Saturday, September 12, 2026

When Science Retracts Its Word

 

When Science Retracts Its Word

On truth, trust, and scientific integrity

I am leafing through Nature when I stop at a page I would normally have passed over.

The heading begins with a single word:

Retraction.

A scientific article has been retracted.

The article was published in Nature in November 2024 and concerned how the protein HDAC6 senses the abundance of the amino acid valine and regulates DNA damage. The biology is complicated for me. I have no qualifications for assessing the experiments themselves.

But I understand the retraction.

Because it concerns something far more fundamental than HDAC6 and valine.

It concerns trust.

After publication, questions were raised about the authenticity of the raw data underlying several of the figures. Unusual similarities were found within and between figure panels that were supposed to report results from different experimental conditions. When the matter was investigated further, concerns were raised about additional raw data as well.

The authors were given an opportunity to explain. Tongji University investigated the matter. But the editors did not find the explanations satisfactory. The concerns about the authenticity of the data were, they wrote, extensive and persistent, and the independent confirmation of the article’s conclusions was incomplete.

Then comes the decisive sentence:

The Editors, therefore, no longer have confidence in the data reported in this article.

The editors no longer have confidence in the data.

The article is therefore retracted.

Science and trust

We often speak of science as knowledge.

That is, of course, correct. But science is also an enormous system of trust.

The researcher must be able to trust the measuring instrument. A co-author must be able to trust that a colleague’s data actually come from the experiment being described. Peer reviewers must be able to trust that the material they are assessing is authentic. The journal must be able to trust the authors. And we, as readers, must be able to trust that what is presented as research really is research.

No one can check everything personally.

I can read an article about DNA damage in Nature, but I cannot enter the laboratory and repeat the experiments. I have neither the expertise, the equipment, nor the raw data.

Yet I can relate to the knowledge it presents.

Why?

Because I place my trust in a chain of people and institutions I have never met.

Scientific integrity is therefore not merely a question of the individual researcher’s morality. It is a precondition for the entire system of knowledge to function.

Trust is not the same as belief

Scientific trust, however, is a peculiar form of trust.

It requires us to trust someone. At the same time, it requires us not to trust them too much.

Science does not say: This is true because the professor says so.

It says: Here is the method. Here are the data. Here is the argument. Examine them yourself.

This is what distinguishes scientific trust from faith in authority.

Trust in science must always contain the possibility of distrust. A result must be open to scrutiny. An experiment must be reproducible. Data must be available for examination. A conclusion must be open to challenge.

That may sound paradoxical.

But perhaps it is precisely organized doubt that makes scientific trust possible.

When something does not add up

Someone, however, has to be the first to say:

Something is wrong here.

Perhaps two images look suspiciously similar. Perhaps the numbers do not correspond to the raw data. Perhaps an experiment cannot be reproduced. Perhaps a reader notices something that the peer reviewers overlooked.

It is an uncomfortable moment.

Because science is not only method and data. Science consists of people. Colleagues. Research groups. Departments. Universities. Careers. Prestige.

To question a published study may therefore also mean questioning people with whom one works.

This is where research ethics becomes practical philosophy.

It is easy to say that truth must come first.

It is harder when truth has a social cost.

Integrity

The word integrity comes from the Latin integer: whole, intact, untouched.

In research ethics, we commonly use the word to mean honesty or probity. But integrity involves more than following a set of rules.

It concerns what we do when different obligations collide.

Loyalty to a colleague stands against loyalty to research. Concern for the institution’s reputation stands against the demand for openness. The desire to protect work to which one has devoted years stands against the obligation to acknowledge that the result may no longer hold.

It is then that integrity is tested.

Not when doing the right thing is easy.

But when it costs something.

Peer review is not a truth machine

There is something else about this small page in Nature that fascinates me.

The original article had passed peer review.

It had been published in one of the world’s most prestigious scientific journals.

And yet the story was not over.

This matters.

We can easily begin to treat peer review as a stamp of approval meaning:

True.

It does not.

Peer review is a mechanism of quality control. An important one. But reviewers cannot necessarily detect problematic raw data. They can overlook errors. They can be wrong.

Scientific knowledge therefore does not derive its strength from having been approved once.

Its strength lies in the fact that the approval is never final.

Popper and the right to be contradicted

Karl Popper made fallibility a fundamental feature of science.

A scientific claim must be capable of encountering something that speaks against it. But perhaps the most interesting aspect of this case is not the textbook formula of falsifiability. It is what happens after something has begun to speak against a claim.

Data do not defend themselves. Figures do not admit mistakes. Institutions do not automatically correct their own publications.

Someone has to notice that something is wrong.

Someone has to investigate.

And someone has to be willing to allow the objection to have consequences.

The strength of science therefore lies not only in the fact that its claims can turn out to be wrong. It lies in the existence — or at least the necessity — of a space in which this can be said aloud.

That is an important distinction.

A scientific culture can celebrate critical thinking in theory while making life difficult for the person who actually criticizes.

Then the problem is no longer primarily one of philosophy of science.

It is practical and moral.

But self-correction is not automatic

It is tempting to read a retraction as proof that the system works.

The problem was discovered. The article was retracted. Science corrected itself.

But the short text in Nature tells a more complicated story.

Three of the authors agree with the retraction. Five explicitly disagree. Sixteen did not respond to the editors regarding the retraction, and the editors were unable to contact one author.

It is a remarkable aftermath.

Because here we see that the expression “science corrects itself” is really a shorthand.

Science is not a person who wakes up one morning, discovers an error, and corrects it.

People have to do that.

In this case, it was the editors who ultimately stated that they no longer had confidence in the data. They retracted the article even though several of its authors opposed the decision.

The self-correction, then, did not consist in agreement.

It consisted in someone taking responsibility for acting when agreement was absent.

That may be even more important.

For if criticism is silenced, if the reputation of the institution becomes more important than the investigation, or if collegial loyalty becomes stronger than loyalty to knowledge, science does not correct itself.

Then the system protects itself.

The difference is crucial.

Being able to retract one’s own word

And so I find myself looking again at the word Retraction.

In this particular case, we must be careful not to read the heading as though the authors themselves had withdrawn their work.

They did not — at least not collectively.

Nature retracted it.

And precisely for that reason, the word becomes interesting.

Because it points towards an ideal larger than this particular case.

To practise science means not only daring to make a claim. It must also mean accepting that the claim may be examined after it has been published. That others may inspect the data. That questions may be asked again. That prestige cannot make a conclusion immune to doubt.

It is not necessarily a weakness to say:

This is what we thought we knew, but we can no longer stand behind it.

It can be a strength.

But this case reminds us that such strength does not arise by itself.

Sometimes the person who made the claim is unwilling to withdraw it. Then the institutions surrounding science must be strong enough to allow the question of truth to outweigh the question of who wishes to stand by their word.

For the opposite of science is not doubt.

The opposite of science emerges when doubt is no longer allowed to do its work.

Perhaps this applies beyond the laboratory as well.

To seek truth does not mean being infallible. It means being willing to allow what we already believe we know to be examined again.

Being able to retract one’s own word when one can no longer stand behind it is therefore not a description of what everyone did in this case.

It is the ideal that the case reminds us of.


To seek truth does not mean being infallible. 
It means being willing to allow what we already believe we know to be examined again.

This essay was written in a conversation with Claude/Anthropic and OpenAI/ChatGPT

No comments:

Post a Comment