Assembly Theory: Pseudoscience Masquerading as Revolutionary Insight Exposed
Three new papers, two of which have been published in Nature and PLOS journals, led by my group, have proven Assembly Theory to be equivalent to Shannon Entropy and LZ compression (the same compression used for file formats like ZIP and PNG). The claims by the authors of Assembly Theory about unifying areas of science are unfortunate and only an item in a long list of academic misconduct.
Click here to know more about the story of the Assembly Theory deception after years of calling out the many issues in substance and form of Assembly Theory, or here to learn about a new list of fallacies by Assembly Theory.
Disclaimer: This critique is not about stifling innovation or rejecting a competing view because it challenges the current status quo, knowledge or scientific establishment but exactly the opposite. Despite being presented by the authors as innovative and groundbreaking, it takes research in complexity science about 50 years back by repackaging well-established concepts and methods that have existed for decades, renamed and reintroduced without proper acknowledgment and under new names to look more elevated than it is. At best, this constitutes very poor scientific practice; at worst, it approaches plagiarism and fraud.
The application of Shannon Entropy and statistical compression to molecular spectral data, while potentially interesting, does not introduce a fundamentally new method and is not inherently novel as AT can be replaced by popular compression algorithms or Shannon entropy to reproduce exactly the same results as we have shown therefore making AT redundant and irrelevant. Instead, the authors of AT have made sweeping claims of having developed an original framework that explains fundamental concepts such as selection, evolution, life, time, and matter — claim that is not substantiated by any of their work.
Moreover, even in the specific application of molecular complexity and molecular graphs, much of the ground they cover had already been explored and reported before, and they were evidently aware of this prior work. For instance, our earlier measure, the Block Decomposition Method (similarities are not a coincidence), developed in the early 2010s (Physica A, PeerJ), reported the same results without being framed as a “unifying theory of everything,” as the authors of AT have positioned their work in their choice of paper titles, press releases and all over the media.
The authors ignored our work knowing about it as they invited us on multiple occasions to their seminars and book compilations some of which we honoured and some of which we did not, all on this same subject as it would make obvious to their readers the evident fact that the results reported by Assembly Theory had been widely reported before. Indeed, their main results on which all their arguments were based upon and everything else followed, the separation of organic from non-organic compounds, and the connections of compression theory with selection and evolution, were reported years before their own respective papers on those topics, which they decided to ignore and never mention even after years of reporting the serious omission publicly here and elsewhere since at least 2022.
During the review process of one of our papers criticising AT submitting to a Nature journal, Lee Cronin and Sara Walker attempted to derail the review proces by failing to disclose their competing and conflicts of interest — a violation of the journal’s strict policies they signed for when wrongly accepting reviewing our paper, causing significant delays that they used to their advantage claiming that we were unable to publish our criticisms.
Nature found them at fault and removed them from the review process after finding them at fault and in undisclosed conflict. Furthermore, Cronin and his group publicly posted their review on the scientific preprint server arXiv (see details below) while the review process was ongoing, violating the confidentiality agreement with the journal, again. Cronin later attempted to withdraw the review that he should not have been made public on the arXiv, but the breach of ethical standards had already occurred and in fact, he was unable to remove previous versions that fully contained his conflicted review. All other four reviewers of this paper also had access to their review and disagreed with it ultimately leading to the publication of the paper in a Nature journal almost at the same time a second one was published by PLOS, and a third one was made available, hence making all papers so far published in our first choice of journals, unlike Cronin open practices of submitting his papers to multiple journals even after been told he is wasting journal editors’ time (end of video).
The scientific integrity of the authors of Assembly Theory scientific reporting is also highly questionable. In their latest paper, for example, a figure we highlight below presents incomplete experiments and selectively reported results, potentially designed to mislead reviewers and readers. But this is only one of many examples attempting to mislead the public.
The conduct of Cronin and Walker raises serious ethical concerns that should not go unnoticed but current academic systems encourage and does not punish. These issues, collectively, represent one of the most egregious cases of scientific misconduct, spearheaded by Lee Cronin and supported by Sara Walker. We know these accusations are serious, and we do take them seriously.
Assembly Theory ‘groundbreaking finding’
The core of Assembly Theory rests on a glaringly false premise: that counting copies of molecular structures somehow unlocks the secret to life and the universe. This concept is nothing more than the recognition that self-replication and the accumulation of matter are central to living and physical systems — an idea that has been widely accepted for over a century. Assembly Theory attempts to pass off this banal observation as a groundbreaking insight, all while ignoring decades of prior work in algorithmic complexity, information theory, systems science, and molecular biology. Cronin and his group have gone so far as to claim that Assembly Theory explains not only life, but also time, matter, life beyond Earth, and even the expansion of the universe and its cosmic expansion in videos promoted by the Templeton Foundation — a spectacular overreach that reeks of pseudoscience.
Cronin and Walker’s continued refusal to engage with the vast body of prior research is telling. They have consistently failed to cite or acknowledge our work and that of many others who have demonstrated, long before them, the utility of information theory and algorithmic complexity in fields ranging from evolutionary biology to genomics. Instead, they have resorted to renaming well-established concepts and presenting them as novel discoveries. For example, in one of their recent papers, the first sentence absurdly claims that “current approaches to evaluate molecular complexity use algorithmic complexity, rooted in computer science, and thus are not experimentally measurable.” This is a patently false statement, betraying either profound ignorance or deliberate obfuscation. Algorithmic complexity, and its approximations via compression algorithms, has been successfully applied to experimental data for decades. Our own research has demonstrated that Assembly Theory’s assembly index is nothing more than a poorly disguised approximation of algorithmic complexity, bounded by Shannon Entropy, and functionally indistinguishable from LZW compression.
The fallacy that algorithmic complexity and information theory are somehow unrelated to the experimental sciences is central to the false narrative that Cronin and his group propagate. In one of their most cited papers, the Molecular Assembly Index is shown to correlate with experimental data, and they present this correlation as evidence that Assembly Theory can be measured experimentally.
For example, just in this paragraph from their recent paper in ACS Central Science, there are more falsehoods than thruds. According to the authors, “Traditional methods in evaluating molecular complexity largely rely on algorithmic complexity, which is a concept normally seen in computer science applications... However, algorithmic complexity does not work in molecular analysis because it cannot be directly measured through experiments…” First, it is false that traditional methods in evaluating molecular are based on algorithmic complexity, they are mosty based on classical information theory, namely Shannon Entropic-methods of which Assembly Theory is only one more and equivalent as per our mathematical proof and empirical comparison. Then, the second part is even more striking, the fact that Cronin and his authors think that anything done in computer science is not relevant to science or experiments. Without invoking the fact that the Nobel Prizes of Physics and Chemistry went to computer science for their contributions to physics and chemistry, the only way that algorithms from computer science could not be applied to any type of data of molecular chemistry is that they were using magic. There is no other option. But not even that because their Assembly index is an algorithm itself, that they do run on a computer. As soon as their data has a representation, any algorithm can run on it, in particular Shannon Entropy or Assembly Theory (which is Shannon Entropy with extra steps). In other words, everything they write and claim is self-contradictory at levels rarely seen before.
What they conveniently omit, on purpose, (as they know our research for at least 10 years, and we have been public about all this for about 5 years now) is that our group has demonstrated through mathematical proof that the assembly index can be derived using standard compression algorithms like LZW. Our work shows not only that Assembly Theory offers nothing new but that its methods are embarrassingly simplistic. To claim that a small Pearson correlation “demonstrates” the validity of Assembly Theory is a stretch at best; in reality, it demonstrates the applicability of algorithmic complexity through compression, a field that we have spearheaded reporting similar results but conducting proper experiments without the hype, not introducing it as a revolutionary new theory of everything that created a lot of useless jargon to appear sophisticated, a stunt to make the hypothesis seem “elevated.”
The more the authors of Assembly Theory push their flawed hypothesis into the public eye, the more it becomes apparent that the theory lacks any real substance and is becoming a quintessential example of scientific malpractice. Some colleagues have asked me why I continue exposing it when most scholars do not take it seriously anymore. The answer is simple: it is a matter of defending the integrity of science. It is not just Cronin who is to blame; it is also the fault of those who have amplified the false statements without any balance or restraint, and those that have remained silent who have allowed Cronin to continue spreading misinformation under the guise of academic legitimacy, misleading the public, many amateurs, and even some naive academics.
Spreading misinformation for self-promotion
Lee Cronin and Sara Walker continue spreading misinformation conflating all sorts of basic scientific terms to confuse people in what looks a concerted purpose of academic deception.
With this set of new papers, however, Assembly Theory is exposed for what it is, a rebranding of areas and measures used in complexity science long time ago that has fooled journals, media outlets and science writers alike.
Of course, they do not cite, and even dismiss our work because, they claim, our algorithms are ‘theoretical’ (and theirs ‘are not’), whatever that means. In fact, they can apply theoretical indexes directly to their ‘experimental’ data and get the same results but they never do. We have done so for them though. Here is a key plot on how we have reproduced their results with naive algorithms applied exactly the the same data and producing exactly their same results (often better), no need for a new theory based on thin air or to claim to unify areas of science from results that can be obtained by simplistic indexes invented 70 years ago, including traditional statistics:
All the new literature above from produced by my group led to my School of Biomedical Engineering and Imaging Sciences at King’s College London to distribute this Press Release last month:
100% Marketing 0% Substance
Assembly Theory (AT) began as a hypothesis purported to detect life by measuring molecular complexity, a claim that was subsequently debunked when it was shown to be nothing more than a rebranding of complexity measures established decades ago. Despite this, the proponents of AT then escalated their assertions, declaring that the theory unified biology and physics — again, without presenting any new or valid evidence, relying instead on the flawed foundations of their earlier disproven results. The same group now asserts that AT can explain not only life but also time, matter, the expansion of the universe, and even cosmic inflation. These bold and increasingly preposterous claims, propagated through highly produced videos funded by the Templeton Foundation, lack any scientific basis or supporting data. Rather than grounded in rigorous research, these claims are inflated iterations of an already discredited idea, compounding its scientific vacuity with grandiose, unfounded declarations.
Lee Cronin’s Assembly Theory has emerged not as a scientific breakthrough, but as a scientifically hollow construct, wrapped in misleading jargon and perpetuated by Cronin’s relentless campaign of self-promotion (now joined by Sara Walker). Cronin and Walker, have continuously disseminated false information disguised as cutting-edge research, capitalising on a façade of academic rigour. Despite the lofty claims made by their work, Assembly Theory has been rigorously dismantled and shown to be nothing more than a superficial rebranding of well-established concepts like Shannon Entropy and LZ compression, tools that have long been fundamental to computer science and information theory.
Constantly crafting new fallacies in their relentless arguments
In their constantly shifting narrative, my team and I have noticed that they are now employing a fresh array of classic fallacies on top of the other 8 fallacies we identified before:
- We-are-better-because-we-are-worse fallacy: The claim that while the authors may now be accepeting that they have no other option but to acknowledge they are like a compression algorithm, they insist they are not or are very different because their Assembly index is not optimal and is not looking to compress data. However, they would still need to explain what the new algorithm does differently (and it is not its ‘stochasticity’ for which Shannon Entropy was introduced in the 1940s) and what advantage being not quite like a compression algorithm gives, if they claim that the evidence that they quantify evolution and selection are the results that all other compression algorithms are able to replicate. In other words, why to use an algorithm that is like Shannon Entropy or like a popular compression algorithm like LZ but with extra steps rather than simply applying Shannon Entropy or like a popular compression algorithm like LZ in the first place, without having to create a whole new dictionary of things that have been introduced before to look elevated (assembly pathway, assembly space, etc).
- Middle Ground fallacy: Related to the above, the claim now that instead of being either completely different from or totally equivalent (as in fact it is, as we have demonstrated), they are claiming now that Assembly Theory has “similarities” with computer compression. This seems to be done in the hope to divert the audience to make their stance seem more moderate and reasonable than considering one of the both extremes. Notice that in the past, they had said they had nothing to do with computer compression and were totally orthogonal. What is clear is that they are little by little accepting their algorithm is a compression algorithm, something we have criticised since the beginning because they started saying that compression was irrelevant and that they had nothing to do with computer science or algorithms or even mathematics (whatever that means, given that their algorithm is written and executed in the language of mathematics and computers).
- Fallacy of Origins: (related but not explicitly in conjunction with the *wishful thinking fallacy* previously covered). Their new main fallacious argument goes like this: because we are physicists aiming at a theory for physics, Assembly Theory (AT) would not be equal or equivalent to computation and compression that is only ‘theoretical’. This has no basis and is false that only because they are physicists, their compression algorithm is more relevant to physics. Under this fallacy, all mathematics and computer science would be useless to describe and explain the world. It is actually the opposite, turns out that highly and purely mathematical and geometrical theories can explain physics, like general relativity and quantum mechanics. Yet, the authors of AT seem to suggest that math and computer science are inadequate to describe the world. And even if they were correct, they are offering nothing but a purely computational simplistic algorithm that is no different to a computer compression that has not proven any advantage over other compression algorithms that do better than theirs at their own alleged breakthroughs, like separating organic from nonorganic molecules or reconstructing phylogenetic trees. The fact that is a deception is notorious from the fact that they have never tested their algorithms against others in any of their papers and refuse to do it over and over.
Masters of deception
The assumption that life is increasing in complexity is surely nothing new, and certainly it was not Assembly Theory that brought it up. This notion is at least as old as the field of complex science itself, and no, Assembly Theory does not quantify life (selection or evolution) any better than much better-grounded complexity measures as we have proven in two of the three papers.
Another false argument they have tried to advance is that their measure is ‘stochastic’ in nature (so does Shannon Entropy). However, they would need to show:
- They can derive any new insight or result that other measures would not because they do not have any evidence that their assembly algorithm is better than any other uncompression (assembly) algorithm. The evidence today says that this is not the case, because other compression algorithms do better at what they take as the main evidence in their favour (e.g. separation of organic v nonorganic molecular compounds).
- They can’t arrive at their results, whatever they are, with other existing naive statistical algorithms (we have reproduced all their results with existing simple statistical algorithms, all of them do the same, with their assembly index adding nothing new or better, this is the most basic thing they should have started with, but they did not and refuse to do).
- We already know most processes don’t assemble like Assembly Theory suggest and that most assembly pathways will have operations other than simple join and concatenation operations (this is, just putting ‘things together’) for which they cannot account for hence missing key steps, if not most, because exact copies to build with in nature are increasingly unlikely the higher the complexity layer of description (no two fingerprints are the same even when they may be built from the ‘same’ kind of skin cell proteins or ‘same’ protein molecules). But their assembly index cannot even account for reversions, so it would even miss chirality, something our measures would not.
From Ignorance to Fraud and Plagiarism
Indeed, our Block Decomposition Method already accounts for counting exact copies that can account for join operations but also all other mechanistic and causal operations. And do so also stochastically because it takes into consideration all the possible mechanistic pathways that build an object from its most likely computational origin. With this, we have explained many genetic, molecular and evolutionary processes within a probabilistic framework given that we consider all possible causal (assembly) paths to mechanistically build an object from its most likely origin up to the amount of computational resources available. This means that the more computational resources we have access to, the more precise and thorough we can be, something that other closed measures of complexity can’t do, hence unable to catch up the uncomputable nature of living systems (they are not constrained in any way). The more resources, the many more pathways our framework we can systematically find into how an object may have been generated. However, having not much resources also allow us to produce as many pathways as possible. Doing this, we have been able to reverse dynamical systems like cellular automata, pinpoint potential cancer pathways that agree with the literature, reconstruct the relationship of animal cells back to their stem cells, find regions of high genomic encoding content, or find new ways to cluster data by most likely causal origins. All things that have never been done before with similar tools.
Assembly Theory, a narrow unidimensional Theory of Nothing
The key takeaway from this investigation is that scientific tools used to measure the complexity of life should not only focus on the internal structure of living systems (like molecules or cells) that can account only for join operations like trivially putting things that look exactly the same together. But also on how these systems interact and adapt considering their internal causal structure that can account for any mechanical operation and noise beyond the only one trivial statistical regularity that assembly theory looks for. For this, we introduced hybrid measures, such as the Block Decomposition Method that combines the best of two worlds, the computable world of Shannon Entropy that can account for statistical operations, and the world of algorithmic complexity that can account for all mechanical operations. The two together equip a framework we called Algorithmic Information Dynamics, that takes into account all the many complex ways in which an object can be built, looking at how a system may adapt and react to perturbations or interventions of the environment on the agent and vice versa. We were able to prove that with these indexes we could also separate organic from nonorganic molecular compounds years before Assembly Theory (that they decide purposely not to cite) and also connect these ideas on complexity, causality and block composition to selection and evolution. So, the authors of Assembly Theory managed to plagiarise a dumbed-down version of our work where they say they want to do what we did, but actually don’t, and they have rather proposed a completely trivial version of our own measures that turns out to be equivalent to the work done in the field but 70 years ago.
The implosion of Assembly Theory
The implosion of Assembly Theory is not surprising to those familiar with its foundations. Three recent papers — two of which have been published in high-profile journals like Nature and PLOS — produced by our group have mathematically proven that Assembly Theory’s so-called “revolutionary” assembly index is, in fact, functionally equivalent to Shannon Entropy and basic compression algorithms like LZW. These are the same algorithms that underlie common file formats such as ZIP and PNG. The notion that Assembly Theory offers any novel insights is therefore a charade, one that has managed to slip past the review process through persistent manipulation of the academic system. Eventually, the media will hopefully grow tired of these author circus, not science.
Cronin’s group has become adept at gaming the media and the publication process. Time and again, they manage to push through papers that perpetuate scientific misconceptions by appealing to the ignorance of certain reviewers and editors, as Cronin himself has openly admitted. The result is the publication of papers that are riddled with scientific inaccuracies, lacking in rigour, and, frankly, a disservice to the scientific community. It is a tragedy that a once-respected institution like the University of Glasgow now harbours such charlatanism. The publication of Assembly Theory in prestigious journals is reminiscent of the infamous anti-vaccine paper published in The Lancet: it serves as a stark reminder that even top-tier journals are not immune to being exploited by those with the resources who have learned to manipulate the system. Such publications, however, do nothing to legitimise Assembly Theory; they merely expose the vulnerability of the peer review process when faced with relentless and unscrupulous authors.
Snake-oil science
It is particularly disheartening to see how Cronin’s association with Sara Walker has only amplified the dishonesty. Together, they have escalated their claims to the point of absurdity, suggesting that Assembly Theory offers a unified explanation for life, matter, time and the universe. This is nothing more than the academic equivalent of snake oil. Assembly Theory has now reached the point where it can no longer be taken seriously by any informed academic. Its proponents have become the purveyors of a modern-day deception in the post-truth era, peddling empty promises to the naive and uninformed, all while doing significant harm to the credibility of science as a whole.
In sum, Lee Cronin’s Assembly Theory now also pushed by Sara Walker represents one of the worst examples of academic malpractice today. It is not just a flawed theory; it is a deliberate attempt to deceive, to appropriate the work of others while masquerading as a revolutionary new framework for understanding life. It is an affront to the principles of scientific inquiry, and it is high time that the academic community calls it out for what it is.
Click here to know more about the full story of the Assembly Theory deception
