[cross-posted from my blog https://pchvykov.com/blog]

I see a problem with current science. It’s not the reproducibility crisis, nor the toxic work culture, nor the misaligned incentive. But in another sense, it is all of these – or perhaps the root cause behind them. It’s hard to name it exactly, but in a way, it’s the dissociation between the romanticism of a selfless “quest for truth,” and the career-success incentives that run academia. In another way, it’s Goodhart’s law (optimizing for citation counts rather than for meaningful progress). In yet another way, it may be an issue with the scientific method itself (or at least how it’s taught). But I think the overarching issue is with trying to remove our humanity from the scientific process. 

Let’s begin with something concrete – Goodhart’s law (“when a metric becomes a target, it ceases to be a good metric”). This is a major (perhaps the major) problem of the western social order, which relies heavily on metrics. We optimize for GDP, and forget about actual well-being and psychological happiness of the population. We optimize for profits, and forget about the social value we are creating. We optimize for grades, and forget about learning. We optimize for longevity, and forget about meaning. And so in academia, we optimize for citations, and forget about progress, about building a better world. Goodhart’s law comes up because the things we actually value cannot be accurately encoded into metrics – they will always be slightly misaligned. People’s ingenuity will then always find a way a way to leverage this misalignment to game the metric. But metrics and reproducible measurements are at the core of the scientific method, thereby seeding the problem (it's easier to write a paper showing some high metric score than one that actually matters).

At the same time, upon reflection we somehow have an intuitive understanding of whether something is or isn’t a “good” metric, or whether it achieved some “desired” outcomes. So what metric inside us tells us what’s good? For this I like the dialectic of “measuring” vs “sensing” – the former being about finding the value of a pre-defined metric, and the latter about pattern-recognition. Perhaps one can say that in these terms, science is more about measuring, while art is about sensing? Either way, measuring has a linear quality to it – it’s the “forward path” of the scientific method. In contrast, sensing is inherently a feedback loop – where question informs the answer, which in turn changes the question, and so on. Such dynamics may mitigate Goodhart’s law as the metric  (i.e., question) here is never static, but continuously updated and refined in response to the outcomes. The problem I’m referring to is insufficient amount of this feedback in science and its application.

But I think we must take it deeper than this. We must include not just the question, but the observer themselves in this feedback loop. To see this, we need to ask ourselves honestly why do we study the questions that we do? The real reason is often circumstantial – “my PhD adviser worked on this,” or “there was a job available for this research,” or “this topic sells well.” But if we go beyond all this and honestly ask ourselves which questions are really important? Which questions are worth our time, our effort, and money? Which questions really make meaningful progress towards a better world? Such sincere inquiry quickly makes us realize that the answers are far from obvious, are very subjective, and are highly sensitive to personal and cultural values, traumas, fears and hopes. And these are the foundation of our science. Thus, we cannot pretend to be the “objective observers,” standing outside the scientific method – we are part of it. And as such, we must have the humility to ourselves become subjects to it. On the one hand, thus goes back to the willingness to update our beliefs about the world, our behavior, our personality, our sense of self even, in response to new evidence – which is already hard enough. But on the other, it also means asking the questions that actually matter to us, that actually have the potential to change our lives (cf. active learning in ML). I think this is the only way to really do “honest science.”

And once we come to this, we basically come to spirituality (in some idealized sense). If science is the study of the external, then spirituality is the study of the internal (-Carl Jung, Carl Sagan, Fritjof Capra, etc.). But if we allow our science to be guided by the quest for personal transformation, for greater joy, and for a better world, then the distinction begins to blur. I believe that this integration of science with spirituality, with our humanity, with our inner goals and aspirations, and paradoxically, with our subjectivity, is the only way to overcome Goodhart’s law and get fulfilling outcomes. The notion of a separate objective observer in the scientific method is an impossible idealization – and therefore misleading, leaving much of academia to study incremental technical minutiae that has little relevance to our lives. Even the relevant technological achievements often end up divorced from generating greater well-being. It is no accident that many of the early scientists were motivated by their faith – they studied the external to better understand the internal. I find that the schism between the scientific and the spiritual that emerged since has been unproductive for both domains, just as any hard disciplinary boundaries lead to siloed inefficient work.

So what do we do? Well, I’m not really sure. The rationalism movement may be on the right track here. Personally I think it would be fun to build a research institute that really focuses on enabling this feedback between research results and our personal inner values. I’ve been quite interested in Eastern Philosophy and mindfulness practices for a while – and these might give a good approach to really help learn from and internalize our scientific insights. On the other hand, complexity science (my research field) may be a good framework to research all these questions more systematically, perhaps in the context of “science of science.” The key, I think, is to integrate the theory and the practice here – to “walk our talk.” I would love to see science become a joyful and deep practice that leads its adepts to ever-greater personal fulfillment and wisdom, not mere knowledge. Scientists serve the role of shamans in modernity, and so their wisdom and personal attainment (or lack thereof) spreads to the rest of the human tribe.