The ideal scientist, we have convinced ourselves, is a robot: dispassionate, in search of an objective truth, embarrassed by any attachment to their own ideas. An ideal so coveted, yet never attained! It is in fact flagrantly violated all the time: scientists deny or explain away every new piece of information depending on how it aligns with their way of viewing things; they write papers pushing methods they’re convinced are superior; they roll their eyes at lines of work they’ve developed a hatred towards; they jump at any opportunity to find tenuous connections to their own extremely important and groundbreaking body of work. Revered scientists in particular are dogmatic and defend their traditions beyond the weight of growing evidence, all in a grand abrasive way. There are modest scientists whom I deeply admire, but who too are, in an endearing way, emotionally attached to their ideas. Of course, I don’t exclude myself from any of these generalizations.

While such narcissistic tendencies at their worst pose hazards to both personal reputation and to the collective—such as when attachment turns into malice, or when the field is led astray by a scientist of influence, or when there are direct consequences to the way society is run—I want to temporarily suspend those concerns and instead explore the other extreme: at their best, what productive function do such tendencies perform in science, and what lessons can we learn from noticing them?

The Courtroom of Science

Much of my argument takes the (admittedly exaggerated) view of science as a courtroom:

The Courtroom is an institution interested in arriving at a collective decision, remaining as objective as it can. The individuals within do not subscribe to this objectivity: the job description of an Attorney is to put forth the best, fullest and strongest version of their client’s case to the court. Thus, in the Courtroom of Science, every Idea—a hypothesis, an algorithm, a theory, a way of viewing things—is a Defendant who has hired for themselves a Scientist as a zealous advocate¹. Time is the neutral Judge that pronounces what is right and wrong, not the advocate.

This is not a prescription. Rather, this is one descriptive lens to make sense of the machinery of science.² Let me add that scientists do try to view their ideas neutrally, but such attempts only succeed at a lower level (such as in calculations, code, data reporting etc.,) and in the initial iterations of an idea. Invariably, there comes a time in the life cycle of an idea, when the idea takes its full form and roots itself into the scientist’s brain, manipulating them into a zealous advocate.

Some paranthetical caveats though: I do not wish to stress as part of this analogy, the combative tone of a courtroom, or the binary nature of legal decisions; we should also be vigilant about scientific integrity, just like codes of conduct lawyers are bound by.³

For the rest of the essay, I want to dwell on the personification of a scientific idea as a “defendant hiring a zealous advocate”. Why would this be a reasonable paradigm of doing science—why on earth is it reasonable that scientists pledge allegiance to their idea, rather than critique all ideas equidistantly and objectively?

---

Reason 1: Uncertainty and subjectivity

When in search of treasures in an uncharted territory, we are forced to take decisions whose fate is unknown. Which directions are just some more effort away and worth gnawing away at; which ones lead to a dead-end soon? What is the best way to cast a certain direction as a concrete research problem? What assumptions are safe to make and what are not? Is the more valuable a simpler, approximate solution or the more precise, inelegant one? What is “simple”? What is “precise”? What is “valuable”?

One must also deal with a second, subtler variety of uncertainty due to the nature of empirical knowledge. How should a scientist interpret an observation when drawing conclusions from it? Of the many variables involved in an experiment, which ones must they take seriously and which ones must they casually overlook? When constructing a plot, what goes on the X axis and what on the Y axis? Given varying pieces of contradicting evidence for a hypothesis, what should be weighed more than the other? Where a connection is unclear (say, between a theory and the real world), how can the missing bridge be speculatively envisioned, so the scientist can continue moving on?

Such uncertainties lurk between the polished lines of every publication—including major discoveries that we now accept as obvious;⁴ the uncertainties of the second type are so delicate that they escape our notice unless we are explicitly looking for them. In all these cases, disagreements arise between scientists, not because one scientist knows what the other doesn’t, but because no one knows! No scientist’s guess is right or wrong a priori; only a posteriori, after exploration, does the ground truth become visible—allowing time to come to a judgement. To get there, an initial playful gamble must be made informed by some heuristic. It is when defining this heuristic that a scientist, whether they are aware of it or not, appeals to their private sensibilities (or to their advisor’s sensibilities), becoming an advocate of that sensibility. ⁵

Reason 2: Efficient exploration

What though is the benefit in having each idea appear with its own attorney in a courtroom? One answer is efficiency: how does a community, tasked with an open-ended exploration into the dark, efficiently cover ground under all the uncertainty? In an authoritarian strategy, the community would deem all uncertain questions as having an undisputed current-best answer (or an undisputed current-best probability over answers) determined by an establishment—a central committee, a revered figure, the zeitgeist or the vocal members of the town hall—and they would shepherd every scientist so that the scientist’s allegiance is towards this centralized understanding, not to their pet idea. As this centralized understanding evolves, so would the direction of the herd. We can contrast this against the more efficient strategy that science offers, which is one where the crowd is let loose, the labor of exploration divided, and every spot on the map gets assigned its own dedicated advocate.

Reason 3: Emotions and Fair chance

But what good is it for the advocate-scientist to be bound to their idea by emotions?

In the Courtroom of Science, for an Idea to have a fair chance⁶ in its fight against other ideas, the Idea finds for itself a zealous Attorney; such a (Dawkinsian) system is put in place by emotions.

The phrase “fair chance” means that in the courtroom of science, no idea is to be dismissed by the judge simply by virtue of its case being half-heartedly built; the idea should be dismissed only after it is challenged at its best.

The word “zeal” needs more justification. First, an idea that has emotionally latched itself onto a scientist, has found for itself someone with the grit and even a level of delusion required to execute that idea, however arduous and bleak the journey gets; a zealous lawyer does not just skim the arguments on the surface, but has the optimism and a “goal-directed” curiosity required to dredge out creative arguments championing the idea; they have conviction in the face of opposition to build a persuasive case for the idea that has hired them. A second argument is the strange human phenomenon of gut feeling: sometimes, especially in theoretical fields, ideas take birth unconsciously in an ineffable form and need nurturing until they attain articulation.⁷ Emotions provide the drive to carry out this the mystical journey from mentalese to material words. A third argument is about attention. Without a scientist’s single-minded focus on their pet idea, idea after idea, and worse, doubt after doubt, would flit in and out of the scientist’s mind, diluting their attention, as if the lawyer keeps inviting one client after the other into their office—including even the oppositon—never committing to a single case.

Reason 4: Unlearning

The above arguments rest on the computational limits of a human. I also hypothesize an “information-theoretic” limit. The premise is that a strong emotional obsession with one’s ideas creates a deep-seated style of thinking (in machine learning terms, an inductive bias) and a sort of negative knowledge (in machine learning terms, a biased training set). The resulting community of scientists becomes an ensemble of highly unique vantage points, that when subject to a tribunal, produces a fuller collective wisdom.

Under this premise, my speculative argument is that a single individual, however computationally-blessed—say even an AI scientist with infinite inference-time compute—cannot simulate this ensemble all by themselves. They cannot simply try to context-switch between various deep-seated views; likewise, simply prompting a single AI model with various contexts (“think like a mathematician”; “think like an artist!”) will not simulate such deep-seated views. In both cases, every thought process may be surreptitiously adulterated by all the biases and knowledge baked into brain/model; such baked-in concepts, I suspect, cannot be unlearned. This is the curse of knowledge. Lifting this curse may require maintaining multiple non-overlapping brains trained independently on different subsets of data—which is precisely what a community of scientists can simulate well.

---

Lessons

A PhD student starts out with belief that their subjective choices begin with “what question do I pick?” and end with “what solution do I come up with?”, with everything else governed by rigid logic. Likewise, the PhD student recognizes their role as a lawyer only in a limited sense: at the moment they subject their work to peer review. Maturity comes from noticing all the deeper ways in which research involves personal interpretations and choices, and from noticing that in those contexts interspersed throughout the life cycle of an idea, one unknowingly behaves like a lawyer. Once noticed, it allows the scientist to be intentional in their choices; it allows them to examine their own implicit style and mould it over the years; it allows them to resist losing themselves in the wild stampede around them.

On a personal note, the courtroom view has helped me deal with the thorny interface between the external world (disagreements, rejections) and the internal world (confidence, humility and self-doubt). When in heated disagreement with someone, the courtroom is a lens to pause and observe our respective behaviors from a distance, even if momentarily, and even if I continue to disagree. The courtroom frames disagreements as a feature in the grand scheme of Science; it creates a safe space (towards others and myself) by reframing “being proven right/wrong” as “being (mis)fortunate in the gamble of Science”.

The courtroom tempers my internal world and be strategic about my emotions. I am less often at the mercy of my own nature, vacillating between utter self-doubt and utter delusion in my idea (which is dangerous to be unaware of). I can be strategic about when I emotionally attach myself to an idea: done too early, I end up sinking costs into a dead-end; done never, I never let an idea materialize. I am also better at consciously calibrating between conviction (that I may be proven right even if the other advocate sounds confident in their opinion) and humility (I may be proven to be wrong, and if so, on what precise accounts). Indeed, I can also very well simulate the extremes: the sense of delusion needed to explore an idea while knowing fully well that I need to doubt my own ideas at opportunte moments.⁸ All this comes from being aware of which decisions are open-ended and therefore require a tribunal where Time can tell what is right or wrong.

---

Notes

---