Censorship in the Name of Greater Good

In the seventeenth century, a Dutch drapery, fabric, and clothing merchant by the name of Antonie van Leeuwenhoek (see Figure 1, left) had an unusual hobby: he liked to play with specially polished glass that could magnify things. By improving lens-making techniques, he created the first microscopes capable of more than 200× magnification. Leeuwenhoek was the first to observe blood cells and microorganisms (see Figure 1 right), and he is considered (together with Louis Pasteur and Robert Koch) one of the founders of microbiology (Lane Reference Lane2015: 20140344).

Figure 1. Left: Antonie van Leeuwenhoek (1632–1723), portrait by Jan Verkolje. Right: A replica of Leeuwenhoek’s microscope. (Image credits: Wikimedia Commons.)

Importantly, Leeuwenhoek did not keep his observations to himself. He carefully documented what he saw through his microscope and communicated his observations to the editor of the Philosophical Transactions of the Royal Society. He published nearly 200 letters describing his observations in words and drawings, such as those reproduced in Figure 2.

Figure 2. Left: Illustrations to Leeuwenhoek’s Letter on the Animalcules found on Duckweed (Leeuwenhoek Reference Leeuwenhoek1702: 1304). Right: Sperm from rabbits (1–4) and dogs (5–8), drawn by draughtsman of Antonie van Leeuwenhoek in 1677. (Image credits: Wikimedia Commons.)

Leeuwenhoek was also the first to observe spermatozoa in semen (Figure 2, right), a breakthrough in understanding sexual reproduction. Yet Leeuwenhoek hesitated to communicate these findings, concerned that they might be offensive. In the cover letter accompanying his report to the president of the Royal Society he wrote: ‘If your Lordship should consider that these observations may disgust or scandalise the learned, I earnestly beg your Lordship to regard them as private and to publish or destroy them as your Lordship sees fit.’ Fortunately for science and humanity, his Lordship published Leeuwenhoek’s revolutionary observations in Philosophical Transactions in 1678 (Leeuwenhoek Reference Leeuwenhoek1677: 1040).

Things might have played out differently for Leeuwenhoek in 2023. According to new guidelines of the Royal Society of Chemistry (RSC), editors must now

The document says,

The guidelines provide 15 indicators of what constitutes offensive content, including content that is ‘likely to be upsetting, insulting or objectionable to some or most people’ – a criterion so vague and broad that it could be used to censor just about anything in chemistry and beyond.

So what could be offensive in a scientific paper? One might suppose that apart from areas of research in the life sciences that touch upon such polarizing topics as evolution, differences between sexes and races, sexual reproduction, and stem cell research, the rest of STEM (science, technology, engineering, and mathematics) would be safe. After all, what could be so offensive in chemistry and physics?

According to CSJ ideologues, the very language we use to communicate our findings is a minefield of offences. Professional societies, universities, and publishing houses have produced volumes dedicated to ‘inclusive’ language that contain long lists of proscribed words that purportedly can cause offense and – according to the DEI bureaucracy that promulgates these initiatives – perpetuate inequality and exclusion of some groups, disadvantage women, and promote patriarchy, racism, sexism, ableism, and other isms. The lists of forbidden terms include ‘master database’, ‘older software’, ‘motherboard’, ‘dummy variable’, ‘black and white thinking’, ‘strawman’, ‘picnic’ and ‘long time no see’ (Krylov Reference Krylov2021: 5371; Krylov et al. Reference Krylov, Frenking and Gill2022a: 32; McWhorter Reference McWhorter2021; Paul Reference Paul2023; Packer Reference Packer2023; Anonymous 2022). The Google Inclusive Language Guide even proscribes the term ‘smart phones’ (Krauss Reference Krauss2022a). The Inclusivity Style Guide of the American Chemical Society (2023) – a major chemistry publisher of more than 100 titles – advises against using such terms as ‘double blind studies’, ‘healthy weight’, ‘sanity check’, ‘black market’, ‘the New World’, and ‘dark times’ (see Figure 3). Recently, the School of Social Work of the University of Southern California declared the word ‘field’ to be racist (Krauss Reference Krauss2023). ‘Why,’ you ask? Because slaves used to work in fields. The memo recommended replacing this newly identified racist F-word by ‘practicum’, so that instead of ‘field of study’, we are to say, ‘practicum of study’. By extension, we take it that physicists should get used to Electromagnetic Practicum and Quantum Practicum Theory.

Figure 3. Recommendations from the Inclusivity Style Guide of the American Chemical Society (2023). The online guide is a massive document, with theoretical background, examples, and even exercises.

Today’s technology enables language policing in ways Orwell could only dream of. Software tools have been developed to ferret out offensive language and recommend Newspeak alternatives. For example, Microsoft Word offers an inclusive language plugin (Baker Reference Baker2022), and Google has developed software for checking computer codes for offensive words (Krauss Reference Krauss2022a).

Not only can common technical terms and plain English words be offensive, but names of scientists, we are told, can too, and so they should be obliterated. A rapidly growing list of scientists who have had their names stricken from buildings, textbooks, awards, and more (Krylov Reference Krylov2021: 5371; Krylov and Tanzman Reference Krylov and Tanzman2021; Bodmer Reference Bodmer, Bailey, Charlesworth, Eyre-Walker, Farewell, Mead and Senn2021: 565; Powell Reference Powell2022; Krylov Reference Krylov, Tanzman, Frenking and Gill2022b) includes such luminaries as:

• William Shockley (inventor of transistors);

• Fritz Haber (developer of nitrogen fixation process);

• Erwin Schrödinger (discovered the wave equation of quantum mechanics);

• Isaac Newton (Newton laws, calculus);

• Robert Millikan (oil drop experiment);

• Ronald Fisher (modern statistics);

• Thomas Henry Huxley (evolutionary biologist, ‘Darwin’s bulldog’);

• Edward O. Wilson (evolutionary biologist);

• James Webb (former head of NASA).

Some of these scientists, such as Shockley, did hold racist views or, like Milliken, were affiliated with organizations that promoted eugenics. Some had a hand in wartime atrocities, such as Fritz Haber, who oversaw the development and deployment of chemical weapons in the First World War. Others, who have no record of unsavoury beliefs or behaviours, stand among the cancelled because of how their work was used by others. For example, only because his statistical methods were allegedly used to justify racism was Fisher accused of being a eugenicist (Bodmer Reference Bodmer, Bailey, Charlesworth, Eyre-Walker, Farewell, Mead and Senn2021: 565). According to the award-winning physicist Prescod-Weinstein, it is sufficient to be a dead white man to get on the cancellation list (Powell Reference Powell2022).

The case of James Webb illustrates that facts do not matter to the cancellation mob (Powell Reference Powell2022). Webb was the head of NASA during the heroic period of the agency that culminated in sending a man to the moon. As an effective administrator and leader, Webb – according to many – deserved a fair share of the credit for the agency’s success, and, accordingly, it was decided to name the Webb space telescope after him. Unfortunately, Webb led NASA during a period when the US government discriminated against gay people. Predictably, this has led to accusations that Webb, himself, was a homophobe who discriminated against gay employees of NASA. A cancellation campaign against Webb by a Twitter outrage mob and a petition demanding that NASA rename the telescope ensued. In response, NASA conducted a formal investigation. Their conclusions, contained in an 80-page document, were that there is no evidence that Webb participated in any anti-gay discrimination or held any homophobic views. But these findings have not shut down the cancellation mob, which continues the campaign.

A disturbing aspect of this story is how scientific publishers have reacted to the cancellation campaign. Rather than following the example of NASA leadership, which evaluated the validity of the claims of the activists before making a decision, the Royal Astronomical Society responded with the following instructions to authors: The Royal Astronomical Society ‘expects authors submitting scientific papers to its journals to use the JWST acronym rather than the full name of the observatory. In this case, the previous requirement for the acronym to be spelled out at first mention will not be observed’ (Kahlon Reference Kahlon2022).

It may be tempting to dismiss these and similar acts as frivolous and irrelevant to the actual practice of science. After all, quantum mechanics does not change whether we say the ‘Schrödinger Equation’ or the ‘Great Wave Equation of the Microscopic World’; the informational content of images of stars and galaxies (see Figure 4) does not depend on the name of the space telescope.

Figure 4. Image of a star-forming region in the Carina Nebula obtained by the James Webb Space Telescope. (Image Credit: NASA; https://www.nasa.gov/webbfirstimages.)

However, the cancellation of words and names does not stop there – it extends to the suppression of ideas, viewpoints, and even research (Krylov et al. Reference Krylov, Tanzman, Frenking and Gill2022b: 12; Stevens Reference Stevens, Jussim and Honeycutt2020: 82). A case well known to the chemistry community is the cancellation of Tomáš Hudlický, who wrote an essay entitled ‘“Organic synthesis – Where now?” is thirty years old. A reflection on the current state of affairs’. Hudlický’s concerns about the integrity of the scientific literature, the decline of the work ethic, and the growth of identity-based discrimination enraged CSJ activists and, in response to the ensuing Twitter outrage, the journal (Angewandte Chemie) removed his peer-reviewed, published essay (Krylov et al. Reference Krylov, Tanzman, Frenking and Gill2022b: 12; Deichmann Reference Deichmann2023).

In a similar incident in 2022, the astronomer John Kormendy had written a research paper entitled ‘Metrics of research impact in astronomy: Predicting later impact from metrics measured 10–15 years after the PhD’, in which he presented statistical methodology he developed to help predict the ‘future impact’ of an astronomer’s research in order to ‘inform decisions on resource allocation such as job hires and tenure decisions’ (Krauss Reference Krauss2022b). Kormendy was, in part, motivated by the goal of reducing bias and improving fairness in evaluating prospective hires. He submitted the manuscript to the Proceedings of the National Academy of Sciences (PNAS) and uploaded it to a preprint server. But the evaluation of merit, a concept the adherents of CSJ ideology find odious, generated a Twitter outrage storm. In response, Kormendy withdrew the paper from the journal, removed it from the archive, and issued the following abject apology:

These two examples are not outliers; rather, they typify the current state of science publishing. In 2022, Nature Human Behaviour (NHB) published an editorial stating that the journal will not publish valid research that the editors consider ‘harmful’ to groups (Unsigned 2022: 1029). The editorial proclaims: ‘Although academic freedom is fundamental, it is not unbounded’ and, ‘Advancing knowledge and understanding is a fundamental public good. In some cases, however, potential harms to the populations studied may outweigh the benefit of publication.’ In response, Jonathan Rauch published the following searing critique:

Rauch goes on to explain (and illustrate with examples) that predicting future effects of research is not possible, and that enabling the authorities – whether spiritual leaders, totalitarian governments, benevolent dictators, or advocacy groups – to suppress knowledge that they would deem harmful is perilous to science and society:

Rauch notes: ‘Nature Human Behaviour’s manifesto … represents an explicit endorsement of social-justice gatekeeping by a respected scientific journal.’

Such attempts to censor ‘harmful’ research go beyond the publishing stage. For example, a recent research paper in PNAS calls for the establishment of ‘Ethics and Society Review Boards (ESRBs)’ at universities and research institutions (Bernstein Reference Bernstein, Levi, Magnus, Rajala, Satz and Waeiss2021: e2117261118). At present, research proposals are evaluated by institutional review boards to ensure that human or animal study subjects are ethically treated. ESRBs are intended to extend this practice to a new domain; namely, the ESRB will evaluate whether the proposed research could harm groups or populations outside the study, very much in line with the sentiments expressed by the NHB editorial (Unsigned 2022: 1029). If such anticipated harms are found, the research would not be funded. The paper cites the following as examples of future harms:

• ‘Risks that arise due to the technology being coopted for nefarious purposes – such as governments employing mass surveillance’;

• ‘Potential harms to any population – such as job loss due to automation’; and

• ‘Potential harms to specific subgroups – such as technical barriers to using technology that is prohibitive to poorer populations’.

These risks and potential harms go beyond NHB’s domain, social science, and are directly relevant to STEM. The vague and subjective notion of such ‘harmful research’ can be stretched to encompass all knowledge production and technological progress. Research in computational chemistry (of the type Krylov does), for example, could lead to new solar energy technology that could cause job losses in the coal industry. The algorithms for high-performance computing that computational chemists develop could be co-opted by bad agents for nefarious purposes. So by these criteria computational chemistry research can be censored out for good. ESRBs are still a proposal, but we know how quickly proposals become policies in the absence of resistance.

Critical Social Justice is Changing the Telos of the Scientific Publisher

CSJ ideology subverts scientific publishing beyond censorship and scholarship suppression by replacing the traditional purpose of a publisher with the CSJ agenda. The Royal Society of Chemistry (RSC) has issued a Joint Commitment for Action on Inclusion and Diversity in Publishing (RSC Joint Commitment n.d.), a declaration pledging to ‘engage all relevant stakeholders to improve outcomes on inclusion and diversity, at all stages of the publishing process’. This amounts to changing the mission of the organization from the publication of valid research to the promotion of DEI. Most – if not all – scientific publishers promptly signed this commitment, without input from the scientific community – including authors, reviewers, and editors. It might again be tempting to dismiss this manifesto as vacuous and harmless. One might even think that advancing DEI could improve knowledge production, for example, by reducing biases and discrimination. To examine if this is true, let us consider the specific proposed policies and the arguments put forward to justify these new policies.

The RSC lays out its plans in a document called the Framework for Action (RSC Framework 2020). The Framework calls for concrete steps, such as implementing identity-based considerations in appointing editorial boards and having authors recommend reviewers from under-represented groups. It attempts to justify these policies with the claim that there is significant gender bias in chemistry publishing and that women are disadvantaged:

The Framework cites the RSC’s own report, entitled Is Publishing in Chemical Sciences Gender Biased? (RSC Report 2019), which purports to have found the following answer to the titular question:

This is a strong claim – that the publishing outcomes are gender-biased. It flies in the face of my (Krylov’s) personal experience as an author of more than 300 papers and as an editor and reviewer of thousands more. But perhaps I am wrong and the data collected by the RSC will reveal the truth.

The only evidence for their claim, which both RSC documents (2019, 2020) cite, is the RSC-commissioned study ‘Is there a gender gap in chemical sciences scholarly communications?’ (Day et al. Reference Day, Corbett and Boyle2020: 2277). According to the RSC, Day et al. (Reference Day, Corbett and Boyle2020) documents bias against women. But Day et al. does nothing of the sort.

The authors of Day et al. (Reference Day, Corbett and Boyle2020) analysed numerous publishing outcomes, such as rates of manuscript acceptance, responses of reviewers, and actions of editors, with the stated goal of identifying whether and where gender imbalances exist. Although they found statistically significant gender differences for a variety of outcomes, the magnitude of the differences was small – e.g., 1–1.5 percentage points in reviewer recommendations for manuscript acceptance, revision, or rejection – and, as explained below, do not demonstrate gender bias. Figure 5, taken from the paper, serves as an illustrative example of their findings. Note that because the y–axis is clipped to the range of the data, the difference depicted in the left panel looks large, while it is actually only two percentage points (see the figure legend for details).

Figure 5. Illustration of how often editors agree with reviewer recommendations depending on the gender of the editor and the reviewer. The figure shows that female editors (left panel) agree with female reviewers 84 per cent of the time and with male reviewers 82 percent. Male editors (right panel) agree with reviewers 83.5 percent of the time regardless of gender. From this the authors conclude: ‘[T]he interaction of editor gender and review gender shows that the relationship is significant. Female editors agree with female reviewers significantly more than male reviewers’ (emphasis ours). The difference is ‘significant’ in the statistical sense only; in terms of magnitude, the two-percentage-point difference shown in the left panel is not much larger than the zero-percentage-point difference in the right, although it looks so because the y–axis is clipped to the range of the data. Reprinted from Day et al. (Reference Day, Corbett and Boyle2020) (license CC by 3.0).

Fundamental to drawing causal conclusions in observational studies such as Day et al. (Reference Day, Corbett and Boyle2020), is controlling for extraneous, or confounding, variables that could be the underlying causes of the observed results. That is to say, before concluding that an observed disparity between male and female scientists is due to gender bias, one must rule out other potential differences between the male and female scientists that could be responsible for the disparity. Such possible group differences might include, for example, the age, years of professional experience, academic rank, and chemistry sub-field of the scientist, or characteristics of the scientist’s institution, such as its size, rank, and whether it is primarily research- or teaching-oriented. To control for such confounding effects in observational research, investigators normally collect information on potentially causative variables and include them as covariates in the statistical analysis of their data. Such statistical control is considered obligatory in fields such as epidemiology, which routinely rely on observational (as opposed to experimental) methods. For example, in a recent study demonstrating a possible causal link between the intake of artificial sweeteners and heart disease, the investigators adjusted for 16 demographic, medical, lifestyle, and dietary variables, in order to rule out the possibility that they were responsible for the observed relationship between intake of artificial sweeteners and heart disease in the study population (Debras et al. Reference Debras, Chazelas, Sellem, Porcher, Druesne-Pecollo and Esseddik2022: e071204).

An observational study that does not control for confounding cannot draw any conclusions about why observed differences between the groups under study exist. Although the authors of Day et al. (Reference Day, Corbett and Boyle2020) tendentiously refer to the ‘small but significant drops in the female percentage of authors’ that they observe through the publication process as ‘death by a thousand cuts’, which they deem to be ‘disadvantaging’ women, they concede in the report that they did not have access to data that would have permitted them to control for confounding and, thus, they explicitly state that they are ‘identifying gender gaps and differences, rather than gender bias’.

Yet, despite the gross methodological limitations of Day et al. (Reference Day, Corbett and Boyle2020) and the authors’ explicit admission that their results do not imply gender bias, the paper has been repeatedly cited as evidence of anti-female bias in chemistry and used to justify imposing gender quotas on editorial boards and reviewer pools (RSC 2020; Burrows et al. Reference Burrows, Huang, Wang, Kim, Meyer and Schanze2020; ACS DEIR 2023).

Returning to RSC’s assertions of gender bias quoted above, we see that they are a mixture of unsupported and false claims. Although the RSC’s data (Day et al. Reference Day, Corbett and Boyle2020) do show small differences in manuscript acceptance rates depending on author, reviewer and editor gender, RSC’s claim that this represents a ‘disadvantage’ to women is unsupported due to the study’s limitations, and the RSC’s claim that Day et al. (Reference Day, Corbett and Boyle2020) reveal gender differences in the decision-making process is false. Contrary to the RSC’s claims, Day et al. (Reference Day, Corbett and Boyle2020) do not show bias at any stage of the publication process, much less at ‘every stage’. The RSC’s conclusion that action is required to mitigate ‘the risk and impact of bias in decision-making’ is unjustified, because the data they are relying on do not indicate gender bias in the first place, as explicitly stated by Day et al. (Reference Day, Corbett and Boyle2020).

In contrast to Day et al. (Reference Day, Corbett and Boyle2020) and the claims of the RSC, when confounding factors are properly controlled, evidence of gender bias in STEM all but vanishes. Controlling for confounding variables, a recently completed quantitative synthesis of the literature on gender gaps in journal acceptance rates as well as five other academic science domains (recommendation letters, tenure-track hiring, grant funding, salaries, and teaching ratings) found convincing evidence of bias only in teaching ratings, and the often-cited gender pay gap of 18% (Shen Reference Shen2013: 22) was reduced to 4%. In the other four domains (including manuscript acceptance rates), the authors concluded that there has been ‘no systematic gender bias in the last 10–20 years.’ (Ceci et al. Reference Ceci, Kahn and Williams2021: 40; Ceci et al. Reference Ceci, Kahn and Williams2023). Similarly, a recent encyclopaedic review of the literature on gender gaps in STEM found that ‘the evidence for endemic anti-female bias is inconclusive at best,’ and that, instead, ‘the main cause of the gender gaps in STEM appears to be average sex differences in people’s vocational preferences’ (Stewart-Williams and Halsey Reference Stewart-Williams and Halsey2020: 3).

Despite a complete lack of data demonstrating gender bias in chemistry publishing (and in the face, actually, of data that suggests gender fairness in scientific publishing), the RSC calls for heavy social engineering. For example, in 2020, editorial boards of RSC’s journals were asked to pledge their commitment to reach a target quota of 36% of women among editors and reviewers by 2022. The proposed quotas do not reflect the demographics of the field: women constitute about 20% of chemists in tenure and tenure-track positions (Wang and Widener: Reference Wang and Widener2019).

Echoing the gender bias narrative, the American Chemical Society lauds the Day et al. (Reference Day, Corbett and Boyle2020) study:

The ACS Diversity Data Report 2022 (2023) goes down the same fallacious path: first, it makes unsupported claims of biases and discrimination; then it calls for identity-based quotas on authors and editors and other DEI measures. For example:

The ACS reports that in 2022, 36% of its editors, 41% of EAB members, and 55% of topic editors are women. This representation exceeds the percentage of women in the author pool (women made up 26% of corresponding authors on both submitted and published manuscripts in 2022) and in the field (around 20% in tenure and tenure-track positions, Wang and Widener: Reference Wang and Widener2019), showing evidence of heavy social engineering (one could even consider these data as evidence of barriers for men). Yet, the report continues to call for additional efforts towards the goals of promoting DEI and ‘dismantling barriers to success’.

Other publishers engage in similar activities in order to promote the CSJ agenda. For example, Wiley has introduced a clause in their standard contract with editors that pledges the editor to commit to DEI (Krylov Reference Krylov and Harlaar2022a: 223). The manuscript submission system of PNAS advises authors to ‘consider [the] gender, race/ethnicity, and country of origin’ of potential reviewers (see Figure 6). ACS seems to be considering instituting similar practices:

Figure 6. A screenshot from the PNAS manuscript submission system (underline added).

CSJ activists are also calling for ‘citation justice’ (Dworkin Reference Dworkin, Linn, Teich, Zurn, Shinohara and Bassett2020: 918), and many publishers are dutifully complying (see, for example, Figure 7). ‘Citation justice’ views citations as a commodity to be distributed equitably among various identity groups (Anonymous 2019), rather than the mechanism by which a publication provides background for the reported work and context for the new findings, as well as acknowledgement of previous contributions to the subject. ‘Citation justice’ uses citations as a social justice mechanism by increasing citations of favoured identity groups in the name of ‘equity’, regardless of the actual demographics of the authors in the field. Some journals suggest that authors run their manuscript through software (Sumner Reference Sumner2018: 396) that evaluates the gender balance of cited references (Figure 7).

Figure 7. Screenshot from Public Opinion Quarterly. The Gender Balance Assessment Tool is described in Sumner (Reference Sumner2018: 396).