Table of Contents


Perspectives on Human Gene-Editing

My perspective on potentially having gene-edited children, and a very brief overview of what the public and bioethics community have to say about “designer babies” and about human gene-editing; these, I believe, are useful precursors for forecasting human gene-editing futures.

In the 2021 Dictionary of Global Bioethics, Henk ten Have and Maria do Céu Patrão Neves1 offer the following definition of a “designer baby” (pp. 405-406):


The children of parents who wish to design their offspring are called “designer babies.” The term was coined by the media to refer to genetic interventions in preimplantation human embryos in which the aim was to select or alter the traits future children would have.

Personal Take

As of the last edit date of this post2, my partner and I have had on the order of 10 or so discussions about whether we want to have children, and if so, whether we want to have our eggs3 or sperm4 frozen, have in-vitro fertilization5 done, have our embryos screened6, or have our children genetically edited7. We operate, in our talks, as though these things will be affordable and available to us when the time comes. Neither me nor my partner are very familiar with the costs, benefits, and feasibility of these procedures; we are, however, interested in improving, in as many ways as possible, the lives of our potential progeny.

  I would like for any children I have to be physically healthy and to be, in the following order of preference, highly intelligent, conscientious, creative, and amiable, among other things8. My partner is primarily concerned with her children being bipolar or depressed, or being prone to taking dangerous risks; she doesn’t want these things to occur. Both of us are interested in having genetic-engineering employed to achieve these ends, but are still quite uncertain about its practicality. It’s highly likely that we won’t have any children until at least 5 years from now. Will gene-editing be legal? effective? expensive? when we decide to have children? How useful, available, and painless9 will polygenic or embryo screening be, potentially as an alternative to gene-editing? These and similar questions motivate me to forecast on this topic.

  In considering the lives of our children, my partner and I have tried not to neglect exploring the “nurture”10 or environmental component of development, but have really not thought deeply about what parenting practices we’d like to adopt, or about other environmental features, such as whether we want to raise our children in an urban or a rural setting. Regardless, human gene-editing - done correctly - seems like something that could benefit society extensively, and is something my partner and I presently believe we want for our future children. What do other people think?11

Public’s Take

In 2020, the PEW Research Center performed a survey12 of 20 publics13 in an attempt to gauge global support for biotechnology research, and to observe how religion was associated with people’s approval levels. Regarding human gene-editing, they found the following:

A 20-public median of 63% say scientific research on gene editing is a misuse – rather than an appropriate use – of technology, according to the survey fielded in publics across Europe, the Asia-Pacific region, the United States, Canada, Brazil and Russia.

However, views on specific instances where gene editing might be used highlight the complex and contextual nature of public attitudes. Majorities say it would be appropriate to change a baby’s genetic characteristics to treat a serious disease the baby would have at birth (median of 70%), and somewhat smaller shares, though still about half or more, say using these techniques to reduce the risk of a serious disease that could occur over the course of the baby’s lifetime would be appropriate (60%). But a median of just 14% say it would be appropriate to change a baby’s genetic characteristics to make the baby more intelligent. A far larger share (median of 82%) would consider this to be a misuse of technology.

So, it appears that a majority of people approve human gene-editing for treatment purposes, and most people disapprove of gene-editing for enhancement14.

  The remainder of the survey is worth looking over, and generally indicates that those who are male, those who are the median age or younger, and those who are more educated are more supportive of human gene-editing; these properties I believe are fairly robust, i.e. I don’t expect the predictiveness of these properties in 2022 to have changed much since 2020, and also don’t expect them to change much in the coming years.

  Any change in approval sentiments (in 2020, median 63% believe gene-editing research is a misuse, and median 70% and median 60% support for using gene-editing for treating illness at birth and illness over lifespan, respectively) I believe will be in the direction of more support; I expect human gene-editing to become safer, more effective, etc… as it is researched more. Religion seems to be declining, which will very likely erode sentiments that human gene-editing is a misuse of science. There’s a chance that, given my support for human gene-editing, my belief that the public’s approval of human gene-editing will gradually increase reflects some degree of optimism bias, but I’ve tried to control for this.

  A final and highly intriguing note on the public perception’s of gene-editing in humans is that, above all other publics surveyed, India stands out as most readily accepting of human gene-editing, with 56% of adults (highest percentage measured in the survey; the next highest was South Korea at 47%, Taiwan at 44%, and Singapore at 38%) considering research on gene-editing to not be a misuse of science. In terms of using gene-editing for human enhancement, India stands out even more, with 64% of respondents supporting changing a baby’s genetic characteristics to increase intelligence (Malaysia came next at 44%, and then Taiwan at 31%). This indicates to me that while India might not be the first nation to legalize human gene-editing, it might eventually be the largest market for both treatment and enhancement purposes of gene-editing, which is something that might lead to it becoming, at some point, the nation with the most gene-edited babies.

Bioethicists’ Take

Eugenics stands out in my mind as something that most people (laymen and scientists alike) associate with human gene-editing. While I am not too familiar with the history of this topic, my present intuition is that most researchers whose work falls into fields involving genetics, gene-editing, CRISPR-Cas9, heritability, etc… are incentivized to dissociate their work, as much as possible, from anything that might conjure up in people’s minds the idea of eugenics.

  Partially out of social desirability, some researchers who study gene-editing might detest its application in humans, or might appeal to ridicule in their assessments of the long-term prospects of human gene-editing, at least for enhancement purposes. This is not to say that geneticists and those with the capacity to alter human genomes care little about the ethical and environmental considerations of genetically engineering humans, just that some of these researchers might cautiously support genetically engineering humans but not advocate for or support it publicly. It seems somewhat difficult to argue against gene-editing in humans for treatment purposes, so I suspect that bioethicists may be more split on this issue.

  Ever since the He Jiankui affair, where He Jiankui of the Southern University of Science and Technology (SUSTech) in Shenzhen used CRISPR-Cas9 to genetically modify a set of embryos with the intent of immunising them against HIV (ultimately three gene-edited babies were born, a set of twins and another child), bioethicists seem to have taken, generally, a firm stance against gene-editing in humans15.

  Such a stance was illustrated when, following the He Jiankui affair, the World Health Organization (WHO) recommended discontinuing applications of human germline genome editing for the near-term future in their 26 July 2019 statement:

The WHO expert advisory committee on governance and oversight of human genome editing convened on 18-19 March 2019. At this meeting the Committee in an interim recommendation to the WHO Director-general stated that “it would be irresponsible at this time for anyone to proceed with clinical applications of human germline genome editing.”

More recently, the WHO’s Expert Advisory Committee on Developing Global Standards for Governance and Oversight of Human Genome Editing provided an updated stance on human germline genome editing in their 2021 report Human Genome Editing: Recommendations:

Human genome editing has great potential to improve human health and medicine. Human genome editing technologies can be used on somatic cells (non-heritable); germline cells (not for reproduction) and germline cells (for reproduction). Potential benefits of human genome editing include new strategies for diagnosis, treatment and prevention of genetic disorders; new avenues to treat infertility; new ways to promote disease resistance; contribution to vaccine development and enhanced knowledge of human biology. For example, application of somatic human genome editing has already been undertaken, including in vivo editing, to address HIV, sickle-cell disease and transthyretin amyloidosis1. Germline human genome editing contributes to deepen our understanding of the role of specific genes and processes in early human development, physiology and diseases. However, there are important areas of ongoing uncertainty as to potential benefits and risks, and gaps in scientific understanding in such key domains as off target effects and long-term risks.

At the same time, however, somatic, germline and heritable human genome editing raise important and outstanding ethical and social issues. Challenges associated with somatic human genome editing include, for example, rogue clinics, medical travel, as well as the reporting of illegal, unregistered, unethical or unsafe research and other activities including the offer of unproven so-called therapeutic interventions. Heritable human genome editing also gives rise to great concerns as the edit might be passed to subsequent generations. Additional issues include enhancement to improve certain traits, the lack of diversity in collections of human samples and associated data, the need for equity of access to and benefit from human genome editing. There are important differences in the scale of the current challenges posed by somatic, germline and heritable human genome editing.

This outlook on human gene-editing appears more optimist than those in the recent past, but is still highly critical of the plausible ill-consequences of the technology.

  The National Human Genome Research Institute (NHGRI), another organization I found that oversees, in great depth, research activities pertaining to human genome editing, takes a similar stance, i.e. they acknowledge that human gene-editing could greatly help humanity, but also place immense weight on considering unintended deleterious societal and biological consequences that might emerge as a result. In their Eugenics timeline, the NHGRI writes

Recently, groundbreaking technologies such as CRISPR-Cas9, have raised concerns about using genome editing methods to make genetic enhancements. In response, some prominent geneticists have publicly requested to stop such enhancements.

Eugenics remains a constant issue in society and the scientific community, and the NHGRI is committed to monitoring its presence and confronting its inaccuracies.

One might suggest that the official statements made by the WHO’s or NHGRI’s ethics boards fail to accurately capture the diverse sentiments held by the researchers with the technical know-how to edit human genomes or by those who work adjacent to such researchers. The paper Attitudes of Members of Genetics Professional Societies Toward Human Gene Editing (Armsby et al.) reports the results of a 2017 online survey given to those with “genetics training”; the paper’s abstract:

Gene-editing technologies have improved in ease, efficiency, and precision. Although discussions are occurring around acceptable uses of human gene editing, limited data exist on the views of genetics-trained individuals. In 2017, we distributed an anonymous online survey to assess the attitudes of members of genetics professional societies toward gene editing (N = 500). Virtually all respondents were supportive of somatic editing in basic- science (99.2%) and clinical (87.4%) research on nonreproductive human cells. Only 57.2% were supportive of germline-editing basic-science research; 31.9% supported the transfer of viable embryos to humans for clinical research. While most favored future therapeutic uses of somatic (96.6%) and germline (77.8%) editing, there was little support for enhancement in somatic (13.0%) or germline (8.6%) cells. This study describes attitudes toward gene editing from genetics professionals worldwide and contributes to ongoing discourse and policy guidance in this domain.

Note that this survey16 occurred prior to the He Jiankui affair. If this same survey had been conducted not long after the He Jiankui affair, I believe its findings might have changed considerably, with more researchers not supporting germline-editing in humans. Nonetheless, I believe that researchers’ real sentiments (including those they might keep private) wouldn’t have changed much due to the affair. One way to think about this might be by imagining the following situation:

It's 2017. You have genetics-training and believe human gene-editing for enhancement purposes is entirely valid (the whole deal - intelligence, attractiveness, hair color, etc...). A researcher in China oversteps current ethical boundaries, producing 3 gene-edited humans and shifting the window of social acceptability of human gene-editing more towards it being perceived as "deplorable", "a misuse of science", etc... Do you stop believing human gene-editing for enhancement purposes is entirely valid?

I don’t think such an individual would cease believing this; I think this individual would simply avoid publicly advocating for human gene-editing for enhancement purposes, out of concern for their research funding and social desirability. So, controlling for the sentiment drift driven by continuing progress (safer, more precise, cheaper human gene-editing and better governance measures) in the field, I don’t think the distribution of sentiments of researchers with genetics-training has changed drastically since the He Jiankui affair.

  What were the demographics of the survey participants? At a glance, there were 500 participants; the largest age groups by a sizable margin were the 25-29 and 30-34 year-olds; most participants came from North America (63.5%); most participants (42.5%) were genetic nurses or counselors, and the next largest group (29.7%) was the research scientists; and a combined (71.3%) were not religious at all or were slightly religious. For the full demographics table of Armsby et al., see the Appendix.

  The opinions most of interest (at least to me) are those of the 36 clinical laboratory scientists, 148 research scientists, and 17 ELSI17 researchers (I suspect these individuals would actually be the ones responsible for performing experiments on, clinically applying, or developing policy for human gene-editing) in the study, but the paper does not partition the results by Profession, which I find unfortunate.

  What did these researchers think of the acceptability of human gene-editing? The study uses the categories “Agree Strongly/Agree”, “Neutral”, and “Disagree Strongly/Disagree”. As an aside for later, I believe people in the “Neutral” category will, for the most part, drift into and out of this group depending on future affairs or progress in human gene-editing. So, if there is some debacle, many people with neutral sentiments might shift to the disagree camp and people with agree sentiments might shift to the neutral camp. Likewise, if there is some clear form of progress in human gene-editing technology, application, or governance, the reverse might occur: disagree → neutral, and neutral → agree.

  At a glance, ~88% of participants agreed that gene-editing human somatic cells18 (treatment of heritable diseases; edits won’t be passed on to future individuals) is morally acceptable; ~48% agreed that gene-editing of human germlines is morally acceptable (~32% neutral); ~41% agreed that parents have a right to gene-edit their children (~32% neutral); and ~36% disagree with the idea that unborn individuals have a right to a genetic inheritance free from genetic engineering (~30% neutral). Armsby et al. also looked at specific use cases of gene-editing in humans - generally, around ~70% of participants were strongly supportive of using somatic gene-editing for treatment of physical or intellectual disabilities, while only ~41% were strongly supportive of using germline gene-editing for treatment of physical or intellectual disabilities. There was little support across the board for using somatic or germline gene-editing to improve appearance, physical abilities, or cognitive abilities. See Appendix for the graphs of acceptability of human gene-editing and its uses.

  If we take this opinion landscape to generalize well to other people with genetics-training and to not have changed very much since 2017, then it seems that the stance that people with genetics training take on human gene-editing is very similar to that of the general public (both were around 70% in support of gene-editing for treatment). The surveyed opinions of those with genetic-training appear to support the clinical application of gene-editing fairly more than the WHO’s and NHGRI’s positions do. While the total picture is incomplete to me - I am generalizing my fractional perspective of the stances on human gene-editing taken by the WHO, the NHGRI, and those with genetics training - the consensus appears to be most people (including the scientists with the capacity to perform gene-editing) support somatic cell gene-editing for treatment purposes, but most do not support it for enhancement; organizations involved in governance are slightly more weary of gene-editing, even for treatment purposes. Germline genome editing of human embryos is not well supported for treatment, and is supported even less for enhancement..


The Near-Term Future

A discussion of some considerations and scenarios for forecasting the number of designer babies in the near-term future (2030), along with my specific forecasts of this quantity.

There are many things to consider when forecasting how many designer babies19 there will be in the near-term future, as it is a rare (as of March 2022, only 3 gene-edited humans are known to have been born - those exclusively from the He Jiankui affair) and divisive event. From my present vantage point (not thinking about analogies or scenarios), my intuition is that there will not be many more gene-edited individuals born anytime soon, simply given the lack of support it receives from the organizations responsible for its governance. Putting an exact number on this is more difficult; nonetheless, this is what I try to do below.

  Let’s be clear with what we are forecasting20; for this post, I will operationalize “designer baby futures” using Pablo’s question, on the forecasting aggregator platform Metaculus.

Genome editing is a type of genetic engineering in which DNA is inserted, deleted, modified or replaced in the genome of a living organism (Wikipedia). The first gene-edited babies — Lulu and Nana — were reportedly born in October 2018.

How many gene-edited babies will have been born worldwide by the end of 2029?

Question resolves according to birth counts given in the first authoritative report (so judged by the admins) to cover the entire 2029 calendar year, as well as all years preceding it.

Categories: Biological Sciences – Bioengineering Biological Sciences – Genetics Human Sciences – General Impactful Forecasting Prize Social issues

Even though this question only asks about the number of designer babies by January 1st 2030, thinking about other years as well (e.g. 2040, 2050, etc…) could be fun, but 2030 seems decent at adequately capturing the “just past the horizon of certainty” and “a milestone year for humanity” feeling (2025, which is 3 years away, feels very temporally close and, with regard to human gene-editing, feels predictable, but 2030 is farther away, and what happens with human gene-editing by then seems much less certain).

  To make more accurate forecasts on Pablo’s question it can be useful to break the interval he used in his question ([0, 1e6] log-scaled) into smaller intervals. These smaller intervals can then be assigned a likelihood, and after all the intervals are given a likelihood, each can be given normalized probability21.

import numpy as np

# Take Pablo's interval and break it into 5 groups
log_space_interval = np.logspace(start=0, stop=6, num=5)
intervals = [int(np.ceil(elt)) for elt in log_space_interval]
int_tups = list(zip(intervals[:-1], intervals[1:]))
int_tups = [(elt[0], elt[1]) for elt in int_tups]
int_tups[1:] = [(elt[0]+1, elt[1]) for elt in int_tups[1:]]

print(int_tups)
#[(1, 32), (33, 1000), (1001, 31623), (31624, 1000000)]

My naive estimate (without weighting different scenarios or looking at analogies) for how many designer babies will be born by January 1st 2030 goes as follows:

Likelihood (0/100) for 1-32 babies?: 95
Likelihood (0/100) for 33-1000 babies?: 80
Likelihood (0/100) for 1001-31623 babies?: 20
Likelihood (0/100) for 31624-1000000 babies?: 10

You assigned the following probabilities to the intervals:
1-32 : 46.34%
33-1000 : 39.02%
1001-31623 : 9.76%
31624-1000000 : 4.88%

With 90% confidence, you expect the # of gene-edited babies to be between 1 and 15556.

Now, I will try to make a more detailed estimate. From my explorations of human gene-editing, 3 principal components, which I believe subsume most of the plausible factors guiding the future application of human gene-editing, stand out to me: legality, availability, and desirability. Generally, I would describe these as

  • Legality: The state of global governance of human genome editing for treatment or enhancement, including how states respond in the future to situations such as the He Jiankui affair.
  • Availability: How easy, globally speaking, it is to have your child be gene-edited, regardless of whether it is illegal.
  • Desirability: To what degree society finds human gene-editing unacceptable and dissuades its application, to what degree people generally desire to have their children be gene-edited, and to what degree some people are willing to go to have their children be gene-edited.

Here are some questions regarding desirability, availability, and legality seem useful for thinking about what factors influence the number of gene-edited births by 2030.

  • How many people desire to have their children gene-edited?
    • How many people desire to treat their children through gene-editing?
      • How might this change?
      • Are there are alternative treatments outside of gene-editing?
    • How many people desire to enhance their children through gene-editing?
      • How might this change?
      • Are there are alternative treatments outside of gene-editing?
    • Will there be any social stigma around having gene-edited children?
    • How painless9 will human gene-editing be?
    • How effective (works as desired) will human gene-editing be?
  • How available will human gene-editing be?
    • Will private organizations or individuals offer gene-editing services?
      • If so, where, when, and to what extent?
      • How easy will it be for a researcher to perform human gene-editing?
    • How expensive will human gene-editing be?
  • Will human gene-editing be legal?
    • If so, where, when, and to what extent?
    • If not, then what will the repercussions be?

In the following section, I generate adoption scenarios for human gene-editing by implicitly tuning the legality, availability, and desirability parameters (this I could be more precise about, but for the sake of time, I am relaxed in my descriptions). These scenarios, when taken together, provide a more thorough picture of how human gene-editing practices might unfold in the coming years. The question to ask for each of these components is what information is important and how should it be weighted?.

Scenarios

There are some general principles to take into consideration for each of the following scenarios. In any case, designer babies will either come from the scientific community through experiments, clinical trials, etc… or from the public accessing the technology to do so, either in hospitals (government funded) or privately (e.g., there may simultaneously be treatment practices occurring in hospitals and enhancement practices occurring in lab experiments). In both instances, there is some population22 who desires for their child/children to be gene-edited, either for treatment or enhancement purposes, and there is some subset of this population that might actually have this desire realized someday (the ones who desired it AND had it done). For every jurisdiction considering gene-editing human embryos, a successful, publicized birth as part of a scientific experiment will very likely precede any widespread public adoption. When and where will these first gene-edited babies be born?

  There are four general scenarios that subsume, in my mind, much of the space of possibility for human gene-editing in the next 8 years, i.e. in the interval [2022, 2030). For each scenario, I access the likelihood of the different intervals that I provided earlier, and then weight each interval across the scenarios to generate a final estimate. The details of this will become apparent soon.

Business as Usual: Progress in human somatic and germline gene-editing continues at its current pace. There are programs that lead to reductions in the cost of gene-editing, more accurate genetic targeting techniques, and improved governance of human gene-editing, but applying gene-editing to human embryos remains illegal for a long time. Stances against the gene-editing of human embyros, such as those of the National Institute of Health (NIH) in 2015 and 2018, are relatively common across university, governmental, and clinical research boards.

[2015 statement] However, NIH will not fund any use of gene-editing technologies in human embryos. The concept of altering the human germline in embryos for clinical purposes has been debated over many years from many different perspectives, and has been viewed almost universally as a line that should not be crossed. Advances in technology have given us an elegant new way of carrying out genome editing, but the strong arguments against engaging in this activity remain. These include the serious and unquantifiable safety issues, ethical issues presented by altering the germline in a way that affects the next generation without their consent, and a current lack of compelling medical applications justifying the use of CRISPR/Cas9 in embryos.

[2018 statement] Lest there be any doubt, and as we have stated previously, NIH does not support the use of gene-editing technologies in human embryos.

It’s a toss-up as to whether another researcher similar to He Jiankui performs an unauthorized gene-editing experiment, despite the stiff international backlash that would be garnered by such a deed, but the possibility nevertheless exists. The opinions of researchers and the public alike drifts back and forth between increased and decreased support for human gene-editing, but the trend seems to be that support is gradually increasing, presumably because religiosity is decreasing, gene-editing is becoming more feasible, and governance and regulation is becoming better established. There may even be a trial or two where, in cases of severe heritable diseases, embryos have their germlines genetically edited. Some gene-editing may occur in human embryos, but it is exclusively for treatment. Subjectively, then:

Likelihood (0/100) for 1-32 babies?: 95
Likelihood (0/100) for 33-1000 babies?: 60
Likelihood (0/100) for 1001-31623 babies?: 10
Likelihood (0/100) for 31624-1000000 babies?: 2

You assigned the following probabilities to the intervals:
1-32 : 56.89%
33-1000 : 35.93%
1001-31623 : 5.99%
31624-1000000 : 1.2%

With 90% confidence, you expect the # of gene-edited babies to be between 1 and 923.

Localized Treatment: Interest in human somatic and germline gene-editing grows; over time, this generates more funding and leads to the creation of more genetics research programs than had previously been expected, which in turn leads to more progress in the field of human gene-editing. Developments in programs such as the NIH’s Somatic Cell Genome Editing initiative have the subsidiary consequence of motivating researchers to explore, at a more rapid pace, the ethics and science of human germline gene-editing.

The NIH Common Fund’s Somatic Cell Genome Editing (SCGE) program is working to improve the efficacy and specificity of gene editing approaches to help reduce the burden of common and rare diseases caused by genetic changes. Genome editing technologies present an exciting prospect for treatments and possibly even cures for these diseases. SCGE is developing quality tools to perform and assess effective and safe genome editing in non-reproductive (“somatic”) cells of the body. These research tools will be made widely available to the research community to reduce the time and cost required to develop new therapies.

Several years after this growth, a few technological leaps or a relaxation of governance platforms, perhaps in some developing countries, prompt clinical trials employing germline gene-editing for the treatment of heritable diseases. While the international rebuke still exists, it is not nearly as strong as it was in 2018 after the He Jiankui affair; rather, researchers and the public alike note cautiously that, while some issues with the application of gene-editing in humans seem to remain, it appears promising and viable for the treatment of certain, highly debilitating diseases. As such, in the 1-3 years preceding 2030, some trial treatments get off the ground sooner than some people would have predicted, though the number of gene-edited babies is still relatively low. Subjectively, then:

Likelihood (0/100) for 1-32 babies?: 75
Likelihood (0/100) for 33-1000 babies?: 90
Likelihood (0/100) for 1001-31623 babies?: 90
Likelihood (0/100) for 31624-1000000 babies?: 2

You assigned the following probabilities to the intervals:
1-32 : 29.18%
33-1000 : 35.02%
1001-31623 : 35.02%
31624-1000000 : 0.78%

With 90% confidence, you expect the # of gene-edited babies to be between 1 and 23558.

Widespread Treatment: This is a variant of Localized Treatment where adoption is, as the scenario title goes, more widespread. This is to say that we underestimated how many unique jurisdictions adopted using or trialing human gene-editing.

Likelihood (0/100) for 1-32 babies?: 50
Likelihood (0/100) for 33-1000 babies?: 85
Likelihood (0/100) for 1001-31623 babies?: 99
Likelihood (0/100) for 31624-1000000 babies?: 70

You assigned the following probabilities to the intervals:
1-32 : 16.45%
33-1000 : 27.96%
1001-31623 : 32.57%
31624-1000000 : 23.03%

With 90% confidence, you expect the # of gene-edited babies to be between 1 and 579091.

Breakthrough: Evolutionary progress is difficult to predict, but with regard to genetically engineering human embryos, it might be the case that such progress befalls this field sometime in [2022, 2030). Perhaps something that is as of yet unpublished, but which might forward gene-editing drastically, increases humanity’s support for germline genome editing. I suspect that, given the existing medical infrastructure globally, such a breakthrough would still take several years to achieve extensive clinical adoption. Subjectively, then:

Likelihood (0/100) for 1-32 babies?: 10
Likelihood (0/100) for 33-1000 babies?: 15
Likelihood (0/100) for 1001-31623 babies?: 65
Likelihood (0/100) for 31624-1000000 babies?: 85

You assigned the following probabilities to the intervals:
1-32 : 5.71%
33-1000 : 27.96%
1001-31623 : 37.14%
31624-1000000 : 48.57%

With 90% confidence, you expect the # of gene-edited babies to be between 1 and 800818.

Updated Forecast

Taking these scenarios together, I would assign a likelihood of 90 to the Business as Usual scenario (this scenario seems most plausible, given that this scenario seems to hold for other scientific enterprises involving emerging and ethically trying technologies), a likelihood of 60 to the Localized Treatment scenario (I imagine this would occur in fewer realities than the first scenario, as much would need to be done in a relatively short amount of time for this scenario to occur), a likelihood of 35 to the Widespread Treatment scenario (as even more would need to occur than in scenario two, which is even less likely to occur), and a likelihood of 5 to the Breakthrough scenario (as these don’t occur very often). So, this would mean that I given the following weights to each of my past estimates:

Likelihood (0/100) for Business as Usual?: 90
Likelihood (0/100) for Localized Treatment?: 60
Likelihood (0/100) for Widespread Treatment?: 35
Likelihood (0/100) for Breakthrough?: 5

You assigned the following weights:
Business as Usual : 47.37%
Localized Treatment : 31.58%
Widespread Treatment : 18.42%
Breakthrough : 2.63%

Thus, a final SPIES estimate would be:

Likelihood (0/100) for 1-32 babies?: (95 x 0.4737) + (75 x 0.3158) + (50 x 0.1842) + (10 x 0.0263) = 78.1595 = ~78
Likelihood (0/100) for 33-1000 babies?: (60 x 0.4737) + (90 x 0.3158) + (85 x 0.1842) + (15 x 0.0263) = 72.8955 = ~73
Likelihood (0/100) for 1001-31623 babies?: (10 x 0.4737) + (90 x 0.3158) + (99 x 0.1842) + (65 x 0.0263) = 53.1043 = ~53
Likelihood (0/100) for 31624-1000000 babies?: (2 x 0.4737) + (2 x 0.3158) + (70 x 0.1842) + (85 x 0.0263) = 16.7085 = ~17

You assigned the following probabilities to the intervals:
1-32 : 35.29%
33-1000 : 33.03%
1001-31623 : 23.98%
31624-1000000 : 7.69%

So, with 90% confidence, I expect the # of gene-edited babies by 2030 to be between 1 and 28683. While I’m not sure how I can exactly take this statement and convert it into a Metaculus prediction, but here is my attempt. I use one component per interval. The weighting of the Metaculus components roughly corresponds to the likelihoods for the different intervals (e.g., I will try to move the slider to the ~78% mark for the 1-32 interval). My confidence intervals are non-symmetric; I haven’t accounted for this in the analysis in this post, and follow my intuition - “where should most of the mass be for this interval”.

For comparison, here is Eli Lifland’s forecast on the question along with his reasoning:

My prediction: 5.7 – 8.7k

Given 3 in the last three years, simple linear extrapolation would put a median at 12. But the median should probably be higher than this. However, I’m not sure how publicized gene-edited babies will be, even if many more gene-edited babies are born.

According to this article, human genome-editing is largely forbidden.

This one feels like it should have a peak at 3 but a long right tail; a decent chance this remains at 3. but also plenty of probabilities on higher numbers if a large advanced country lifts legal restrictions to some extent. My distribution is pictured here.

Adjustments

In this section, I will provide adjustments to my forecasts as new and existing information becomes known to me. I expect to do this every year until 2030.

Retrospective

In this section, I will examine the actual value resolve value, i.e. the number of designer babies born by 2030, and will discuss the manners in which I was incorrect.


Appendix

Armsby et al. Survey Figures

Demographics

Opinions on the Acceptability of Human Gene-Editing

Opinions on the Acceptability of Various Uses of Human Gene-Editing

Approval by Nation for Gene-Editing (PEW)

(these are currently unordered and lack descriptions; my apologies - I will try to fix this in the near-term future)


Page Epistemics

Status: “Working Draft” I don’t believe I will make many more edits to the content of this page in the coming years. There are some adjustments and updates to make to my forecasts, along with the final retrospective to perform in 2030…I hope I am not dead! I will likely only alter the web-infrastructure of this page (e.g., collecting all URLs on the page to generate an ordered link bibliography). Outside of alterations, what I have stands as a working attempt to forecast how many gene-edited humans there will be by the years 2029, 2049, and 2100.

Certainty: “Somewhat Likely” This means I am around 60% confident in the accuracy or thoroughness of my content. This post is no where close to a literature review, and I still lack a deep understanding of molecular biology and of the actual technology for gene-editing humans. My coverage of the policy surrounding gene-editing in humans could have benefitted from more examples. I doubt the actual number of gene-edited children will fall outside my 90% confidence interval, but am somewhat uncertain about my likelihoods for the widespread treatment and breakthrough scenarios. Did I down-weight these too much? Am I too optimistic - will there simply be less than 10 gene-edited babies? Did I not weigh the 1-32 interval enough? That such musing plague me indicates to me that I still have some reservations about the “completeness” of this forecast.

Importance: “6.5/10” I think human gene-editing can be a transformative technology for social good. Imagining people being free from painful and debilitating disease, being physical capable, having a robust immune system, being very cognitively able, etc… is a tantalizing prospect. John von Neumann’s intellect pops into my mind from time to time when I think about human gene-editing. With better genes, humanity could do so much more; nevertheless, there is the potential for things to go very, very wrong.

Impact: “3.5/10” I think some of my friends might read this and appreciate the current state of human gene-editing research slightly more. Since this post is part of a contest, there is a slightly heightened potential for increased readership.


Notes

Cover Images

The cover photo for this page was likely taken by Birmingham Museums Trust. I found the photo on Unsplash. To my knowledge, my use of this photo is permissible under Unsplash’s license: “Unsplash grants you an irrevocable, nonexclusive, worldwide copyright license to download, copy, modify, distribute, perform, and use photos from Unsplash for free, including for commercial purposes, without permission from or attributing the photographer or Unsplash. This license does not include the right to compile photos from Unsplash to replicate a similar or competing service.

Footnotes

  1. For Henk ten Hank’s personal website, see here, and for Maria do Céu Patrão Neves’s faculty webpage (I was unable to find her personal site after looking for a few minutes, which makes me believe she doesn’t have one) see here

  2. Every source accessed and all personal takes are accurate at least up to the date of last edit (the second date beneath the tags; this date is currently ). 

  3. What is egg freezing, and why is it used? Here is one source’s description of egg freezing - UCLA’s Fertility and Reproduction Health Center’s (see here) definition:

    Egg freezing, or oocyte cryopreservation, is a process in which a woman’s eggs (oocytes) are extracted, frozen and stored as a method to preserve reproductive potential in women of reproductive age. The first human birth from a frozen oocyte was reported in 1986. Oocyte cryopreservation has advanced greatly over the past few years, with improved overall success of eggs surviving the freezing process. It is no longer considered an experimental procedure by the American Society for Reproductive Medicine. The techniques leading to enhanced gamete survival, potential fertilization and live birth rates allow women a much greater degree of autonomy than was possible even in the past 5 years.” 

  4. What is sperm freezing, and why is it used? Here is one source’s description of sperm freezing - the University of Utah’s Health Care Center’s definition:

    Sperm cryopreservation or sperm freezing is a way for men to preserve their sperm and store it in a bank so it can be used in the future. Many medical treatments can damage sperm quality, including several types of cancer treatment like chemotherapy and radiation. Some men choose to freeze their sperm before getting medical treatment.” 

  5. From Wikipedia’s in-vitro fertilization page:

    In vitro fertilisation (IVF) is a process of fertilisation where an egg is combined with sperm in vitro (“in glass”). The process involves monitoring and stimulating a woman’s ovulatory process, removing an ovum or ova (egg or eggs) from their ovaries and letting sperm fertilise them in a culture medium in a laboratory. After the fertilised egg (zygote) undergoes embryo culture for 2–6 days, it is implanted in a uterus, with the intention of establishing a successful pregnancy. 

  6. For descriptions of polygenic or embryo screening, check out the National Human Genome Research Institute’s article on polygenic risk scores, Wikipedia’s page on polygenic scores, Penn Medicine’s review of embryo screening procedures, IVI’s (“the largest Assisted Reproduction group in the world”) article What is embryo screening and is it right for me?, and Wikipedia’s page on preimplantation_genetic_diagnosis

  7. The National Human Genome Research Institute, human genome editing describes human genome editing as

    …a method that lets scientists change the DNA of many organisms, including plants, bacteria, and animals. Editing DNA can lead to changes in physical traits, like eye color, and disease risk. Scientists use different technologies to do this.” 

  8. There are many qualities that I would like my children to have. Two things concerning this should be noted: 1) the term “qualities” is vague, and the things people routinely associate with them (e.g., intelligence or IQ, OCEAN measures, physical attractiveness, creativity, the absence of disease, the absence of a cognitive ailment, etc…) are typically broad in scope, complex (stemming from the interplay of multiple genes), and lack epistemic consensus (multiple competing measures or definitions for the phenomenon), so I believe we should presently be weary or cautious of optimizing for these qualities, given that humanity’s lack of understanding of the biology underlying these things may result in unintended harmful consequences, and 2) my desire for my children to have certain “qualities” is illustrative of the idea that I believe there are genetic traits that are more instrumentally valuable to goals I personally like or that I believe are more “worthy”, which is to say that my selection of which traits I would want to have enhanced in my children is highly biased. I’m fairly certain other people would differ in their trait selections. For example, someone who is as ambitious about creating beautiful art or music as I am about research might be interested in genetically optimizing traits positively correlated with artistic performance, but might not care as much about other traits that would make their children good researchers. These two points together demonstrate that my partner and I still have many considerations to account for in deciding whether to use gene-editing on our children, should we have any. 

  9. My partner is particularly concerned that the potential complications of batch egg-extraction might not be worth it, should we decide to incorporate embryo screening or gene-editing into our reproductive activities. From the “Complications” section of the Wikipedia’s page on transvaginal oocyte retrival:

    “Injection of hCG as a trigger for ovulation confers a risk of ovarian hyperstimulation syndrome, especially in women with polycystic ovary syndrome who have been hyperstimulated during previous assisted reproduction cycles.

    Complications of TVOR include injury to pelvic organs, hemorrhage, and infection. Occurring more often in lean patients with polycystic ovary syndrome, ovarian hemorrhage after TVOR is a potentially catastrophic and not so rare complication. Additional complications may result from the administration of intravenous sedation or general anesthesia. These include asphyxia caused by airway obstruction, apnea, hypotension, and pulmonary aspiration of stomach contents.

    Propofol-based anesthetic techniques result in significant concentrations of propofol in follicular fluid. As propofol has been shown to have deleterious effects on oocyte fertilization (in a mouse model), some authors have suggested that the dose of propofol administered during anesthesia should be limited, and also that the retrieved oocytes should be washed free of propofol. Anecdotal evidence suggests that certain airborne chemical contaminants and particles, especially volatile organic compounds (VOC), may be toxic to and impair the growth and development of embryos if present in sufficient concentrations in the ambient atmosphere of an IVF incubator.

    Endometriosis seems to cause a challenge for TVOR that may have reflection on individual surgeon’s performance rates for the procedure, independently from the diameter of a pre-existing ovarian endometrioma (OMA) or ovarian adhesions. Obesity is another factor that may present a challenge for the procedure.”
      2

  10. I’m unfamiliar with the present consensus on the “nature vs. nurture” debate, and lack a general understanding of heritability, especially with regard to the heritability of intelligence

  11. I don’t cover people’s attitudes towards human gene-editing in detail, but intend to at some point in a Metaculus notebook with the prospective title - Attitudes Towards Human Gene-Editing. When/if I write this, I will link to it in this footnote. I expect to develop forecasting questions on what people’s attitudes will be in the near-term future for enhancement versus treatment uses of human gene-editing and on how the connection between religion and aversion towards human gene-editing will change. 

  12. The PEW Research survey’s explanation for How we did this?:

    This report examines public perceptions of biotechnology, evolution and the relationship between science and religion. Data in this report come from a survey conducted in 20 publics from October 2019 to March 2020 across Europe, Russia, the Americas and the Asia-Pacific region. Surveys were conducted by face-to-face interview in Russia, Poland, the Czech Republic, India and Brazil. In all other places, the surveys were conducted by telephone. All surveys were conducted with representative samples of adults ages 18 and older in each survey public.” 

  13. The 20 publics: Australia, Brazil, Canada, Czech Republic, France, Germany, India, Italy, Japan, Malaysia, Netherlands, Poland, Russia, Singapore, South Korea, Spain, Sweden, Taiwan, United Kingdom, United States. For each public, PEW reports that there were 100 respondents per question, but I do not know whether this means that there are 100 total respondents per public, or that there are 100 x n total respondents, where n is the number of questions. For the survey’s methodology, see here

  14. Within the field of human enhancement, a distinction is often made between gene-editing for enhancement versus for treatment. From what I’ve seen, treatment generally refers to using gene-editing to remove diseases or other malignities, whereas enhancement refers to improving some trait or property of the child (e.g., treatment would be editings genes associated with an increased likelihood of breast cancer, whereas an example of enhancement might be editing genes to make a person taller than they would be otherwise.) 

  15. Vera Lucia Raposo, in her paper 2019 paper The First Chinese Edited Babies: A Leap of Faith in Science, provides a nice section on stances taken against human gene-editing. In summary, they include (1) sanctity of genome, (2) genetic discrimination and eugenics, (3) undermining human genetic pool, and (4) loss of human nature.

    Ethical concerns have long been asserted against genetic interventions. However, most of the objections have been based more on prejudice than substantive arguments. Critics have invoked the sanctity of the human genome, as if changing it would equate to playing God (Habermas, 2003). However, protecting the human genome should not prevent genetic interventions that can improve our lives. What brings real value to our lives is having a genetic code that allows us to live free of severe diseases, not to have an unmodified but unhealthy genetic code. Some have argued the perils of genetic discrimination (Mehlman & Botkin, 1998) and eugenics (Habermas, 2003), but if that were truth no medical treatment would be allowed under the suspicion of discriminating the ones not that are not treated and of aspiring to create a “superior” society of healthy people. The risk of undermining the human genetic pool (Committee on Science, Technology, and Law, 2016) is also a recurrent concern, but “there are more than six billion humans on the planet. Absent some kind of magic wand, it is initially difficult to see how any given genetic intervention could change human nature” (McConnell, 2010). The eventual loss of our human nature (Habermas, 2003) has been also invoked, but changing our genes does not change our human nature. Humanity does not reside in a specific genetic code, but in a certain perception of the world and our role in it. That role adds to the story of how we overcome the surrounding environment and ourselves.” 

  16. After a quick first-pass search, I was not able to find any other surveys on the opinions of bioethicists or people with genetics training towards human gene-editing. When writing this, I could have refrained from including the WHO’s and NHGRI’s stances on human germline genome editing, but after failing to find the opinion polls I was looking for, I decided that they might be useful for presenting some “official” attitudes towards governance of human gene-editing. Research and development in human gene-edited is something that I would expect to be greatly affected by the ethics committees of organizations 

  17. Note that ELSI quite certainly stands for “Ethical, Legal and Social Implications” (see here). 

  18. From the Chapter 4 of Human Genome Editing: Science, Ethics, and Governance (2017):

    The use of human genome editing to make edits in somatic cells for purposes of treating genetically inherited diseases is already in clinical trials. Somatic cells contribute to the various tissues of the body but not to the germline, meaning that, in contrast with heritable germline editing (discussed in Chapter 5), the effects of changes made to somatic cells are limited to the treated individual and would not be inherited by future generations. Morever, germline gene-editing is supported much less than somatic genome gene-editing, even for treatment.” 

  19. I follow some of the principles outlined in Rob J Hyndman and George Athanasopoulos’s chapter on judgemental forecasting in their excellent book Forecasting: Principles and Practice, 2nd ed. (this is just a pointer to advertise their book; I’ve found it to be a nice and useful read). 

  20. The existence of this post is partially due elifland, sam_atis, and yagudin’s Impactful Forecasting Contest. They ranked 50 Metaculus questions by their impact, on the following scales:

    Decision importance: The importance of the decisions which will be affected by this question. Should combine cause area importance + importance within cause area. Decision relevance: How much of an impact would this have on actual decisions if the forecast changed by a substantial amount? This factor is re-used from Nuño Sempere’s An estimate of the value of Metaculus questions. Ease of contribution: ​​How easy will it be for a “median generalist forecaster” to make a contribution to the analysis on this question within a few hours? e.g. questions requiring lots of domain expertise or background reading would score low here. Neglectedness of contributions: How few contributions have there been on this specific question so far? How in need of attention is it? This should be subjectively evaluated using the existing count of forecasts and quantity + quality of comments/writeups.

    A curation score was calculated, weighing decision importance at twice the other three due to it feeling like the most important factor. We chose a set of 25 questions based mainly on the curation score, but also including a diversity of cause areas and question types.

    The question that this post focuses on - How many gene-edited babies will there be by 2029? - was given a score of 17.

    Decision Importance: 3. Since He Jiankui went rogue (assuming nothing weird was going on with CCP) this has become a v. interesting Q. Important path to reduce suffering (some diseases like sickle cell will be easy to eliminate with CRISPR). Ease of contribution: 2. Fairly tricky to forecast on with no domain-specific expertise but developments here do get media coverage. Neglectedness: 3. Think there hasn’t been enough discussion paid to gene-editing so fairly neglected - but decent number of forecasts, just the comments aren’t *that* enlightening (yet).” 

  21. This is the SPIES (Subjective Probability Interval Estimates) method for judgemental forecasting; for a working example in Javascript, check out here.

    This work introduces a novel method for reducing overprecision in estimates. This method, called Subjective Probability Interval Estimates (SPIES), does not directly elicit a confidence interval. Rather, it presents the judge with the entire range of values, divided into 11 intervals of equal width. These intervals can span the entire range, or, in case the value scale does not have pre-determined high and low bounds, a range that includes all plausible values, with an additional interval at each end representing all extreme values which lie outside this plausible range. The judge then estimates a probability for each interval. In case of estimating one true value, this probability is the likelihood that the interval includes the correct answer. For estimates of a population’s properties, this probability is the proportion of the population that is included in the interval. Since the SPIES range includes all possible values, the sum of these probabilities is constrained to equal exactly 100% (see Figure 1 for an illustrative example). The judge’s output, then, is a series of subjective probabilities that total 100%. These subjective probabilities can be computed for numerous types of estimates. In addition to the estimated distribution, it is possible to calculate confidence intervals of virtually any width and confidence level, by combining the SPIES’ intervals within the SPIES task. For example, from the estimate of a future temperature in Pittsburgh, as in Figure 1, it is possible to calculate the most likely 10-degree and 20-degree intervals, as well as the judge’s 70% and 90% confidence intervals, all without having to elicit the judgment from the judge multiple times. The SPIES method, then, offers great versatility and flexibility to the recipient of the estimate. In this paper I make the argument, and present data to support it, that SPIES can significantly reduce overprecision in two ways. As an elicitation method, SPIES forces the judge to consider all possible values, including ones that often go ignored in the estimation process of other, more instantiated methods. This enables judges to produce confidence intervals of greater width and better calibration. As an intervention for reducing bias, SPIES influences subsequent estimates in other elicitation formats, by inducing judges to revise their estimation process. The remainder of this dissertation will be organized in two parts, each presenting data from three 12 laboratory experiments. Part I will focus on SPIES as an elicitation method. It will present a comparison between estimates made using this method and estimates made using other methods and tests of the robustness of this difference. In Part II, I will explore how making an estimate with the SPIES method can improve the calibration of subsequent confidence interval estimates. Finally, I will discuss implications, theory extensions and possible applications of the SPIES method, within and outside the realm of cognitive research.” 

  22. My original intention was to write this paragraph following the current one this footnote is in:

    “The pool of people who desire to have their children gene-edited changes over time, likely due to such things as cultural and generational changes, progress in the safety and effectives of gene-editing, and policy on gene-editing, among other things. To approximate the present size of this population, I refer back to the PEW research survey in 2020 of 20 publics. There are some immediate problems with this approach - the study only covers 20 nations, these 20 nations are relatively “developed”, so My thinking is this: establish how many people might want to use gene-editing for treatment and/or enhancement, and then postulate how different levels of availability, legality, and scientific progress might affect this number over time.”

    I would like to have used the populations, religiosity, approvals, and human development indices of the countries listed to find the correlation between approval, religiosity, and human development index to estimate a median approval for the rest of the world’s population. Then, I would examine how many members of the two subsets of the global population that approve gene-editing for treatment and approve gene-editing for intelligence would be inclined to actually seek out and use gene-editing if it were various degress of legal, available, and desirable.