The Ethical Debt: a preview of the book I'm finishing

A book, and its one-sentence argument

For the past months I have been writing a short argumentative book — around 30,000 words across ten chapters — that gives a name to a problem I keep meeting in preclinical laboratories: the moral deficit created when animals are used to produce knowledge but the resulting data are allowed to become unusable. I call it the ethical debt, and the book's one-sentence argument is that data welfare is animal welfare.

The reasoning is simple. Every animal experiment is a bargain across time: the moral cost is paid now, in full, by the animal; the scientific value matures later — through replication, pooling, control reuse, method validation and, increasingly, model training. When the data lose that future value — left unfindable, undocumented, or impossible to reuse — the bargain is broken after the fact, and the debt falls forward onto the next animals, because the only way to regenerate lost knowledge is to run the experiment again.

The book grows out of a peer-reviewed argument my colleague Benoit Petit-Demoulière and I published — that data stewardship is not a technical afterthought but part of the ethical responsibility attached to animal research [1]. The book takes that thesis and follows it, chapter by chapter, through the machinery of modern research.

When data die (Chapters 1–2)

The opening chapters borrow a metaphor from software — technical debt — and its darker cousin from AI ethics: the insight that those who incur a debt are rarely those who pay it. A dataset can survive as files while dying as evidence. I describe six ways this happens — loss of provenance, context, meaning, interoperability, access and trust — and five of the six leave every byte in place.

This is not hypothetical: the availability of research data declines by roughly 17% per year of article age [5]. The aggregate bill is large — irreproducible preclinical research is estimated to cost the United States alone around US$28 billion a year [2] — but long before it is money, in animal research it is lives.

The 3Rs in the data age (Chapters 3–5)

The middle of the book returns to Russell and Burch's 3Rs — Replacement, Reduction, Refinement [10] — and extends each one to the fate of the data rather than only the design of the experiment. Reduction is the clearest case: better power estimates, avoided repetition and reused controls all depend on evidence from animals already used — and historical control data can more than halve the control animals a new study needs [7]. Poor stewardship of past data therefore forces future animals to repay a debt they did not incur.

The same logic reframes publication bias: when only about half of animal experiments are ever published [11] and the published record is skewed toward positive results [3], other laboratories unknowingly repeat what could have been learned — exposing new animals to rediscover an absence of effect. Publication bias, in animal research, is a welfare problem and not only a statistical one. The stakes are not abstract: on a retrospective harm–benefit analysis of six intervention areas, fewer than 7% of the preclinical studies were judged ethically permissible [4].

Repayment: FAIR, VCGs, NAMs and a data-index (Chapters 4, 6–8)

The book's practical spine is repayment. It starts with metadata as ethical infrastructure and the FAIR principles — Findable, Accessible, Interoperable, Reusable [9] — argued as a condition of ethical data generation rather than an administrative label, and made concrete through minimal metadata sets designed to make animal data reusable [13]. The chapter does not pretend this is solved: across the medical literature, roughly 8% of papers declare data are available but only about 2% actually share them [6].

Two later chapters make the argument tangible. Virtual control groups reuse historical control data instead of breeding a fresh control arm, with an estimated reduction of up to a quarter of control animals [8] — but only where the historical animals are described well enough to belong in the comparison, and early validation shows they reproduce overall conclusions while only partially matching individual findings [15]. New Approach Methodologies get the same treatment: a replacement method earns its place by validation against trustworthy reference data — overwhelmingly animal-derived — so wasting that data can slow replacement itself.

A further chapter asks why none of this is rewarded: data sharing appears in only about 1% of biomedical promotion and tenure criteria [14]. The book sketches a speculative counterweight — a “data-index” that would ask not how often our papers are cited but whether the data generated from animal lives continue to serve science — framed as a review practice, never a single number to be gamed.

A WellFAIR charter (Chapters 9–10)

The ethical debt is “nobody's fault and therefore everybody's” — it accumulates in the space between roles, so no single actor can repay it. The closing chapters turn the argument into commitments by audience — researchers, ethics committees, journals, funders — and lean on a hard lesson from reporting standards like ARRIVE 2.0 [12]: a standard endorsed in name and ignored in practice changes very little. The organising principle is one line: animal ethics should ask not only how many animals will be used, but what will happen to the data those animals generate.

What the book is not

It is worth being clear about the boundaries, because they matter. The book is not an argument against animal research; the claim is narrower — that if animals are used, a duty follows to keep the resulting knowledge alive. It does not argue that all data should be open: sensitive and proprietary datasets can be ethically alive under controlled access, and openly dumped data can be ethically dead. It does not claim virtual control groups can replace live controls in general, nor that a data-index should become a new scoreboard. And it is honest about its own frontier — the chain from better stewardship to measurably fewer animals is argued, not yet measured.

Coming soon

The book will be published shortly, and I will share the release here and on LinkedIn. If the argument interests you — as a scientist, a reviewer, a funder, or someone who simply cares where animal data go after the paper is filed — this note is the shape of it. The debt is repaid not by a breakthrough but by a habit. Every animal leaves data behind; we have a duty to make those data count.