Why a data professional oath (even when we do not need it to behave)

The moment a data project actually mattered was always the same one. Someone specific, whose name I sometimes did not know, got treated differently that week because of a number I had shipped: a candidate placed or not placed, a patient higher or lower on a waiting list, a customer offered one price rather than another. From what I have seen in the field, that is where the ethics of this work lives, well before it lands in a policy document. And it is the question I keep returning to after twenty years of BI work across staffing, healthcare-adjacent reporting and retail analytics: not whether the model was accurate, but whether the person on the other side of it was still a person in the way I designed it.

Where the oath comes from

A version of that question was published formally in 2018, as an appendix to the National Academies of Sciences, Engineering, and Medicine report on undergraduate data-science education. The Data Science Oath is modelled on the Hippocratic Oath, covers eight values (Dignity, Interpersonal Trust, Justice, Self-Control, Autonomy, Reliability, Responsibility and Trustworthiness), is non-binding, and has not been adopted profession-wide. As Baumer, Kaplan and Horton note in Modern Data Science with R, there is still no official data scientist’s oath the field as a whole has taken up. Seven years on the shelf: either a failure of adoption, or a standing invitation.

Why the oath earns its place

From where I stand, most data professionals do not need an oath to behave. The people doing this work are, in the main, careful people. What the seven-year gap reveals is that the profession has not yet agreed on what carrying an oath actually looks like in practice. A value named “Dignity” is easy to write down. The working question it maps to is more specific: at 3pm on a Tuesday, in a stand-up where someone proposes merging a postcode with a health flag in a hiring filter, is the person on the other side of that decision still a person in the way you designed it?

That is where the oath earns its place: as a habit that keeps the “who” in view when the calendar fills up and the reporting layer starts to feel like the whole job.

Orientation versus practice

DJ Patil, Hilary Mason and Mike Loukides made this point in 2018: an oath sets orientation, but a habit, a working question carried into a project meeting, is what changes behaviour on a Tuesday afternoon. Virginia Eubanks sharpened the same observation: the harm patterns already documented in welfare systems, policing algorithms and child-services scoring did not come from practitioners who ignored an oath. They came from practitioners who never turned the orientation into a question they could actually ask in the room.

The cohort I am running over the summer is an attempt to do that translation in the open.

Eight values, one working question each

Between July and September, I am taking each of the eight oath values out one at a time on LinkedIn and on this site. One value, one working question, one post. Dignity on 17 July, Interpersonal Trust on 22 July, Justice on 28 July, Self-Control on 4 August, Autonomy on 16 August, Reliability on 21 August, Responsibility on 5 September, Trustworthiness on 11 September, and a closing piece on 12 September that uses the oath as a mirror: eight values, eight questions, what actually shifted.

Dignity becomes: “Is the person on the other side of this decision still a person in the way I have designed it?” Interpersonal Trust becomes a question about what I would explain to the person the decision affects. Justice asks whether the distribution of outcomes is something I would stand behind if it were made visible. The remaining five follow the same pattern: abstract value, concrete question, one meeting at a time.

A working diary in the open

A signed oath hung on the wall becomes a certificate; the only version that shifts anything is the one you practise on a Tuesday afternoon, in a real meeting, with real trade-offs on the table. This cohort is that second version made public, so anyone doing the same work can compare notes and disagree out loud. There is no badge at the end, no certification, no compliance product. The Data Science Oath is non-binding, and this series is not legal advice. What I am offering is eight working questions drawn from eight values, practised over nine weeks, so the orientation the oath sets has somewhere specific to land.

Where to start

Pick one of the eight values and try the working question it maps to on your next project meeting, before anything else lands on the agenda. Notice what shifts in the room and what does not. That is the whole method. The posts between now and September run the same practice across all eight values, so you can follow the threads as they develop or come back to any single value on its own.

Happy to think this through with you.