Curiosity is a hallmark of any scientific endeavor. Wonderment over how molecules interact, how species evolve, or how stars are formed and die, for example, has guided the work of generations of scientists. When sensory abilities have fallen short, tools such as microscopes and telescopes were invented to explore beyond what we could see. When geographical constraints made it impossible to study individual particles or animals in the world, statistics were derived to make generalized claims about the population from a smaller sample size. Sometimes, scientists have made discoveries unintentionally - like the discovery of the antibiotic penicillin, which came about as a result of Alexander Fleming’s curiosity regarding a bacteria-killing fungus growing in an unwashed petri dish. Other times, scientists have a stronger intuition of what they wish to understand and accordingly have invented tools throughout the generations. But what about a phenomenon so complicated that we wouldn’t know to seek it, or wouldn’t recognize it even if we chanced upon it? What if we had all the information we needed, but were unable to connect the dots of the pattern constructed by nature? Or worse, what if we think we recognize the pattern, but this understanding is tainted by the limits of human cognition? For example, were we to attempt to decode how dogs communicate, would our understanding be biased and/or limited to thinking in terms of human conversation - looking for nouns and verbs among whimpers and barks?
We don’t even need to look outside ourselves to find examples of the potentially non-understandable. Princeton cognitive neuroscientists Uri Hasson and colleagues, for instance, make an argument that for years we have mistakenly segregated the complexity of the brain, studying each of its components individually, and breaking it down into relatively simplistic assumptions and models. In doing so, there was hope that by understanding the simpler, individual parts comprising the brain, we could one day comprehend the entire system as a whole. But Hasson claims that this approach will never really lead us to a truly comprehensive understanding of the brain, but rather a mirage of it, altered by the lens through which we view this enormous complexity - namely, through its individual components. It is important to note that this is not an argument against studying the smaller constituents of the brain; rather, the argument is that no matter the granularity with which we study brains, the ways in which we seek to understand the brain can themselves corrupt our understanding of it. For example, one can imagine attempting to characterize the sizes of fish in a pond, but being limited to using fishnets with only big holes - in such a case, one’s understanding of fish sizes would be adulterated by having missed out on all the fish too small to be caught by the net. Indeed, having no access to ‘nature’s playbook’, our attempts at understanding the brain will always depend in part on how we choose to approach it, and Hasson argues that an approach too reliant on simple models of individual brain regions may lead us astray.
For a more concrete example, consider how we currently understand mental disorders. The Diagnostic & Statistical Manual for Mental Disorders (DSM), referenced by psychiatrists to diagnose and treat patients, categorizes disorders based on symptoms. Our cognitive constraints make it hard to think of disorders on a spectrum with unclear boundaries, and so we rely on oversimplified albeit comprehensible categories, not just for ease of prescribing disorder-specific care, but also for patients, doctors, and society alike to label and put into words as to how people can be atypical. However, this propensity to oversimplify has often undermined the intended usefulness of the DSM as a tool, at times wrongly diagnosing patients and thus causing confusion and potential harm by prescribing inappropriate remedies. The field of psychiatry has since made efforts to approach understanding of mental disorders in terms of the varied overlapping dimensions that they occupy; yet, undoing the entrenched ideas of disorders as distinct categories, both as scientists and as a society, is no easy feat. Psychiatry must continue refining our understanding of disorders to provide robust contributions to mental health.
Several other examples in neuroscience abound. In 2014, the Nobel Prize in Medicine or Physiology was awarded to Drs. May-Britt Moser, Edvard Moser, and John O’Keefe for discovering ‘place’ and ‘grid’ cells: neurons that activate when an animal is in a very specific location, and not otherwise. Yet, since then, scientists have realized that these cells didn’t in fact just encode locations, but instead any relational metric that is regularly encountered. For example, just like locations are relational to one another on a map, individuals can be organized relative to one another in a social network, and sounds can be encoded relative to each other’s pitch. Indeed, ‘place’ cells have been shown to encode both social hierarchy and frequencies of sound in animals that regularly use these metrics—humans and rats, respectively. Using a particular common instance of memory processes, e.g., navigation, led us to an incomplete picture of what these neurons encode, emphasizing that how we pursue questions influences what answer we land on. Yet, these examples show that with time and thought, we have the ability to recognize our biases and tweak how we approach and answer scientific questions.
Nonetheless, it is difficult to know just how big our blind spots are, and how many supposedly answered questions may instead constitute mere doppelgängers of reality. Further, as-of-yet unanswered or unexplored questions are likely to be increasingly complex, and may challenge our intuitive notions of the nature of reality. How then must we approach understanding? How can we find truer approximations of reality with mental models that will inevitably be constrained by our cognitive abilities? And when our abilities fail us, how can we be sure to recognize this failure, and to not get blinded by the mirage of reality?
On one hand, scientists should not solely rely on simple-to-understand models. Scientists could instead embrace models that are inherently complex and perhaps not fully understandable, but which can perhaps map onto the equally complex natural phenomena that we seek to explain. Where we cannot surpass our cognitive constraints to conjure complex models, we can look to nature for inspiration - indeed, neural networks and quantum computing are just such examples. While we may never fully understand every single component of these models, we can understand the ingredients we use to build them, and analyze the output using the heuristic, ‘if it looks, swims, and quacks like a duck’ - or, in our case, if it appears to be a good model by accurately reflecting the phenomenon we are investigating - then it probably is one. This can be thought analogously as not needing to come up with a completely original recipe, but instead being given a hint of it by nature, and then studying and adjusting it to make the dish and understand the phenomenon. These models are therefore capable of helping scientists better understanding entire phenomena we otherwise would only be able to study at a small scale.
On the other hand, it would likely be advisable for scientists to remember that a complete understanding of a complex phenomenon does not need to stand in the way of practical science. For example, in 2022, a bombshell study argued that our understanding of the functioning of the pharmacological drugs commonly prescribed to treat depression and related disorders - i.e., serotonin reuptake inhibitors (SSRIs) - is outdated, despite the evidence of their efficacy being largely intact. If the perfect understanding of specific mechanisms is the sole goal of science, one could imagine a scenario in which many scientists might be disappointed with our current understanding of SSRIs and thereby find their usage objectionable. However, such an impractical view would favor a kind of ‘pure’ science over utility, failing to take into account the fact that SSRIs are safe, with minimal and manageable side effects, and have a profound positive impact on patients. A drive to better understand how SSRIs work may indeed increase the efficacy of the drugs and reduce the variability of their effects, but to view this understanding as the end goal in-and-of-itself would be tantamount to sacrificing human health for the sake of scientific hubris.
Thus, embracing complex models to most accurately understand nature, while recognizing our epistemological limitations and understanding that science for its own sake is not necessarily the end goal, could provide a multi-pronged framework for advancing science. Of course, identifying the blurry line between knowing when phenomena must be fully understood (e.g., in making accurate medical diagnoses) and when a full understanding may not be necessary (as in the case of SSRIs) is a challenging philosophical endeavor in itself. Given the increasing need and public desire for accurate scientific communication, it may do scientists well to emphasize this distinction, so as to secure funding when needed to build complex models, while not letting incomplete understanding be an impediment to societal benefit. As for increasingly complex phenomena, it is perhaps fitting that as our curiosity exceeds our ability to conjure models on our own, nature itself provides us the inspiration and certain frameworks to understand its best kept secrets.
Edited by Nick Bulthuis
References
Gaynes, Robert. "The discovery of penicillin—new insights after more than 75 years of clinical use." Emerging Infectious Diseases 23.5 (2017): 849.
Hasson, Uri, Samuel A. Nastase, and Ariel Goldstein. "Direct fit to nature: an evolutionary perspective on biological and artificial neural networks." Neuron 105.3 (2020): 416-434.
Shorter, Edward. "The history of DSM." Making the DSM-5: Concepts and controversies. New York, NY: Springer New York (2013): 3-19.
Lane, Christopher. “Pathological: The True Story of Six Misdiagnoses”. psychologytoday.com/us/blog/side-effects/202206/pathological-the-true-story-six-misdiagnoses (2022).
Craske, Michelle G. "The R‐DoC initiative: Science and practice." Depression and Anxiety 29.4 (2012): 253-256.
Burgess, Neil. "The 2014 Nobel Prize in Physiology or Medicine: a spatial model for cognitive neuroscience." Neuron 84.6 (2014): 1120-1125.
Schafer, Matthew, and Daniela Schiller. "Navigating social space." Neuron 100.2 (2018): 476-489.
Aronov, Dmitriy, Rhino Nevers, and David W. Tank. "Mapping of a non-spatial dimension by the hippocampal–entorhinal circuit." Nature 543.7647 (2017): 719-722.
Dreyfus, G. "Neural networks: an overview." Neural Networks: Methodology and Applications (2005): 1-83.
Vedral, Vlatko, and Martin B. Plenio. "Basics of quantum computation." Progress in quantum electronics 22.1 (1998): 1-39.
Moncrieff, Joanna, et al. "The serotonin theory of depression: a systematic umbrella review of the evidence." Molecular Psychiatry 28.8 (2023): 3243-3256.
Pies, R., and Dawson, G. “Serotonin or Not, Antidepressants Work”. psychiatrictimes.com/view/serotonin-or-not-antidepressants-work (2022).