A Pledge for Scientists
In the wake of yet another major conference getting swamped by too many submissions, and many participants using AI in irresponsible ways,1 I would like to suggest a simple social fix:2 Like the Order of the Engineer, our field should also have a pledge, serving as some form of commitment device. My (arguably naïve) idea is that such a pledge will help us maintain consistency and remind us about our values, as well as our intentions.
Shaped by discussions with many of my collaborators, in particular Richard (who suggested such a pledge in the first place) and Mathieu (who suggested adding sections on responsibility and disagreement), here is a first version of such a pledge. I hope it will serve as a building block and potential source of inspiration.
In my work as a scientist, I will strive to never forget that what I do has consequences in the world. I make these commitments freely and will hold myself accountable to them.
On honesty. I will report what I find, not what I wish I would find. I will be honest about my findings, my methods, their limitations, and how they relate to the works of others, refusing to inflate or exaggerate. Whenever possible, I will share my code and data. I will describe my findings as transparently as possible, making it possible for others to build on them, reproduce them, or even challenge them.
On responsibility. I will own any errors or mistakes in my work, even the ones introduced by my tools. I will not let fear or false pride prevent me from revising false conclusions since I recognize that corrections are a natural part of the scientific method.
On disagreement. I will engage seriously and respectfully with critics and criticism. I will likewise criticize ideas without criticizing the people who express them, and I will remain open to being persuaded.
On tools. I will learn how to use my tools in the most effective and efficient manner possible so that they do not use me. In particular, I will treat AI systems as useful tools, which nevertheless require my judgment. I will verify what they produce and will not let them speak for me, even though it would be expedient to do so. As for the tools I create, I will think about those who might be harmed or helped by them. I will refuse to build something that I believe would do more harm than good.
On credit. I will acknowledge those who served as inspiration or precursors to my work, just as I will acknowledge those who undertake my work with me. I will never knowingly claim that as mine which was originally created by others, be they human or machine.
On community. I will remember that I am a member of a community that extends well beyond my daily work. I will make sure that the trust in science and the scientific method remains warranted.
This I pledge, in the full knowledge that I am fallible. I know I may sometimes fall short of these ideals, but I commit to acknowledge it and try again whenever I do.
-
Including, but not limited to: (i) Providing a framing that is too broad, (ii) providing responses in a rebuttal that do not answer a question but instead wax lyrical about unrelated points, and (iii) “spamming” reviewers (or authors) with long comments that have not been read by a human. ↩︎
-
Even though the efficacy of such a fix is very much debatable but having something is probably better than nothing in this particular instance. ↩︎