Research Ethics

Collaboration

Astrophysics depends on large, multi-institutional and international collaborations, and I value them. What concerns me is a division of intellectual labor so complete that it resembles an assembly line. If no one understands a project as a whole, the science becomes hard to check, and so less reliable. This is one reason I care about transparent, checkable deliverables (see AI in Science).

Mentoring and a healthy environment

A mentor is more than a research advisor, and anyone who gives advice, academic or not, can be one. Students and postdocs are vulnerable in several ways, and protecting them is part of responsible research and of building a healthy research community. Mentoring is also highly personal, and should be adapted to each mentee.

I have been lucky to work in respectful environments, but a healthy academia should not depend on luck. Early-career researchers who are bullied are often afraid to report it, so supervisors and colleagues who notice inappropriate behavior should speak up. And safety comes first, even when rules slow research down.

Data and records

I keep notes on almost every part of my research, which I call work logs, and I try to name files consistently. As a theorist and a builder of methods, I run many trial versions of every calculation, and it is not reasonable to store them all forever. The principle I follow is to always keep the data from which scientific conclusions are drawn.

Authorship

Contribution statements are a good way to make authorship clear, though asking coauthors to fill in a formal statement can feel awkward. That is why my publication list says what I contributed to each coauthored paper. How I acknowledge AI is on Working with AI.

Peer review

Peer review can be ineffective, biased, inconsistent and unfair; in my view it is needed only as long as journals are. Preprints are fast and open but less trusted, so for now I see the combination of journals and preprints as a reasonable balance. Most early-career researchers receive no training in reviewing, which journals and supervisors could change by involving them. One idea of mine: authors could prepare a few questions that check a reviewer has read the manuscript carefully.

Conflicts of interest and commitment

I used to think conflicts of interest mattered more for scientists whose work has commercial uses than for astrophysicists. But I have been funded by research grants, and visits and collaborations with other institutions can create conflicts of commitment. Disclosing a conflict is not enough on its own: admitting possible bias does not make anyone objective. Managing conflicts works better, with extra checks where they arise. Rules can have gaps and go out of date, so what we need is a balance that is carefully maintained.

Misconduct

The U.S. federal definition of research misconduct (fabrication, falsification and plagiarism) focuses on what can be documented, but questionable practices can also hide in the invisible parts of the process. Some questionable practices are a matter of degree: splitting work into several papers is fine when a project grows naturally, but overstating the significance of results is a real pressure in publishing. I respect people who expose misconduct, often at a personal cost, and I think tools that help detect problems such as manipulated images are worth building, with human checks to avoid false alarms.