When Evidence No Longer Settles the Argument
The Key Today is Not Just Better Evidence. It’s Better Conditions for Evidence to be Trusted
Einstein was right: politics is more difficult than physics
Evidence is everywhere, yet agreement seems to be nowhere. We live in an age of unprecedented data abundance—of real-time metrics, sophisticated models, and artificial intelligence capable of detecting patterns invisible to humans. Yet even the most basic facts are now fiercely contested. Elections, pandemics, and climate change no longer provoke disagreement primarily over values or trade-offs, but over what is real and true. Appeals to “follow the science” not only fail to settle arguments; they often intensify them. In a media environment shaped by algorithmic amplification, tribal identity, and AI-generated persuasion, evidence is steadily losing its authority as a shared reference point. Where does this leave us?
We are left with an uncomfortable question: does evidence still lead us toward truth in any meaningful sense, or has it become just another instrument in political conflict? The short answer, I will argue, is that evidence still matters profoundly—but it no longer functions as a neutral or decisive arbiter of truth in complex, identity-laden political contexts. Understanding why requires a clearer view of what we mean by “truth” in the first place.
The belief that empirical evidence leads invariably to truth amounts to a secular creed—an article of faith born of the Enlightenment and reinforced by the extraordinary achievements of science. While appealing to empirical evidence is usually far superior to the alternatives in decision-making, the relationship between evidence and truth is far more subtle than we often admit. It helps to begin by going back to basics and distinguishing three different conceptions of truth, which are frequently conflated.
First, there is faith-based truth. This includes religious belief, moral conviction, and the kind of foundational certainty captured by cogito ergo sum. Such truths are not typically falsifiable. They are prior commitments—starting points rather than conclusions derived from evidence.
Second, there is formal truth: the inescapable truth that emerges from mathematics and logic. Once axioms and rules of inference are specified, conclusions follow mechanically. These truths are exact and internally consistent, but only within the closed world of their assumptions.
Third, and most contentious, is empirical truth: the kind of truth associated with observation, experience, and the scientific method. It’s the claim that vaccines reduce mortality, or that increasing atmospheric CO₂ raises global temperatures. This is the form of truth most directly relevant to public policy—and also the most fragile.
Empirical truth differs fundamentally from the first two in that it’s always provisional, forever subject to revision as new evidence accumulates. The history of science is replete with overturned truths: Newton yielding to Einstein, classical determinism to quantum uncertainty. Yet this tentativeness does not make empirical truth arbitrary or merely subjective, as some post-modernist critics suggest.
Empirical truth has force because the world exhibits regularities which, once recognized, enable predictions to be made. Why such regularities exist in the first place is a deep mystery—one that cannot be explained in terms of more primal concepts and must simply be accepted as a feature of nature. At this fundamental level, science does rest on some foundational assumptions: that the world is not wholly chaotic, that observation is informative, and that patterns discovered yesterday are not irrelevant tomorrow. This is a very different kind of “faith” from belief grounded in doctrine, identity, or authority.
Once regularities are observed, evidence allows us to navigate the world far more effectively than intuition or ideology alone. In some fortunate cases, empirical evidence can be translated into quantitative models that generate testable predictions—physics being the canonical example. Engineering works not because our theories are true in some ultimate sense, but because they are true enough for practical purposes. Bridges stand. Planes fly. Microchips function.
Public policy, however, operates on far less stable terrain. The distinction matters because policy depends almost entirely on empirical truth, while often treating it—implicitly or rhetorically—as if it possessed the certainty of faith or formal logic. But the fact is that when evidence is invoked in policy debates, especially those involving economics, social outcomes, or political behaviour, it is constrained by three structural limitations.
The first is that the empirical base is always messy. Data are incomplete and often poorly measured. Controlled experiments are rare, ethically constrained, or infeasible. Social science has nothing like the laboratory conditions of physics, and few double-blind randomized trials that cleanly isolate causation.
Second, even when patterns are observed, it is extraordinarily difficult to identify the dynamic “rules” governing complex social systems. There is nothing analogous to Newton’s three laws of motion. Feedback loops, nonlinearity, and adaptive behaviour dominate. Prediction degrades rapidly. This is why economic forecasting so often fails, why many social policies fall short of expectations, and why political outcomes—despite increasingly sophisticated polling—continue to surprise experts. Once human volition enters the system, uncertainty explodes—a reality that led Einstein to remark that politics is more difficult than physics.
Third, and most troubling, these limitations create ample space for prior beliefs, tribal identity, and self-interest to intrude and overwhelm evidence. When empirical conclusions are clouded by uncertainty or ambiguity—as they almost always are—beliefs and biases rush in to fill the gap. Once formed, these beliefs are remarkably resistant to correction. The conviction that vaccines cause autism, or that Barack Obama was not born in the United States—despite overwhelming empirical evidence to the contrary—illustrates how fragile evidence can be when it collides with identity and ideology.
The reality, then, is that evidence rarely delivers an unambiguous policy “truth”. More often, it serves humbler but still essential functions: ruling out clearly bad options, narrowing the plausible range of outcomes, and improving the odds that we are asking the right questions. Eventually, decisions must be made before uncertainty is resolved. At that point, value judgments inevitably take over, layered on top of incomplete and contested evidence. It has always been thus.
There remains a recurring hope that one day we might develop a genuine “physics” of society. Indeed, that ambition may seem to be more achievable as big data, social media, and artificial intelligence have dramatically expanded the range of observable behaviour. In principle, this should allow us to uncover new regularities in how individuals and groups respond to incentives, information, and persuasion. In that imagined world, factual disputes would largely recede and politics would consist primarily of moral choice—how much equality we want, how much efficiency we are willing to sacrifice.
Unfortunately, on present evidence, the hope that improved data and modelling alone will resolve our deepest political conflicts appears illusory. The lesson isn’t that the science or the evidence is too weak. It’s that human conviction—shaped by millions of years of evolution—is influenced far more by identity, interest, and group affiliation than by detached empirical reasoning. Only when evidence becomes so overwhelming that it directly and powerfully alters lived experience does it begin to override those forces—and by then, it may be too late for effective collective action.
The digital ecosystem has, so far, worsened rather than improved this dynamic. Online platforms reward engagement, not accuracy. Media echo chambers reinforce prior beliefs. Falsehoods propagate faster than corrections. Seeing is no longer believing in a world of deep-faked video and synthetic audio. It’s one of the great ironies of the information age that as the deluge of data proliferates, shared agreement on empirical reality becomes harder to achieve.
A sustainable society depends on some minimal consensus about facts, even if empirical evidence can never confirm absolute truth. When that consensus collapses, disagreement over values mutates into conflict over reality itself. Evidence loses its coordinating role, polarization deepens, and collective decision-making begins to grind down.
It should by now be clear that the challenge ahead is not simply to generate better evidence, important as that remains. The deeper challenge is to sustain and renew the social and institutional conditions under which evidence can once again command broad public trust: institutions that mediate between data and decision, that reward empirical discipline, and that make disagreement over values possible without disagreement over reality. Without such conditions, even the most rigorous facts—and truth itself—risk becoming just another opinion, and democracy loses its footing.
 | A guest post by
|