Science's iterative process is broken

Why our society can't seem to agree on what is true

In order to understand how and why science is failing, one must first understand what science is. There is perhaps no noun more commonly misunderstood in the year 2021 than ‘science’. Some think science is knowledge acquired through the analysis of data, others think science is knowledge that is agreed upon by relevant experts, and others are naïve enough to still think science is an understanding established through an iterative process of eliminating ideas that are proven false (i.e,. falsified). None of these are a complete picture, though the latter is the closest.

Science is a method for humans to utilize their God-given ability to perceive reality to inform a process of testing speculations for the underlying mechanism giving rise to those observations. The process of testing one’s speculations is known as an experiment. This is the first hurdle science usually fails to clear. Science requires a carefully controlled experiment, where artificial conditions are simulated, to isolate a speculative effect. If the experimenter does not account for all possible confounding variables, spurious conclusions about the nature of the isolated effect are inevitable. In situations where a variable cannot be controlled, statistical methods are employed to mathematically isolate the variable of interest from one or more confounding variables. Data science has progressed towards increasingly complex statistical models to control for confounds post-hoc, and in turn experiments have become increasingly lazy, opting to control for confounds statistically rather than experimentally.

There are two unintended consequences of this pivot away from rigorous experimental controls and towards sophisticated mathematical modelling: firstly, it disadvantages those who are not trained or proficient in statistics but are otherwise competent in their field; and secondly, it applies an unnecessary layer of statistical discourse over all fields of science – one that is often left unresolved. In other words, irreconcilable arguments about the proper statistical approaches and interpretations of data prohibits the progression of scientific theory. One may argue a stasis in scientific theory is preferable to establishing spurious theories established on misinterpreted data, but I will argue that the scientific process is designed to do the exact opposite. It is our human desire to never be wrong that is fomenting a stasis in scientific progress.

The elucidation of data through experimentation is followed in science by an explanation of the mechanism that gave rise to the data pattern. This process is the hypothesis formulation stage. Many scientists mistake this as the point at which it is essential to “get it right”, but that is not true. It is perfectly acceptable for a spurious conclusion to subsequently result in a hypothesis that turns out not to be true. In fact, the scientific process depends upon there being hypotheses that are not true. Science is an iterative process – it is not over once a single experiment has been run and the dataset analyzed. In fact, it is never over. Every hypothesis and theory (a hypothesis that has been reaffirmed independently multiple times) is up for grabs. One could design a novel experiment to test whether or not the Earth is round, if one wanted to. This example is important because it elucidates an asymmetry that plagues science – the asymmetry of payoff, and this is science’s second hurdle. A quick aside here to explain this asymmetry, and why it’s corroding our scientific process, and then back to the iterative process of science.

Like it or not, science costs money, and money requires funding, and funding requires justifying the expenditure. Most scientific endeavours do not receive funding, and those that do must justify themselves as contributing significantly to the body of knowledge. It is for this reason that scientific experiments seeking to assess, corroborate, or refute prior experimentation, is fiscally disincentivized. The iterative process of science is disincentivized, and thus we depend on scientists to get it right on the first try, lest false hypotheses persist. An experiment designed to test whether or not gravity exists would never get funded – in part this is because this is “settled science”, but it illustrates the broader point that novel scientific research is more likely to be funded than iterative research. The scientific process is fragile when its findings are not checked and re-checked, by independent researchers and clever new experiments. The scientific process is anti-fragile when its findings are repeatedly checked, through corroborative experimentation, not through pedantic statistical debates. By and large, science has become fragile, save for a few fields where reassessments of cutting edge findings are at the centre of the funding process.

The iterative scientific process is imperative to weed out the mistakes, spurious conclusions, misinterpretations of data, and statistical errors that are otherwise unavoidable. It becomes exponentially more unlikely that the same spurious conclusion would be reached repeatedly in subsequent independent experiments. In contrast, it becomes increasingly likely that a hypothesis is true when it is corroborated by subsequent independent experiments. The current landscape of science is largely populated with independent novel experiments, each occupying a position in a network of related research, because these studies are attractive to funders. With novel research, scientists must interpret how their findings fit into the surrounding landscape of literature, and sufficiently illustrate these connections for other researchers to build off. There is no formal structure for doing this, but it takes place in the Discussion section of a published article, and it involves more story-telling than science. In contrast, the anti-fragile landscape of science is governed by directly linked experiments building off of predecessors in a vertically structured way, such that it is clear which hypotheses are being assessed in which experiments. There is a question with this approach, which is, where do new vertical pathways of knowledge originate if all researchers are busy confirming or rejecting existing research? I suspect, however, that the human inclination to discover new knowledge will not waver anytime soon, and thus this hazard is mostly not worth worrying about.

(There is a question here of how to go about incentivizing the checking of existing hypotheses. It is not clear that there is an intuitive sense of the importance of doing so, and it may instead rely upon scientists who recognize this issue establishing a central pot of money for funding research that seeks exclusively to check existing hypotheses.)

The vertical structure of science is necessary to the iterative process. In this context, science can thus be said to evolve. Evolution is simply the process by which replicators constantly change towards more stable forms. In science, the hypotheses are the replicators. The reason science must be allowed to evolve naturally is to achieve more stable forms – in this case, hypotheses that are more likely to be true. In this framework, one does not trip over the oft-made misconception that science achieves revelations of truth. The truth about this universe is elusive, a realization that philosophers have repeatedly discovered but remains largely unknown to the general public. For this reason, a mainstream understanding of science is generally misinformed at the fundamental level. Science is the evolution of knowledge to more likely hypotheses, not some mystical process by which “Truth” is discovered plainly. It is also for this reason that no hypothesis is immune to doubt. It is that the appropriate amount of doubt is proportional to the stability of the hypothesis. More stable hypotheses (i.e., theories) are less likely to be false, and thus are less worthwhile to attempt to falsify, and less sensical to be doubtful of. The inverse — unstable hypotheses — are more likely to be falsified, and these low hanging fruits of science are where most iterative research is oriented.

None of this is to say that one couldn’t make a living out of doubting highly stable hypotheses. For example, there is an intellectual market for Creationists who cast doubt on Darwin’s theory of natural selection. None of these arguments emanate from alternative hypotheses that are more stable than natural selection, and they are readily falsified or do not meet the logical rigours required to be taken seriously, but that doesn’t stop millions of people from finding them compelling. Science would not say these people should be silenced, or even ridiculed, it merely asks for better attempts. And therein lies the rub.

Science relies on being falsified. Without falsifications, science stops evolving, and without evolution, science remains in a stasis of unstable hypotheses. And this is science’s third and critical hurdle. The Achilles’ heel of science is that it is done by scientists, and scientists are almost all human, and humans really fucking hate being wrong. It’s true – many of the revelations in the field of Psychology for the past two centuries has centred around the ways in which the human mind constructs entire illusions about reality just to avoid being proven wrong. Thus, science is inescapably bound to the limitations of human psychology, and the two are in direct conflict. Science relies on being wrong, scientists rely on being right.

I do not see a good solution to this conundrum. You could simply ask scientists to embrace being wrong, but the young ones will inevitably be surpassed in their career aspirations by those willing to feign never being wrong, for institutions are incentivized to hire better scientists, and better scientists are less likely to be wrong. It is therefore untenable to expect scientists to embrace the fallibility necessary to push science forward at the cost to their individual careers. Perhaps instead someone will craft up a clever parallel institution to Academia science where being wrong is rewarded, allowing the self-aggrandized scientists to continue playing their game in the universities without sacrificing the entire scientific project. Until then, science is stalled out, and our entire society is suffering for it. It is for this reason that debates about the nature of reality seem like they can persist for eternity without any reconciliation. Disagreements about everything from vaccines to UFOs, diets to 9/11, biological sex to economics, are a reflection of the stasis in sensemaking and knowledge acquisition, and at their core share one thing in common: no one wants to admit they’re wrong. In reality, all of these disputes could be settled using the scientific method; a well-structured public forum allowing people to put forward falsifiable hypotheses, attempt to falsify others, and defend against challenges would lead to increasingly stable forms of knowledge and understandings of our reality. Instead, it seems for now we’re content with petty Twitter battles and tribal warfare bereft of any concern for epistemology or empiricism. In the immortal words of Friedrich Nietzsche,

Science is dead. Science remains dead. And we have killed him. How shall we comfort ourselves, the murderers of all murderers? What was holiest and mightiest of all that the world has yet owned has bled to death under our knives: who will wipe this blood off us? What water is there for us to clean ourselves? What festivals of atonement, what sacred games shall we have to invent? Is not the greatness of this deed too great for us? Must we ourselves not become gods simply to appear worthy of it?

(disclaimer: I substituted "God" with "Science" for those of you unfamiliar with the original quote)