Many disputes persist because the underlying questions can’t be cleanly tested.

Evidence often informs debates without settling them, especially in complex systems.

Values shape which outcomes people prioritize, even when the facts are agreed upon.

Differences in values create the biases that lead experts to stress different findings.

Every few months, a new study on the four-day workweek makes headlines. Depending on who’s describing it, the shorter week is either a revolutionary breakthrough or a productivity death spiral. Companies report higher morale and fewer sick days; critics counter that the research is patchy and highly industry-specific, the studies often lack anything resembling proper controls, and the hype ends up wildly disproportionate to the evidence. The same basic data can be framed as proof of a workplace revolution or a sign that we’re wishing our way into bad policy.

This piece isn’t about settling that debate. The four-day work week just happens to illustrate a broader problem that shows up whenever we try to turn complicated, value-heavy questions into crisp scientific ones. We tend to assume the right dataset should deliver a clean, universally correct answer. Then we look around, see experts pulling in different directions, and conclude something must be wrong with the science.

That’s usually the wrong diagnosis. The real issue is that people approach certain topics expecting a level of clarity that the underlying evidence can’t actually provide.

When the Question Can’t Deliver the Certainty We Expect

Brown (2025) recently mapped out scientific claims along two dimensions—how testable they are and how strongly values shape the conclusions drawn from them (see Figure 1 for an adaptation). Those two factors help explain why some claims feel solid while others generate years of disagreement.

Some problems come with clear, noncontroversial results. If a material fails, the bridge collapses. If a medication behaves differently at different dosages, a clinical trial will show it. In settings like that, claims converge because the world corrects errors quickly and with little ambiguity. But issues involving workplace policy, nutrition, mental health interventions, and many public-health recommendations don’t operate under those conditions. Parts of them can be tested, but the broader judgments they result in rarely fit inside tidy experimental designs. The systems are large, the evidence varies, and people care about very different outcomes.

In that kind of territory, evidence rarely produces a single, decisive answer. It informs the discussion, but it doesn’t end it. Disagreement persists because the questions themselves won’t yield clean, final conclusions. These are also the places where gathering more data doesn’t necessarily reduce the uncertainty.

The Testability Problem

Claims differ in how easily real-world evidence can confirm or contradict them. For the four-day workweek, the testability problem shows up immediately:

Organizations differ immensely in industry, size, staffing, and constraints.

The “four-day week” isn’t standardized—some compress hours; others cut them.

Pilot programs are often voluntary, attracting motivated participants.

Many effects are long-term and hard to measure cleanly.

If we can’t run repeated, real-world trials under stable conditions, the evidence will always be provisional. That’s not the fault of the researchers; it’s a feature of the domain. And this leads to conclusions that will always be somewhat contestable.

The Values-Pull Problem

Then there’s the part we’re less eager to admit: What we hope is true can shape what we see in the evidence.

If you believe that modern work culture is corrosive and that people desperately need more time off, you may interpret the same findings differently than someone who prioritizes profitability and staffing stability. Both sides can pull selectively from the data without technically misrepresenting anything. That’s how value-laden domains work: Facts matter, but the framing determines which facts feel decisive.

This doesn’t mean people are irrational or malicious. Humans—even experts—simply don’t approach data as blank slates. Our aims, experiences, and priorities leak into how we define the problem and how we interpret a result. When a topic is emotionally or politically charged, the gravitational pull toward a preferred conclusion gets even stronger.

Where People Go Wrong

Put limited testability and strong values pull together, and you get a kind of epistemic fog. Claims become contestable because they blend facts with judgments. Yet, we keep demanding a level of certainty science can’t deliver in those areas.

That mismatch fuels three common mistakes:

Expecting scientific consensus where consensus isn’t possible: Domains like metallurgy or microbiology converge because the evidence is testable and repeatable. Domains like workplace policy or public health interventions often don’t converge because the evidence is harder to interpret. When we expect the former from the latter, we interpret normal disagreement as incompetence or bias.

Expecting p-values to resolve value tradeoffs: A statistical model can estimate changes in productivity or sick days. It can’t tell you how to weigh those findings against fairness concerns, hiring patterns, or the intrinsic value of more time off. Models presented with mathematical precision can hide the value assumptions built into their conclusions.

Confusing debate over values with debate over facts. People talk past each other because they’re actually answering different questions. One person is asking, “Does this improve well-being?” Another is asking, “Does this threaten business viability?” Both can cite evidence. Both can be right. And neither is answering the same question.

What Disagreement Reveals About the Question

When we see legitimate[1] experts dueling on topics like work schedules, nutrition, stress, or public policy, we tend to conclude that the science should have produced a clearer answer. In reality, the disagreement often tells us something useful: We’re in a domain where evidence is limited, the testability is poor, and values do a lot of the heavy lifting. In those domains, disagreement is a feature, not a failure. Science can inform the conversation, but it can’t resolve it. So what should we do with claims that sit in this middle zone between easy empirical truth and pure opinion?

A few simple heuristics can help:

Ask which part of the claim is testable.

Ask whose values define success.

Treat confident pronouncements skeptically.

Expect persistent disagreement.

None of this means we should abandon scientific evidence or treat all conclusions as equally valid. It means we need to be more precise about which kind of claim we’re dealing with—and more honest about the role our values play in interpreting results.

The four-day workweek debate isn’t uniquely confusing. It’s just a clear example of a broader pattern: Many of the issues we care most about live in regions where science can offer some insight but not decisive conclusions. The trouble starts when we pretend otherwise. If we recognized the difference between claims that can be cleanly tested and claims that inevitably mix evidence with values, we’d spend less time assuming ignorance or bad faith and more time having the conversation we’re actually trying to have.

[1] I use the term “legitimate” here to distinguish between domain-qualified experts and the broader class of public-facing figures often treated as experts despite limited or uneven expertise. I discuss some of the incentives that contribute to this dynamic in a recent Substack essay called The Counterfeit Era of Expertise.

There was a problem adding your email address. Please try again.

By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy

© Psychology Today