This is part 3 of a series of 7 posts by Lars Johan Erkell, with comments on each by Ola Hössjer and a reply by Erkell. Part 1 will be found here. They are translations of 2020 posts in Swedish from the Biolog(g) blog of the Department of Biology of Gothenburg University.
Breakthrough for Intelligent design? (part 3)
November 13, 2020
by Lars Johan Erkell (with comment by Ola Hössjer)
Miracles as science?
Thorvaldsen and Hössjer give a number of examples of systems that they think are suitable for the statistical model we discussed in the previous post. One of the examples concerns human evolution.
One may also study the more restricted problem of human/chimp ancestry, and compare a model M2 with common ancestry of the two species, with a unique origin model M1, according to which each species is founded by one single couple (Sanford and Carter, 2014; Hössjer et al., 2016a; 2016b, Carter et al., 2018, Hössjer and Gauger, 2019).
Since Thorvaldsen and Hössjer raise this question as suitable to study with their statistical model, we shall look at it more closely.
The question they ask is whether it is possible that humanity has its origins in a single couple, a man and a woman. It is of course about Adam and Eve, even if they are not mentioned by name - otherwise there is hardly any reason to ask the question. The common scientific view is that humans did not originate in two individuals, but that our species passed a bottleneck perhaps 70,000 years ago, when there were perhaps 10,000 individuals. One can of course imagine that the bottleneck was only two individuals, but to be able to explain the genetic variation we see today, we then have to extend the time to 500,000 years or more. This is not acceptable for a young earth creationist scenario. There, creation must have taken place 6,000 to 10,000 years ago if you want to read the Bible literally.
Ola Hössjer is a co-author in three of the referenced articles: “Genetic Modeling of Human History Part 1: Comparison of Common Descent and Unique Origin Approaches” and “Genetic Modeling of Human History Part 2: A Unique Origin Algorithm”, both from 2016. Here he reports that he, along with two co-authors, developed a population genetic model to simulate human descent. Together with biologist Ann Gauger (who is affiliated with the Discovery Institute), he has also written “A Single-Couple Human Origin is Possible” from 2019, where they developed their earlier model. They write that an origin from a single couple 500,000 years ago is consistent with the model, but that only small changes in the model can give an age of 100,000 years or less.
The three articles are published in the online journal BIO-complexity, published by the Discovery Institute, the central organization of the ID movement. The journal is not recognized as a “real” scientific journal. For example, it is not included in the scientific database SCOPUS.
In “Genetic Modeling of Human History Part 1” we read on the first page:
We argue that a unique origin model where humanity arose from one single couple with created diversity seems to explain the data at least as well, if not better.
Ola Hössjer and his co-authors thus say that their model that humans originated in a single couple matches the data better than the standard evolutionary model. But what data? All you find is qualitative reasoning about published data of various kinds. Neither of the two articles with the title “Genetic Modeling of Human History” contains any "Genetic Modeling"; no quantitative data, no results from simulations with the model developed. The approach is recognizable from Thorvaldsen and Hössjer's current article (see the first post in the series). Here the authors also pretend to have results they don't have.
The quote also mentions "created diversity". Now what is that? I have never heard of such a thing before. The text discusses five different mechanisms for how genetic variation can develop and be limited. But then it introduces a sixth possibility. The authors write on page 5:
… But if an intelligent designer is invoked as a possible explanation, and if humanity originates from one single couple, it is possible that their chromosomes were created with considerable diversity from the beginning [19,20]. This gives us a sixth mechanism of genetic change.
VI. Created founder diversity is biologically plausible for DNA of non-sex chromosomes. Since there are four copies of each non-sex chromosome in two individuals, we may think of one of the founding male’s two copies as a reference or template for the other three. All differences between his reference chromosome and the other three founder chromosomes can be thought of as a very large number of mutations, all of which occurred in one generation. Since the founding pair had three copies of the X-chromosome (one in the man and two in the woman), for the same reason they may also have been different. Some founder diversity is possible for mitochondrial DNA as well. Since the founding female carried hundreds of mitochondria that could have been diverse, it is possible that she passed on some of that diversity to her daughters.
The genome contains two copies of each gene. The idea is that Adam and Eve would have been created with two different variants of each gene in their DNA, which together would give four variants. That would increase genetic variation so much that one could imagine a creation 6,000 to 10,000 years ago. References 19 and 20 are not to the scientific literature, but to two creationist journals. The first of the two articles does not justify why introducing an engineered genetic variation would be scientifically justified. However, the other one, “On the Origin of Eukaryotic Species' Genotypic and Phenotypic Diversity: Genetic Clocks, Population Growth Curves, and Comparative Nuclear Genome Analyzes Suggest Created Heterozygosity in Combination with Natural Processes as a Major Mechanism”, by Nathaniel Jeanson and Jason Lisle, does. They write on p. 86-87:
No known non-miraculous mechanism exists that could play a role in this process, and invoking miracles at every juncture without biblical or scientific justification quickly moves a hypothesis from the realm of science to the realm of ad hoc speculation.
… On an individual organism level, hypotheses must invoke observable processes or biblically-justified miraculous processes. … Again, observable processes or biblically-justified miraculous processes must be invoked, or the hypothesis quickly drifts into the realm of the ad hoc.
A hypothesis could thus be justified either scientifically or with the support of the Bible. However, I find no discussion of exactly how the Bible can justify God creating Adam and Eve with maximum genetic variation. Ola Hössjer and co-authors thus refer to a created variation that is not justified either scientifically or biblically.
But surely this is a convenient explanation. Invoking a designer suddenly makes it possible that genetic variation may be greater than established science admits, and that Adam and Eve may have lived 6,000 to 10,000 years ago. It is not seen as necessary to show that there really is a designer or that the designer really had the desire to create exactly the genetic variation that is needed.
There are problems with this model. Without going into the genetics, one can, for example, state that in this model there can be a maximum of four gene variants - alleles - of a certain gene. However, there are gene families with thousands of alleles. Where do they come from? But once you have started to explain things with miracles, you can surely solve those problems as well.
And where does the designer come from? Out of thin air. A designer has been invented that creates genetic variation to solve the problem that Adam and Eve must not have lived hundreds of thousands of years ago. And thus they have completely left science and floated out into the world of fantasy.
If Ola Hössjer was faced with a new statistical model that was based on an unproven theorem, I believe he would react sharply. Theorems must be proven, otherwise you cannot know that they are true. If there is any part in the derivation of the model that is unproven, it cannot be used. Grabbing something out of thin air – a theorem or something else – is called ad hoc assumption and is not considered good scientific practice.
If you can't get your theory together, you can't just pull something to save it. All parts of a mathematical derivation or a biological theory must be based on knowledge, not on grabbing something out of thin air when needed.
Nevertheless, this is exactly what Ola Hössjer and co-authors are doing. The introduction of a supernatural designer makes their work lose all scientific credibility and thus all scientific interest.
Let me add that in principle one can consider a designer as a scientific explanation, provided it is known for sure that the designer exists, and has such properties that they can explain a certain phenomenon. If we have actual knowledge about the designer (as we have about the five genetic mechanisms discussed) we can use that knowledge to explain things. But in this case we have no knowledge of the designer at all. And then we cannot use design as an explanation. at all. And then we cannot use design as an explanation.
Comment by Ola Hössjer June 29, 2021 3:19 am
Part 3: Miracles and Science
In Part 3, Erkell makes a detour and comments on a number of articles on the origin of man that I wrote in collaboration with Ann Gauger and Colin Reeves3. In these articles we use the design model M1 that was described in Part 2 in a creationist context. More precisely, our model M1 assumes that all humans are descended from a first couple, Adam and Eve. The model thus includes the aforementioned microevolutionary mechanisms (i)-(v), as well as Adam and Eve being created with genetic variation (vi). In other words, according to this model, Adam has two genetically different copies of each non-sex chromosome, and so does Eve. Thus, in the first founder generation we get four genetically different variants of each non-sex chromosome. The degree of genetic dissimilarity between these variants, i.e. the degree of created variation, is a parameter that can be varied in the model.
Erkell calls our creationist model M1 a miracle model (which, by the way, I think is a good name)4), and he has similar high demands on this model as he had on the design model in the previous section:
“A hypothesis could thus be justified either scientifically or with the support of the Bible. However, I find no discussion of exactly how the Bible can justify God creating Adam and Eve with maximum genetic variation. Ola Hössjer and co-authors thus refer to a created variation that is not justified either scientifically or biblically.”
To support his thesis, Erkell starts from the quote “On an individual organism level, hypotheses must invoke observable processes or biblically-justified miraculous processes” by the creationists Jeanson and Lisle.5 I support this formulation, if the purpose is to choose between a creationist model M1 and a naturalistic model M2. But the fact that a hypothesis can be justified based on the Bible does not mean that it is a direct quote from a Bible verse. Erkell has simply misunderstood how creationist research is conducted. In fact, it is perfectly possible to let the Bible inspire the formulation of falsifiable hypotheses. In other words, creationism can be reconciled with the hypothetical deductive method, regardless of whether we start from direct biblical quotes or simply are inspired by the contents of the Bible, according to the following three steps:
Propose a falsifiable hypothesis that partly contains natural explanatory mechanisms and partly explanatory mechanisms that are inspired by the Bible.
Evaluate and check if the hypothesis in (1) is consistent with data.
If the agreement in (2) is not good, go back to (1) and propose a better hypothesis.
As a Christian, I believe that God wrote two books, the Bible and nature. Therefore, it is ultimately God who is behind what we call “natural explanations” as well as the “biblically motivated explanations”. Of course, one must place the same demands on a creationist model M1 as on a secular model M2. Erkell argues that our proposed creationist miracle model is firstly ad hoc and secondly not falsifiable. I believe that both of these statements are completely wrong. The Bible states that man was created uniquely. Given this, I mean that it is not at all a far-fetched ad hoc assumption to assume that God created Adam and Eve with genetic variation, because He is creative and did not want all men and women to look alike, like monozygotic twins. By creating the first two humans with variation, the risk of inbreeding depression was also reduced when harmful mutations made their way into future generations.
Note, however, that we only include created variation in our model, not why God created the first two humans with variation. Given this, the model is falsifiable, because then, due to recombinations (iv), our DNA is a mosaic of the four building blocks from each chromosome, with which Adam and Eve were created. Such building blocks, called haplotype blocks, were discovered by secular scientists about 20 years ago. We mention this in two of our articles, with numerous references.6 Either Erkell has missed these references or he has not realized that the existence of haplotype blocks makes our miracle model falsifiable. And regardless of whether the haplotype blocks are identified or not, I still believe that our miracle model is falsifiable, since there are indirect methods of testing the presence of created variation by comparing the allele frequency spectra and linkage disequilibrium plots calculated from the data with those predicted by the miracle model M1. This latter approach exactly the one we use in one of our publications to test M1 against data.7 However, Erkell (in Part 5 of his article series) objects to our way of fitting the miracle model to the data:
“Then you add an unknown designer as a sixth mechanism. Then all opportunities to conduct a fact-based discussion disappear; we have no idea when or how such a designer would intervene. Hössjer and his colleagues seem to assume that the designer would create just enough genetic variation that their model fits. How can they know that? How can they test their model? It’s not possible. You have simply invented a designer who has the qualities you want.”
Here Erkell confuses two things, firstly, whether the miracle model M1 is falsifiable or not, which I have already commented on, and secondly, how the adaptation of the model to data has been done. All statistical models contain unknown parameters, so does our miracle model. This is precisely why the data is used to estimate the degree of created variation. Our approach is very similar to an evolutionary biologist who fits a naturalistic model M2 to data by estimating mutation rates, population sizes, and recombination rates. Of course, it is possible to overfit if one uses a model with too many parameters, without validating it. But in my joint article with Ann Gauger, on the contrary, we used a very simple model with created variation, but without geographic variation (v). We also managed to get a good fit of our miracle model to data, if Adam and Eve lived several hundred thousand years ago. However, we write that with a larger (and more realistic) variant of the miracle model, the age of the first two people can probably be lowered significantly. In addition, it is perfectly possible to test the miracle model by first estimating parameters on a data set (a so-called training of the model) and then evaluating how well the obtained model adapts to another data set (a so-called validation of the model).
Erkell finally delivers a third critique of the miracle model:
“There are problems with this model. Without going into the genetics, one can for example state that in this model there can be a maximum of four gene variants - alleles - of a certain gene. However, there are gene families with thousands of alleles. Where do they come from? But if you start to explain things with miracles, you can certainly solve those problems too.”
This is a rather astonishing statement from a biologist. In fact, it is perfectly possible to explain gene families with a large number of alleles in a model where Adam and Eve were created with variation. Although there were only four raw copies of each non-sex chromosome in the first generation, mutations (i) and recombination (iv) in future generations can generate a large number of variants of different genes.
Erkell concludes, in Part 3 of his article series, by asking how I, as a mathematician, can allow models with ad hoc assumptions:
“f Ola Hössjer was faced with a new statistical model that was based on an unproven theorem, I believe he would react sharply. Theorems must be proven, otherwise you cannot know that they are true. If there is any part in the derivation of the model that is unproven, it cannot be used. Grabbing something out of thin air – a theorem or something else – is called ad hoc assumption and is not considered good scientific practice.”
There is much to say about these sentences. First, Erkell confuses mathematics (a deductive science) with biology (an empirical or inductive science). In mathematics, one starts from axioms (assumptions of faith) and proves, given these axioms, various statements. Here, the demands on accuracy are high, because a single small error can overturn an entire proof. However, there are no requirements for a theorem of mathematics to correspond to reality (despite this, mathematics has nevertheless proven to be very useful). In biology, the aim is instead to adapt models to empirical data, using the hypothetico-deductive method (and here mathematics comes in as a tool). It is therefore never possible to prove that a biological theory is true. On the other hand, if models that do not explain data well are falsified we may be left with plausible models that so far have survived all tests. Of course, even in biology you have to be strict when proposing and testing models against data, but as I have shown above, our miracle model is firstly not ad hoc and secondly it is falsifiable.
Hössjer, O., Gauger A. and Reeves, C. (2016a). Genetic modeling of human history part 1: Comparison of common descent and unique origin approaches. BIO-Complexity 2016(3):1-15. Hössjer, O., Gauger A. and Reeves, C. (2016b). Genetic modeling of human history part 2: A unique origin algorithm. BIO-Complexity 2016(4):1-36. Hössjer, O. and Gauger, A. (2019). A single couple of human origin is possible. BIO-Complexity 2019(1):1-21.
Our sixth mechanism for genetic change, created variation (iv), assumes that this variation is part of God’s creation. However, miracles are generally defined as exceptions to the laws of nature that God has created already, see for example Keener, C.S. (2011), Miracles, the credibility of the New Testament accounts, Baker Academics, Grand Rapids, Hössjer, O. (2018). Becoming a Christian, Combining prior belief, evidence, and will. Wipf and Stock Publishers, Eugene, OR, and Grenholm (2019), Documented Miracles, Sjöberg Publishers. Alternatively, all of God’s creation can be seen as a miracle. For these reasons I don’t mind using the name miracle model for our creationist explanatory model of human genetic variation, as suggested by Erkell. 5) Jeanson, N.T., and Lisle, J. (2016). On the origin of eukaryotic species’ genotypic and phenotypic diversity: Genetic clocks, population growth curves, and comparative nuclear genome analyzes suggest created heterozygosity in combination with natural processes as a major source mechanism. Answ Res J 9:81-122.
See section 3C of Hössjer et al. (2016a) and section 1.3 of Hössjer et al. (2016b). Note, however, that created variation can be defined in different ways. For example, all of Eve’s egg cells, which were with her from the beginning, may have been created differently. In our articles (see Footnote 3) we have chosen to make the model of created variation as simple as possible, assuming that all of Eve’s eggs were genetically identical. However, one can think of different (falsifiable) models for created variation.
Hössjer, O. and Gauger, A. (2019).
Reply by Lars Johan Erkell September 8, 2021 6:17 am
In this section we return to the question of how to formulate scientific hypotheses. I refer to my discussion in the post “Science, hypotheses and Intelligent design” https://biologg.wordpress.com/2021/09/08/vetenskap-hypoteser-och-intelligent-design/.
So you find it astonishing that I, as a biologist, do not agree that today’s genetic diversity could be explained on the basis that Adam and Eve must have lived 6000 years ago. But I’m not the only one who thinks so. I assume that you know that Steve Schaffner and Joshua Swamidass, with their respective allele frequency spectra and recombination analysis, concluded that from a genetic point of view it is indeed quite possible that Adam and Eve could have existed, but that they must have lived at least 500,000 years ago (see eg http://richardbuggs.com/2021/03/15/video-adam-eve-and-human-genetic-diversity/). In your own article “A Single-Couple Human Origin is Possible” (https://bio-complexity.org/ojs/index.php/main/article/view/BIO-C.2019.1) you and Ann Gauger arrive at the same results. So why are you surprised that I don’t believe genetics shows a creation 6000 years ago?
The critical point here is whether there are genetic mechanisms that in 6000 years can generate today's genetic variation from a single couple. You write "Even if there were only four raw copies of each non-sex chromosome in the first generation, mutations (i) and recombination (iv) in future generations can generate a large number of variants of different genes". Here's one thing I have to ask you: one of the main points of young earth creationism is that mutations are basically always harmful and destructive. They would cause irreversible losses of "information" and an ever-increasing "genetic entropy". Are you prepared to argue contrary to this that mutations and recombinations (which are a form of random mutation) can generate new, functional alleles of a gene? HLA-B, the gene with most variation to my knowledge, was reported in June 2021 to have 8,181 alleles. So can a blind process of random mutation and selection generate thousands of functional, new gene variants?