This exchange consists of seven posts, each with a comment by Ola Hössjer and a reply by Lars Johan Erkell, plus one other post by Erkell. It is a translation, mostly by Google Translate, from the original 2020 Swedish posts and comments. Erkell and Hössjer have kindly corrected the translation. You and they are welcome to comment at Panda’s Thumb. Keep in mind that the posts and these appended comments were written two years ago and thus cannot take more recent comments here into account. We hope to publish the sections on seven consecutive Wednesdays (we chose the day of the week without remembering its connection to ancient days in Sweden). Readers and these authors are invited to contribute further comments in our Disqus comment system.
This week’s post starts with an introduction by Lars Johan Erkell, followed by the first of his posts at Biolog(g) and a comment by Ola Hössjer followed by a rejoinder by Lars Erkell.
by Lars Johan Erkell
A breakthrough for Intelligent Design?
Two years ago, Steinar Thorvaldsen and Ola Hössjer, professors of Informatics and Mathematical Statistics, respectively, published an article titled “Using statistical methods to model the fine-tuning of molecular machines and systems” in “Journal of Theoretical Biology” a regular scientific journal. Within the Intelligent Design (ID) community the publication was seen as a breakthrough for ID as science. A close reading of the paper, however, is not convincing. In seven blog posts on “Biolog(g)” (a blog run by the staff at the Institute for biology and environmental science at the University of Gothenburg, Sweden) I therefore have detailed my critique on four specific points, and also asked how this paper could be published.
1) Where are the results?
The article title, “Using statistical methods to model …” as well as wordings in the article clearly state that statistical methods have been used to demonstrate fine-tuning in the cell. However, no results or calculations are presented to support this claim.
2) How can design be calculated?
The authors present a statistical model comparing M1, a design model, and M2, a naturalistic model, in their ability to explain the occurrence of complex cellular structures. However, M1 is misconceived and M2 is unworkable.
3) Miracles as science?
The authors also suggest that their statistical model can be used to study human evolution. Here, they compare M1, a model assuming that humanity arose from a single human couple, and M2, assuming common ancestry with chimpanzees. However, to explain how present human genetic diversity could be generated from single couple within a time frame of 6000 to 10 000 years, they resort to divine intervention.
4) Non-physical information?
In their article, the authors introduce the revolutionary concept of “non-physical information”, but without discussion or definition. The only reference goes to a text by philosopher Del Ratzsch, who does not use the term. Instead, the authors have fabricated a hybrid sentence, part of it by Del Ratzsch and part of it written by themselves, attributing it to Del Ratzsch. That is, they have presented no evidence whatsoever for the existence of “non-physical information”.
5) Why is ID not regarded as science?
Some reasons ID is not regarded as science are briefly outlined. It is pointed out that ID does not formulate any concrete theories or predictions and consequently cannot be falsified. ID is a theory without a theory, without method and without results.
6) Why was this article written?
One may ask why two established professors at respected Scandinavian universities would write an article like this. The reason is conceivably that they both are connected to Discovery Institute, the central hub of the ID movement.
7) Why was this article published?
The article has been heavily criticized for being pseudoscientific, and the review procedures of the Journal of Theoretical Biology have been questioned. After the first publication of this post in 2020 (in Swedish) additional information seems to confirm the suspicion that editors and reviewers were pushing the article through the review process, against the policy of the journal.
Breakthrough for Intelligent Design? (Part 1)
originally posted November 13, 2020
by Lars Johan Erkell
Where are the results?
That articles openly arguing for Intelligent Design (ID) are published in the established scientific press is not something one expects. Therefore, it is worth noting that Ola Hössjer, Professor of Mathematical Statistics at Stockholm University, and Steinar Thorvaldsen, professor of Information Science at Norway’s Arctic University in Tromsø, published a paper this summer entitled “Using statistical methods to model the fine-tuning of molecular machines and systems” in the Journal of Theoretical Biology. On July 8, the Swedish Christian newspaper Världen Idag had a report about the publication under the heading “Breakthrough: Research points to intelligent design”. The Swedish creationist association Genesis writes that Norwegian-Swedish research points to intelligent design in our cells. Now is this a scientific breakthrough for ID?
Perhaps not the breakthrough one had hoped for. When the article was printed in the September issue, there was a disclaimer from the magazine’s editors-in-chief on the website. They distance themselves from the concept of intelligent design and claim that they were misled by the fact that the authors did not include the keyword “intelligent design” during the review, but added it later.
There’s much to say about this business, so much that I’m devoting seven posts to it.
The concept of fine-tuning is one of the more common arguments for ID. It is argued that the natural constants that determine the properties of matter and the universe could have been radically different, and that the probability of a universe with properties that allow intelligent life is vanishingly small. Nevertheless, we are here, and that cannot be a coincidence. It shows design – a cosmic mind has created a universe that suits us. That’s the argument in a nutshell.
The purpose of the work is to now go one step further and to show that fine-tuning also exists in the living cell and its components:
This paper describes molecular fine-tuning, how it can be used in biology, and how it challenges conventional Darwinian thinking. We also discuss the statistical methods underpinning fine-tuning and present a framework for such analysis.
Fine-tuning means that a number of natural constants have such values that life is possible. But the constants are still not so life-friendly that life could arise and develop through evolutionary processes. Life would therefore not be possible without the intervention of a designer. Thorvaldsen and Hössjer claim to see clear traces of design in the cell’s proteins and in its complex mechanisms. You get the image of a designer who doesn’t quite manage to create a sufficiently life-friendly universe, and is therefore forced to build all living beings together, in the smallest detail. The universe is both life-friendly and life-hostile at the same time.
Now, fine-tuning is a notoriously hard-to-define term, so one is glad to actually get a definition:
We define fine-tuning as an object with two properties: it must (a) be unlikely to have occurred by chance, under the relevant probability distribution (i.e., complex), and (b) conform to an independent or detached specification (i.e., specific).
The first condition is therefore that it is unlikely that the object in question could have arisen by pure chance. The second is that it must respond to some sort of independent specification. The question is whether it is possible to detect fine-tuning with advanced statistical methods:
Is it possible to recognize fine-tuning in biological systems at the levels of functional proteins, protein groups and cellular networks? Can fine-tuning in molecular biology be formulated using state of the art statistical methods, or are the arguments just “in the eyes of the beholder”?
That complex biological structures cannot have been formed by pure chance is so obvious that one hardly needs advanced statistical methods to convince oneself of the matter. The real problem is the concept of “specification”; how to define specification in concrete terms is a question the ID movement has wrestled with for decades, without coming to any solution. The article’s discussion of the concept is purely mathematical, and contains no clues as to how specification could be demonstrated in practice. The authors refer to an article by George Montañez, “A unified model of complex specified information”, but this too is completely mathematical and says nothing about how one could determine whether a certain structure or molecule is specified or not.
At the end of the article, section 7, we read:
Functional proteins, molecular machines and cellular networks are both unlikely when viewed as outcomes of a stochastic model, with a relevant probability distribution (having a small P(A)), and at the same time they conform to an independent or detached specification (the set A being defined in terms of specificity). These results are important and deduced from central phenomena of basic science.
It is therefore not possible to give an account of what specification actually means in practice. Instead, the authors refer to “central phenomena in basic science”, without giving any reference. In the last paragraph they write:
A major conclusion of our work is that fine-tuning is a clear feature of biological systems. Indeed, fine-tuning is even more extreme in biological systems than in inorganic systems. It is detectable within the realm of scientific methodology.
This is simply not true. Since there is no working method to scientifically detect specificity in actual biological systems, it has not been possible to demonstrate fine-tuning with scientific methods. It has therefore not been possible to conclude that fine-tuning is characteristic of biological systems. This is not possible because condition (b) of the definition of fine tuning is not met. It is also impossible for the simple reason that no actual data or any calculations are presented to support the claims. The statistical methods developed are not used to solve any actual problems. All you have is speculation.
The title of the article reads “Using statistical methods to model the fine-tuning of molecular machines and systems”. It applies that statistical methods have indeed been applied to model fine-tuning - the word “using” reasonably means that something has really been used. In fact, no methods have been applied at all; nothing has been discussed besides how it might be done.
It is difficult to understand that Thorvaldsen and Hössjer in this way make it appear that they have results they clearly do not have. The reader may get the impression that they are bluffing.
Comment by Ola Hössjer June 29, 2021 2:45 a.m
Has Intelligent Design made a breakthrough as a science?
Introduction and background
Last year I published the research article “Using statistical methods to model the fine-tuning of molecular machines and systems” in the Journal of Theoretical Biology together with Steinar Thorvaldsen, who is Professor of Information Science at Tromsø University. In this paper, hereafter referred to as TH2020, we propose a mathematical and statistical framework to test whether biological systems (proteins, protein complexes, cellular networks) can arise by chance or whether fine-tuning is a better explanation. Fine-tuning of a system means that a number of different parameters must be finely calibrated for it to exist. This requires (a) that the probability of the system having arisen by chance is small, and (b) that there is an independent specification of what is fine-tuned. This specification could mean that the system must have all its parts in place in order to function, that it has arisen in a short time despite its complexity, or that it includes codes, instructions, and organized structures that we recognize from intelligent agents.
In physics, the concept of fine-tuning has been used for 50 years and today a majority of physicists claim that the universe must be fine-tuned in order to exist and harbor life. On the other hand, fine-tuning has not had the same impact on biology. One may ask why. One explanation is probably that in many examples fine-tuning is associated with an intelligent designer, which in turn brings God to mind. Although the definition of fine-tuning does not involve a designer, this is often the most reasonable interpretation or explanation of specification (b) above. This connection between fine-tuning and design has been central for Intelligent Design (ID), a movement that arose in the United States in the late 1980s with the aim of proposing explanations for the origin and diversity of life, that are alternatives to the secular theories (chemical and biological evolution) that dominate within the academic world. And as a Christian I go one step further and identify the Intelligent Designer with God. With a creationist perspective, I can then use the Bible as inspiration to propose and interpret specified properties of biological structures. Thus, while the ID movement goes “halfway” and identifies fine-tuning with a designer, creationists go “all the way” and interpret fine-tuning as something God created or brought about.
For my own part, I have both a creationist and an ID perspective. Depending on the research problem and context, I choose to go either half way or all the way in my interpretation of how the fine-tuning may have arisen, or to make no interpretation at all. Sometimes it is appropriate to leave it to the reader to consider how the fine-tuned structures arose, in other contexts it is important to justify that the proposed fine-tuning is inspired by the knowledge we have about intelligent senders, or by texts from the Bible. At the end of the aforementioned article, TH2020, Thorvaldsen and I go halfway and make a connection between fine-tuning and an intelligent source. As this is unusual within a secular academic environment, our publication has received a lot of attention and circulation. Last year, TH2020 was the most downloaded article in the Journal of Theoretical Biology.
Among those who got involved in the debate around TH2020 is Lars Johan Erkell, who is Associate Professor in Zoophysiology at the University of Gothenburg. On Biolog(g), a blog run by the teachers of the Department of Biology and Environmental Science at the University of Gothenburg, Erkell wrote a series of seven articles in November 2020, entitled “Breakthrough for Intelligent Design (?)”. In these articles he analyzes and comments on TH2020. As an academic, I welcome factual criticism where my research results are analyzed and seen at the seams. At best, such debates contribute to the advancement of science. I also want to thank Erkell for actually taking the time to read, not only TH2020, but also several of the articles that Steinar and I refer to. That makes parts of his criticism well-informed. At the same time, Erkell’s series of articles contains a series of claims, for instance that our article is not only pseudo-scientific but also anti-scientific. I therefore would like to take this opportunity to respond to this serious criticism. I have divided my response article into seven parts (corresponding to Erkell’s seven articles) and a concluding summary.
Part 1. The results of the fine-tuning article
As mentioned in the introduction, the method for detecting fine-tuning in a biological system, which we propose in TH2020, involves two steps. First (a) to show that the probability is small that the system has arisen by chance and second (b) to demonstrate an external specification, which can be recognized by an observer. In part 1, Erkell asks what results are presented in TH2020. According to Erkell, our article contains only mathematical speculations that are not relevant to a biological reality. In other words, Erkell lacks concrete examples where we show how our mathematical theory for detecting fine-tuning should be applied.
It is true that in TH2020 we do not have concrete examples with numerical calculations in order to illustrate our theory. But this is not the purpose of the article. Our aim is partly to provide an overview of fine-tuning and partly to propose a mathematical framework for establishing fine-tuning, which can then be used in subsequent articles to investigate whether a certain biological system (a protein, a protein complex, a cellular network) is fine-tuned or not. At the same time, TH2020 contains about a hundred references to other articles, and at several places we give concrete suggestions on how the specification in (b) should be chosen, as well as detailed justifications that the probability (a) that the biological system arose by chance is very small. In addition, several of the articles we refer to are based on empirical studies or analyses of such. In other words, our mathematical model is not plucked out of thin air.1
At the same time, it is important to point out that mathematical models are simplifications of a more complicated biological reality. Whether the model is successful or not depends on how well it is able to describe empirical data. But this does not mean that every research article needs to contain such a connection between mathematics and empiricism, even if this connection is crucial in the long run. There are many examples of science (such as population genetics and theoretical ecology) which were indeed motivated by biological applications, but where initially the theory was developed quite separately from empirical work. Despite this, population genetics later became an important tool for describing small-scale microevolution, when more genetic data became available. Similarly, theoretical ecology has been used to understand the processes that bring about environmental and climate change. My question to Erkell is therefore whether he believes that such developments in the history of science lack value, where theoretical development sometimes precedes applications. Doesn’t this rather indicate a diversity of the scientific project?
- See for example Axe, D.D. (2004). Estimating the prevalence of protein sequences adopting functional enzyme folds. J. Mol. Biol. 341 (5), 1295–1315. and Kozulic, B., and Leisola, M. (2015). Have scientists already been able to surpass the capabilities of evolution? viXra Biochemistry 1504.0130.
Reply by Lars Johan Erkell September 8, 2021 5:52 am
Thanks for commenting on my posts! There is a lot to discuss here, and I hope we can sort things out.
Let me start with your question about whether I think it is OK to publish theoretical articles without empirical data. Yes, of course! It can lead to faster progress. Like when X-ray crystallographic diffraction patterns were first interpreted and the theory of space groups was already in place. So my criticism is not directed at you publishing a model without data, but that TH2020 clearly pretends to show results, but does not contain any. How can you interpret “A major conclusion of our work is that fine-tuning is a clear feature of biological systems … It is detectable within the realm of scientific methodology” if not that you have concrete results showing that fine-tuning actually does exist in biological systems? I thought I expressed myself clearly here.
So to the problems with the concept of “specification”. It’s notoriously difficult to define and I can’t find a clear definition or proper discussion of it in TH2020. But you suggest how it should be chosen, like “… a molecular machine, and that the specification states how many parts must be in place for x to work.” Another example is that f(x) indicates the biological fitness of an organism x, i.e. its reproductive capacity”. But one is a structure, the other is a property of a living organism. What does a definition look like that covers these very different cases? And what determines the necessary number of parts in a molecular machine? Furthermore, in the next example, if we move an organism from a favorable environment, where it has high biological fitness, to an unfavorable environment, where it has zero fitness, does it lose its specification? It is still the same organism. And is the specification a quality or a quantity? You mention both options in section 6.2. Which is it? You have to look for a more blurred concept.
In this my first post, I had two points of criticism: (1) that you pretend to have results that do not exist, and (2) that you do not define or sufficiently discuss the term “specification”. In your answer you avoid both points. Do you have any comment on that?