countries unwilling to adopt it will risk falling far behind. EAs can therefore do research to track attitudes, and find the right time to begin implementing the strategy outlined above.
Interventions
I have explained why genetic enhancement is an area worth considering, but what can effective altruists actually do to advance the cause? I will now list a few promising, non-coercive forms of genetic enhancement. For each of the following items, EA work could focus on advocacy, research, or technical implementation.
I would highly recommend Gwern’s article on “Embryo selection for intelligence” for a detailed comparison of the feasibility and effectiveness of several different genetic interventions. (Although the title refers to intelligence, the analysis applies equally well to other genetic traits.)
Incentive-based genetic enhancement: We might consider programs that pay for people with desirable traits to reproduce. For example, we could provide high pay to quality sperm and egg donors based on their genetic profiles. We can also leverage recent research in genetics that predicts success based on one’s genetic profile, and pursue further research along these lines. This will help us make the case for incentive-based genetic enhancement, and will provide an effective means to discover high-quality donors.
Embryo selection: First performed in 1990, preimplantation genetic diagnosis (PGD) is a process by which a handful of eggs fertilized via IVF are screened for genetic diseases prior to being implanted in utero. The screening has traditionally been done using FISH (to detect chromosomal abnormalities like Down’s syndrome) or PCR (to diagnose monogenic disorders like sickle-cell anemia). Of the screened embryos, the healthiest one is implanted. There are several different potential variations of the procedure. As would be expected:
- The more embryos produced, the better (but diminishing marginal benefits)
- The more traits screened for, the better.
Germline gene engineering: There is a lot of hype surrounding the idea of using CRISPR to modify embryos. As many of you know, it’s already been done. While CRISPR does have a great potential to e.g. treat monogenic diseases, Gwern doesn’t think it will produce enormous effects like future versions of embryo selection could. This is partially because genetic studies are not good at isolating which specific single nucleotide polymorphisms are causally responsible for a trait. If we modify a gene which is merely correlated with the trait we are trying to augment, it may be ineffective or possibly even backfire. Moreover, if we want to edit polygenic traits, many individual edits would be required, but currently CRISPR can only be used to make ~5 edits reliably. On the bright side, CRISPR would allow us to increase the frequency of rare beneficial alleles and even create novel “mutations” that we hypothesize to be beneficial (the latter, though, is very risky and wouldn’t pass an ethics board).
Iterated embryo selection: This is a hypothetical technology that could be used to exert a great degree of control over the genome. It involves collecting stem cells from different donors, differentiating these cells into sperm and eggs, and then allowing the gametes to fertilize each other. The zygotes with the most desirable genomes would be differentiated back into sperm and eggs, and the rest discarded. The process can be repeated for several iterations, “compressing multiple generations of selection into a few years or less.” Gwern is quite optimistic about the potential of this technology, expecting it to increase IQ by multiple standard deviations in one generation.
Genome synthesis: This refers to creating a completely new genome from scratch. This procedure would allow the greatest degree of control, and Gwern is quite optimistic about it, but there are several technical challenges that would need to be overcome for implementation.
In general, I think there is high value in having EAs enter government and work on shaping relevant regulations in a more positive direction. It’s additionally possible that conducting more rigorous genetic studies would be useful, but it’s not clear how EA can have a large counterfactual impact there, because academia and the genetic profiling industry are already working on the issue.
Objections and responses
What if increasing these supposedly positive traits results in negative consequences?
It’s important to avoid status quo bias. To quote Bostrom and Ord:
Reversal Test: When a proposal to change a certain parameter is thought to have bad overall consequences, consider a change to the same parameter in the opposite direction. If this is also thought to have bad overall consequences, then the onus is on those who reach these conclusions to explain why our position cannot be improved through changes to this parameter. If they are unable to do so, then we have reason to suspect that they suffer from status quo bias.
But we can give an explanation: evolution! Doesn’t natural selection already work in favor of desirable traits? Isn’t it hubris to think we know better than nature?
Bostrom and Ord give four reasons why this argument is dubious:
- The environment of our evolutionary ancestors is different in many ways from our modern world. It’s possible that what is beneficial today was an evolutionary disadvantage for our ancestors, or vice versa.
- There may have been trade-offs in the past that are no longer relevant. For instance, we no longer have to worry so much about large brains imposing high metabolic costs, because food is widely available.
- Evolution is a blind process, and it’s possible it just never happened to stumble onto the correct combination of genes.
- What we care about is not the same as evolutionary fitness. Evolution doesn’t optimize for happiness. The ability to rape and plunder might increase genetic fitness, but we don’t consider them good. Likewise, there could be traits which humans value but that hurt fitness.
Genetic enhancement technology would only be available to the rich. It would greatly increase inequality.
I would like to point out that genetic enhancement is not necessarily a zero-sum game. Sure, there are some genetic traits that are almost exclusively positional goods, i.e. they benefit one person by increasing their status over others. Examples might include physical attractiveness or height. On the other hand, many other characteristics such as health, well-being, and intelligence are considered good in and of themselves. We should focus on the latter.
The enhancement of economically advantageous traits such as intelligence would grow the overall pie of the economy, which we could then redistribute more equitably. Then we’re back to standard political debates about how to set the marginal tax rate.
While it is quite probable that the wealthy would have earlier access to reproductive technology, the price would eventually drop to the point where it could be made available to anyone. If necessary, governments could provide social security benefits to subsidize access for the poor.
One way of alleviating the harm due to inequality is by advocating a tax on innate, unearned qualities, such as favorable genetics and inheritance. I believe that these policies will be popular once the technology comes up on the horizon, and will likely play a large role in mitigating the worst risks of inequality.
Genetic enhancement is too taboo for advocacy to make headway.
I see a few reasons for optimism. First of all, surveys indicate significant support for genetic enhancement and similar ideas. A Pew Research poll found that a majority of Americans would support using gene editing on embryos to treat diseases, although only 20% supported using the same technique to increase IQ. Furthermore, when push comes to shove, 90% of fetuses diagnosed with Down syndrome are aborted.
The bottom line is that genetic enhancement isn’t too far out of the Overton window for advocacy to be futile. At the same time, it’s not so universally accepted that we can assume highly effective enhancement will happen before artificial intelligence arrives. It’s in the perfect zone where a concerted effort on the part of EA advocates could make a difference, either by shifting the development timelines or by influencing policies and norms surrounding its use.
AI timelines are too short for genetic enhancement to have any impact
AI timelines are quite uncertain. According to a survey of AI researchers, the median estimate for the arrival date of “high-level machine intelligence,” defined as AI capable of exceeding humans at all tasks, is 40–50 years away, which would allow for one or two intermediate generations. On the other hand, the 75th percentile date is more than 100 years away. Personally, I’m skeptical of the reliability of these estimates, and I recommend to take them with a grain of salt.
In any case, Bostrom and Shulman have argued that even a single generation of iterated embryo selection for intelligence, limited to a small proportion of the population, could have a massive impact on society. The same goes for genome synthesis and advanced forms of “one-shot” embryo selection. There is likely substantial room for impact in the median scenario.
Genetic enhancement would greatly decrease genetic diversity, leaving us more vulnerable to pandemics or other unforeseen disasters.
While this is technically true, I’m not too worried about it personally. The marginal risk increase seems small enough that the benefits of genetic enhancement dwarf it. It should be noted that this objection is strongest in scenarios where enhancement is highly widespread and where technology enables large jumps in one or two generations.
Selecting for high IQ would make the world more vulnerable to agential risks (e.g., lone wolf extremists building nukes in their backyard)
(Relevant background reading: Torres, Bostrom)
While a society composed of high IQ individuals would indeed be more likely to contain individuals capable of building WMDs in their backyard, such a society would also have smarter control and surveillance methods for preventing acts of terror. Moreover, intelligent people tend to commit less crime. But even if we assume that this one specific risk would increase, it still seems likely to me that total existential risk would decrease, for the reasons I mentioned in the “long-term” argument.
We risk creating a race of enhanced humans who won’t care about (or will subjugate) the rest of us.
First of all, this concern is usually based on science fiction like Gattaca. I would warn that generalizing from fictional evidence is not a reliable way to arrive at true beliefs.
Beyond that, research shows that intelligent people are more altruistic and less discriminatory rather than the opposite.
Even assuming the above research is false, from a principled perspective, there seems to be no compelling moral reason to keep our gene pool the way it is. It is typically assumed that creating a class of people who are highly cognitively capable would be unfair to the rest of us.
However, disparities in cognitive ability already exist via natural means. To the extent that these natural disparities are acceptable, then inducing further changes doesn’t fundamentally “change the rules of the game.” In fact, these technologies could actually level the playing field, if we allowed broad distributed access to them.
Historically, increases in the average intelligence of populations has been overwhelmingly considered a positive thing. The long running Flynn effect has likely contributed to lifting nations out of poverty, achieving the exact opposite effect of the dystopian worries we are often reminded of.
Conclusion
Genetic enhancement is a plausible candidate for a cost-effective cause area. There are both short-term and long-term arguments for the desirability of genetic enhancement, and several different approaches that could be used to improve the gene pool. The field is highly neglected and important, but tractability remains uncertain.
The Centre for Effective Altruism has stated that they believe EA needs to incorporate a variety of different approaches to addressing existential risk. I believe genetic enhancement is a cause area that belongs within the effective altruist portfolio. At the very least, it deserves much more attention at the level of cause prioritization than it has received until now.
Appendix 1: Selected traits with heritability estimates
You can see more at SNPedia.
Neuropsychological disorders
Physical disorders
Important psychological constructs
- Affective empathy: 52-57%
- Intelligence: high (see Heritability of IQ for background; the topic is somewhat contentious, but most estimates are >50%, with recent ones >80%)
- Well-being: 36%
- Dark Triad traits: 31-72%
- Big Five personality traits: 41-61%
Also worth pointing out: George Church has a list of single gene variants that have already been shown to have a large effect on human welfare.
Appendix 2: Genetic enhancement for animal welfare
We already have evidence that artificial selection pressure applied by humans can powerfully shape evolution. Look no further than the wide variety of domestic animals and cultivated plants that owe their unique features to the whims of their breeders. Unfortunately, animals have generally been bred for their utility to humans even when this comes at the expense of their own welfare. Over the past century, livestock breeders have substantially increased productivity through genetic modifications. The same modifications have introduced a plethora of horrific afflictions that affect billions of farm animals every year.
While animal breeding may seem like a cause for despair, it also offers us a potential solution to reduce farm animal suffering. Adam Shriver has argued that we should genetically engineer livestock that have a diminished (ideally eliminated) capacity for suffering. This is a research area with huge potential for impact, although it does have some drawbacks as well. Specifically, we would need to be sure that we were actually reducing the feeling of suffering, and not just the behavioral reaction to the feeling. Furthermore, some ethical theories suggest that it is wrong to exploit and kill animals full stop, even if the animals are not in pain. Finally, reducing pain may hamper efforts to completely end the practice of animal farming, as animal rights “abolitionists” argue. Despite the concerns, I would nonetheless be excited to see more research on this issue.
In recent years, there has been growing moral concern for the suffering that wild animals experience due to natural processes such as disease, parasitism, and Malthusian scarcity. Geneticist Kevin Esvelt, one of the pioneers of CRISPR/Cas9 gene drive technology, has suggested that we may have a moral obligation to use our technological powers for the benefit of wild-animal welfare. For decades, transhumanist philosopher David Pearce has been arguing along similar lines. While these ideas are still extremely speculative, further research on the topic could be valuable.
(Epistemic disclaimer: My understanding of genetics is very limited.)
If additive heritability for all the relevant personality traits was zero, many interventions in this area are pointless, yes.
I might have underestimated this problem but one reason why I haven’t given up on the idea of selecting against “malevolent” traits is that I’ve come across various findings indicating SNP heritabilities of around 10% for relevant personality traits. (See the last section of this comment for a summary of various studies).
SNP heritabilities of ~10% (or even more) for relevant personality traits seem also not implausible on theoretical grounds. If I understand Penke et al. (2007, see Table 1 in particular) correctly, balancing-selection models of personality predict that personality traits should show less additive heritability than, say, cognitive ability, but not (necessarily) zero additive heritability.
Granted, 10% is pretty low, but is it hopelessly low? According to Karavani et al. (2019), a polygenic score for IQ which explains 4% of the variance, would enable an average increase of 3 IQ points (assuming 10 available embryos). I infer from this that a polygenic score which can explain only ~4% of the variance in, say, psychopathy would still enable the reduction of ~ 1/5 of a standard deviation in average psychopathy scores, assuming 10 embryos. Polygenic scores explaining ~10% of the variance might thus enable considerably larger average reductions of ⅓ - ½ of a standard deviation or so (numbers pulled out of my posterior).
Again, ⅓ of a SD might seem underwhelming but, as you emphasize in your essay on embryo selection, small changes in the mean of a normal distribution can have large effects out on the tails, so this could still lead to surprisingly large reductions in the frequency of extreme psychopathy or sadism (~psychopathy scores 2-3 SDs above the norm), even in “normal” IVF embryo selection. When applied in iterated embryo selection (IES), this could result in much stronger effects still.
Again, I could easily be wrong about any of the above.
Will SNP heritability estimates increase with larger sample sizes?
This is at least what Tielbeek et al. (2017) suggest: “Recent GWASs on other complex traits, such as height, body mass index, and schizophrenia, demonstrated that with greater sample sizes, the SNP h2 increases. [...] we suspect that with greater sample sizes and better imputation and coverage of the common and rare allele spectrum, over time, SNP heritability in ASB [antisocial behavior] could approach the family based estimates.”
Higher additive heritability for personality disorders?
Another point that makes me somewhat hopeful is that specific personality disorders seem to show larger additive heritabilities than personality traits. For example, the meta-analysis by Polderman et al. (2015, Table 2) suggests that 93% of all studies on specific personality disorders “are consistent with a model where trait resemblance is solely due to additive genetic variation”. (Of note, for “social values” this fraction is still 63%).
And a lot of the benefits in this area might come from selecting against, say, antisocial or narcissistic personality disorder (sadly, sadistic personality disorder is not a thing anymore but it was included in the appendix of DSM-II).
But it’s been a while since I read the Polderman paper and I’m also a bit confused by how there can be high additive heritability for, say, narcissistic personality disorder but very low additive heritability for narcissism as a trait, so the above might be wrong.
Some interventions in this area don’t require additive heritability
There are also interventions that work, even if additive heritability is zero though they assume that the non-additive genetic variance is at least partly due to dominance and not solely due to epistasis; I think. For example, ensuring that the parents of the first generation of IES embryos score low on dark tetrad traits or influencing the first genome synthesis projects to make their first genomes as similar to those of people scoring low on dark tetrad traits as possible (alongside edits to achieve substantial IQ increases, of course).
Lastly, there are interventions that have nothing to do with genetic enhancement but would benefit from more research on and advocacy for screening against malevolent traits and are thus somewhat related to the above. For example, it seems valuable to develop better measures of malevolent traits, potentially ones that are impossible to game such as neuroimaging techniques. Such measures could then be used in various high-impact settings to. For example, they would enable decision makers to better screen for highly elevated dark tetrad traits in government officials, humans whose brains will be used to create the first ems, and human overseers in AI projects. (Currently, all measures of malevolence seem to be self-report questionnaires or interviews which seem easily gameable by smart psychopaths.)
Is non-additive genetic variance really useless?
(No need to reply to the questions in this section.)
Assume that all of the genetic variance in trait A is due to dominance. Wouldn’t it still be possible to achieve non-zero increases/decreases in trait A via (iterated) embryo selection?
And what about epistasis? Is it just that there are quadrillions of possible combinations of interactions and so you would need astronomical sample sizes to achieve sufficient statistical power after correcting for multiple comparisons?
Some evidence for non-zero SNP heritabilities of relevant personality traits
Table 4 of Sanchez‐Roige et al. (2018) provides a good summary. Below, I focus on studies examining traits that likely correlate with dark tetrad traits, such as agreeableness and conscientiousness.
The UK biobank (N ≈ 290k) estimates SNP heritabilities of around 10% for various personality traits, some of which probably even correlate with psychopathy, such as the items “do you often feel guilty?” and “Do you worry too long after an embarrassing experience?”. (Don’t get me wrong, I’m not saying that we should select against traits that only correlate with psychopathy while being completely fine in themselves, like e.g. not often feeling guilty. I'm just listing these items to support the hypothesis that as long as we can find SNP heritability for them, we can expect to find SNP heritabilities for related traits as well.) Unfortunately, the UK biobank didn’t seem to measure any personality trait apart from neuroticism (estimated SNP heritability: 11%). Also, they usually didn’t even use likert scales, only dichotomous yes/no responses, which might reduce heritability estimates (??).
A GWAS (N = 46,861) by Warrier et al. (2018) found an additive heritability, explained by all the tested SNPs, for the Empathy Quotient of 11%. (The Empathy Quotient contains items like “I get upset if I see people suffering on news programmes” and “I really enjoy caring for other people” and thus probably correlates negatively with dark tetrad traits.)
Verweij et al. (2012) give a SNP heritability of 6.6% for harm avoidance which likely correlates with dark tetrad traits.
Lo et al. (2017) estimate SNP-based heritabilities of 18% for extraversion, 8.5% for agreeableness and 9.6% for conscientiousness (see supplementary table 2). (N = 59,176).
Granted, Power and Pluess (2015) estimate SNP heritability of agreeableness and conscientiousness as 0%. However, their sample size of 5,011 is much smaller than the sample sizes above and they write: “It is worth noting that the large standard errors around the negative findings suggest increased sample size may identify a low but significant level of heritability.”