Research on the harms of social media — mental health, disinformation, radicalization, echo chambers… — has gone astray by asking the wrong question: "has social media worsened the problem?" It's a tempting but meaningless question that leads to dangerously wrong policy advice.

It is meaningless because social media is not adopted in a vacuum. It’s like asking whether car adoption made roads more dangerous. Maybe, maybe not; but cars reshaped everything about transportation (and neighborhoods and societies). There’s no way to set up a proper comparison.

It doesn’t matter if social media has made these problems worse or not. Banning social media is not a realistic option; no one is advocating for that. So the answer to this question is not useful or actionable for policy making.

Instead, what we urgently need to know is: *can social media algorithms be changed* to minimize harms to mental health and to democracy? (With cars, the relevant question was whether mandating seat belts would save lives!) We can’t turn back the clock. But we can fix the design.

Researchers who frame the issue around the wrong counterfactual are unwitting pawns of tech companies, who benefit from endless debates about an irrelevant question, so that they can say the science is not settled (or, often, that it shows that social media is not a culprit).

The current misguided research also distracts from the fact that to study the burning question—can design changes improve our individual and collective lives?—tech companies must be forced to open up their systems! Kudos to the researchers/journalists/advocates focusing on that.

Usually researchers will skirt the counterfactual issue by studying the effect of “exposure to social media”. This, too, is irrelevant — it frames collective harms in an individualistic lens. Individuals shouldn’t be the target of intervention. Fix the system, not the people.

(That slogan is inspired by "Fix the system, not the women" by Laura Bates of @EverydaySexism.)