Swords into Plowshares: On Open Science and Morals.
Photo by Gioele Fazzeri on Unsplash
One of my core beliefs is that the search for truth is necessarily collaborative. No single study is ever the answer; at best it becomes a contribution to an ongoing conversation that stretches across contexts, disciplines, and time, and that continues long after the original authors have moved on to the next project. For that reason, I have always been a strong supporter of replication, particularly the honest kind that asks under which conditions previously observed results continue to hold and under which conditions they quietly stop doing so. Replication, understood that way, is not an attempt to catch someone out or to stage a methodological ambush. It is an attempt to understand the boundaries of a claim.
We recently did something along those lines ourselves. We published a paper that replicated a study conducted with British entrepreneurs by running a comparable study with entrepreneurs in Germany, holding as much constant as one reasonably can while still accepting that contexts are not interchangeable containers. The result was perhaps the least scandalous headline imaginable: the findings differed. This should not have been surprising. Although I genuinely believe that Germans and Brits have far more in common than either side sometimes assumes, institutional, cultural, and structural differences remain, and those differences do not politely disappear simply because a regression model would prefer them to.
What is noteworthy, however, is what we did not conclude. We did not declare that the original study was wrong, nor did we claim that we had “failed” to replicate in a way that would somehow diminish the earlier work or expose it as defective. Quite the opposite. The divergence suggested that there are institutional context factors at play that neither study captured fully, and there are always institutional factors. The fact that we do not measure them all does not invalidate research; it merely reminds us of its situated nature. If anything, the combination of both studies paints a richer and more differentiated picture than either one could have produced on its own.
To me, this is what Open Science at its best should look like. Not a game of gotcha, not a public ranking of who replicated whom more successfully, and not a scoreboard of methodological virtue, but a cumulative effort to understand where claims travel well and where they begin to bend under contextual pressure.
I have always considered myself a fairly strong proponent of truth. Not in the heroic, Indiana Jones sense of the word, although, strictly speaking, he is a university professor, but I was informed by HR that I am not allowed to bring my whip to work, which limits the cinematic dimension of my commitment. Mine is of a far less dramatic variety. I am attached to truth in the statistical sense, in the quiet and somewhat unglamorous belief that if we gather enough memory, enough data, and enough patience, we might make decisions that are at least slightly less wrong than the ones we would otherwise make. This is not a revolutionary ambition. It does not promise redemption or intellectual purification. It promises incremental improvement, and that has always seemed realistic enough for science.
Recently, I was installed as Open Science ambassador for my school, JADS, at Tilburg University. You can find the interview, which in the finest academic tradition I largely conducted with myself, here. The appointment forced me to articulate what exactly I believe Open Science is, what it is becoming, and perhaps more importantly, what it might risk becoming if we are not attentive to its tone and trajectory. This post is the result of that reflection.
Science has always required a peculiar kind of collective faith. We write papers knowing that most of them will be read by a handful of people, two of whom are reviewers who may already dislike the premise before reaching page three, and yet we persist. We persist because there is a shared assumption that the pile grows, that memory accumulates, and that across decades and disciplines this accumulation produces something that deserves, cautiously and imperfectly but still meaningfully, to be called progress. The individual contribution is usually small and occasionally embarrassing, but the collective project retains its gravity.
Open Science emerges quite naturally from that logic. If better memory enables better decisions, then more transparency, more documentation, and more accessible data should strengthen the enterprise. The argument is, at its core, almost disarmingly reasonable. Share what you did. Make your steps visible. Allow others to build on your work without having to reverse engineer your spreadsheets at two in the morning while quietly resenting your column names. None of this sounds radical. It sounds like good housekeeping for a knowledge system that has grown large enough to misplace its own files.
And yet Open Science has grown beyond housekeeping. It has developed its own language, its own rituals, and its own symbols of virtue. There are canonical readings, preregistration norms, badges, repositories, and increasingly clear signals about what counts as proper conduct. There is rarely a day on which I do not read an abstract that already announces, almost ceremonially, that the study is a “preregistered experiment,” as though the label itself were part of the empirical contribution. While preregistration is often valuable, and sometimes essential, the performative emphasis can feel uncomfortably close to virtue signaling, as if methodological commitments were gradually becoming markers of moral standing rather than tools for improving inference.
This, in fairness, is not surprising. Any movement that aims to reshape institutional practice requires shared norms and a shared vocabulary. Encultification is how collective change happens. People repeat certain practices, internalize them, teach them to their PhD students, and eventually begin to judge deviations from them.
The difficulty begins when the boundary between methodological preference and moral virtue becomes blurred. When language becomes a signal of belonging rather than a tool of clarification, when procedural commitments turn into proxies for personal integrity, and when disagreement is interpreted less as epistemic diversity and more as ethical deficiency, a movement that began as a pragmatic reform risks acquiring the tone of moral judgment. Feeling righteous is an easy temptation in academia, particularly when one sincerely believes to be on the side of improvement.
At this point, religious history offers a cautionary tale. Belief systems that attempted to expand primarily through coercion, whether in the form of crusades, inquisitions, or other spectacles of enforcement, were capable of producing compliance and fear, but rarely deep and lasting conviction. Durable transformation did not arise from the sword. It arose when missionaries entered local communities, learned their language, embedded themselves in existing social structures, and demonstrated through lived example why a particular set of practices might be coherent and meaningful. They did not begin by demanding purification. They began by translating.
If Open Science increasingly resembles a gospel in its vocabulary and normative ambition, then it would be wise to consider which historical strategy it wishes to emulate. Crusading for methodological purity, publicly shaming deviations, or treating colleagues as obstacles to reform may generate visibility and a certain emotional satisfaction, but it is unlikely to produce durable cultural change. What may prove more effective is the less glamorous work of translation, entering different methodological cultures, understanding their constraints and incentives, and demonstrating patiently how certain practices improve collective memory without demanding immediate conversion. Reform tends to spread most sustainably when it is locally embedded and voluntarily adopted rather than externally imposed.
In my earlier post, Causal-ish, I argued that science rarely delivers certainty. It delivers approximations that are useful under specific assumptions and fragile outside them. Models are simplifications. Claims are conditional. Progress is real, but it is rarely pure. If that is the nature of our epistemic enterprise, then Open Science should embody the same humility. Its purpose is not to purify science or to divide the righteous from the misguided. Its purpose is to help the system remember better and, as a consequence, to help it err slightly less often.
If Open Science has indeed become a kind of gospel, then it should behave like a mature one. Maturity means resisting the temptation to define oneself primarily through denunciation. It means recognizing that cultural change is not sustained by perpetual outrage, nor by the ritual exposure of the next big fish to grill or the next glamorous name to sacrifice on the altar of replication. I have never been particularly fond of watchdog blogs whose energy is devoted almost exclusively to uncovering failure while only rarely acknowledging improvement, as though progress were somehow less interesting than scandal. Critique is indispensable. But when critique becomes identity, it narrows the very community it seeks to reform and risks turning reform into spectacle.
A movement that aims to strengthen science should derive its cohesion not from public shaming but from demonstrably better practice. If our shared ambition is to improve how we remember, how we document, and how we build on one another’s work, then our tone and strategy should reflect that ambition. And so, in the best biblical sense, perhaps it is time for us, as Open Science advocates, to turn our swords into plowshares, to lay down the instinct to police and instead cultivate, and to use those tools to work productively on the vast, complex, and imperfect field of science that we all share. The study of British and German entrepreneurs did not weaken either side of the conversation. It enriched it. That, to me, is what careful cultivation looks like. Redemption is not required. Incremental improvement will suffice. In a discipline devoted to approximation, that is ambition enough.