Around the 1950s, we started our interesting journey understanding our physiology — junk food, asbestos, lead, and of course, smoking. For the latter, how carcinogenesis develops in the lungs was for the common man more than a fuzzy concept — it was invisible, contested, and deliberately obscured by an industry with everything to lose. Fast forward to today, as Tim Wu writes, "an underacknowledged truism that, just as you are what you eat, how and what you think depends on what information you are exposed to." And yet neurology — the science of essentially the brain trying to understand the brain — is still in its infancy. So much so, it is important to appreciate that the DSM-5 only recognized non-substance-based behavioral addiction as a clinical category in 2013 to explain the phenomenon of gambling. Not only do we not have answers in the form of measurements, but we hardly know what questions to ask. What are the effects of algorithmic information exposure on our psychology? Today, this mechanism is very real yet just as fuzzy as those smoke-filled lungs, indeed.

Ask yourself the following ridiculous questions we can't even begin to answer: at what follower count, on which platform, will the identity of a self-absorbed influencer completely consolidate? We don't know. What news flow, divided by which particular outlet, in what daily doses, will cause someone to believe that 5G caused a respiratory pandemic orchestrated by "the elites"? We don't know. And what dynamic of likes, comments, and messages is classified as an altruistic-reciprocity networking strategy to build "a brand and followership," versus two human beings trying to build a genuine friendship? Today we don't know. However, one day we will be much closer to answering these and similar questions. We should all strive to be proud empiricists, but we should also recognize that just because you can't measure something yet does not mean it is not, in fact, true.

Unfortunately, some researchers have dismissed the correlation between social media use and declining mental health as negligible, as low as one percent of variance. As Haidt puts it, these people confuse the map for the territory. The fact that the dataset is not reality has been missed by too many on this subject, as Haidt and Rausch make painfully clear. The myriad of disastrous effects on children as a global phenomenon is real, and the rate of degradation can hardly be explained by any other variable. The only variable worth mentioning here: the one that consumes five hours a day on average, grabbing it every six minutes of waking life, tapping it roughly three thousand times a day, hearing a ringing Pavlovian bell with every notification, a continuously unbounded preening of one's appearance for likes, predominantly passive consuming short-form entertainment, and a constant state of downward-social-comparison. All of this, as the new, twenty-first-century predominantly discretionary activity of modern life, is surely three distinct things: unparalleled in its exposure; unprecedented with it being less than 10,000 days old; and lastly, obviously incredibly unsatisfying and regretfully dumb. Other things are going on in the world, of course, but these variables pale in comparison to what is happening to the lives of the first generation raised on screens.

This "correlation is not causation" tactic is the same corporate playbook we have heard before. The canary in the smoke mine was obvious when nonsmoking housewives of chain-smoking husbands developed cancer at a significantly higher rate. So too is it obvious now that nothing else is most probably the cause. Meta's own internal research — 31 studies the company intended to keep hidden, leaked through whistleblowers and litigation (after they had already cut off most outside researchers)—includes experiments that concluded social media causes harm to mental health. Just as with Big Tobacco, the same company that publicly insists "correlation is not causation" privately confirmed causation years ago.

Unlike those who cite the one percent correlation confusion, as Haidt stresses, the burden of proof here is not that of a criminal trial — "beyond a reasonable doubt" — but that of a civil trial: the preponderance of the evidence. We do not need definitive causality to be prudent and wise here. We need probable cause. And the preponderance is overwhelming that this is truly yet another dumb experiment we should not run. The reliable disastrous causality between time-on-device maximizing algorithms, neuronal rewiring, and its effects on our identity, our relationships to others, and ultimately our well-being, will be extremely obvious in retrospect. I am with Haidt on Pascal's Wager here: if the alarm ringers are wrong, the costs are minimal and reversible. If instead the denialists are wrong, we will have stood by doing nothing while we ruined the lives of an entire generation. On the off chance that a few trillion-dollar multinationals are going to miss out on some profit in Q3? Am I really going to take that chance, stand by, and do nothing? No. I, for one, will certainly not.