A Science paper says remote work wrecks mental health. The thread says it never measured a single remote worker

The finding everyone will quote

The Science paper reports that after the pandemic, workers in remotable jobs spent more time alone and avoided social activity, worsening mental health — most acutely for those living alone, who "spent entire days without human contact" with measurable increases in distress, mental-healthcare use and antidepressant use. The quantified hook: remotable workers spent an extra 1.1 waking hours alone and saw a 72.2% relative increase in days with zero human contact.

It is exactly the kind of result that gets weaponized into policy. Which is why the thread mattered.

The methodology problem: there are no remote workers in this study

"They don't compare remote vs non-remote workers. They compare workers in job families that could be remote vs workers in job families that are unlikely to be remote. Their control group is nonsense … The real effect is living alone or not. Also, it conflates mental health utilization with mental health status." — a researcher picking apart the study's design

Another reader spelled out the confound that makes the timing damning:

"they did not actually check if anyone in these jobs was working remotely. They compared job categories. … This is an analysis of the depression of software developers vs the depression of mechanics. … There were mass tech layoffs in 2022-2024, and those people who got laid off were more depressed." — a commenter who dissected the paper's confounds

The same commenter offered a deflating alternative for that headline "1.1 hours alone": it's the commute people no longer make.

The authors saw one objection coming — and missed the bigger one

To their credit, the authors pre-empted the AI angle, and a reader quoted the rebuttal directly:

"We found that the mental health effects load on remotability rather than AI expos[ure]" — a commenter quoting the paper; an ex-TechCrunch editor now at The Aerospace Corporation

But they couldn't rule out the simplest confound of all — self-selection:

"Who's more likely to choose a job that can be done from home? People who already have reasons they'd rather not go out and spend their entire day around other people. How do you control for all those reasons?" — a commenter raising self-selection

Why this one is worth being skeptical about

Isolation is real, and living alone while working from home is a genuine risk worth studying. But a paper that never observes a remote worker, uses mechanics as the control group for developers, runs across the biggest tech-layoff years on record, and conflates seeking mental-health care with needing it — that paper cannot carry the return-to-office argument that will inevitably be hung on it. The thread didn't just nitpick; it identified the load-bearing flaw.