What's Well & Good in Media

Transcript 29 lines

Cold Open Stats Overview Paper Walkthrough free_promo

Cold Open

Jenny Have you ever opened an app just to check one thing and somehow walked away feeling worse?

Davis Constantly, but I don't think the app is always the villain; the same feed can teach you about a health scare, make you feel less alone, or point you toward help.

Jenny Right, and that's the split I'm stuck on: when it's information, it can steady you, but when it turns into a comparison machine, suddenly everyone else has the body, the job, the vacation, the calm kitchen.

Davis So the question isn't whether media is good or bad; it's what the design invites you to do, and whether researchers measure learning, belonging, anxiety, or that little social gut-punch.

Jenny Exactly, and one new review puts a number on that gut-punch: looking up at other people's lives online tracks with worse adjustment at about r equals point three three zero, which means a moderate link in human-behavior research ... welcome to What's Well & Good in Media on paperboy.fm.

Stats Overview

Davis This week, we started with about four hundred papers, and sixty-two made the cut. That pool had 218 unique authors across 12 countries, with Indonesia leading at eight papers and the U.S. next at three.

Jenny The qualified count fell from 81 to 62, so down 19 papers, or about 23 percent. I’d be careful calling that a trend yet, because one week can shift hard if the search catches fewer broad media-and-wellbeing papers.

Davis The bigger swing is upstream. Query hits dropped from 959 to 414, down 545 hits, or almost 57 percent, so the funnel got smaller before review even started; the question is whether this was a quieter publishing week or a narrower topic mix.

Jenny The topic mix is pretty clear. Social media led with 20 papers, mental health had eight, and digital literacy and TikTok had four each, which fits the through-line: media can spiral people, teach them, or help them cope depending on design and context.

Davis Methods were grounded more than flashy. Qualitative studies led with 17, surveys had 14, and meta-analyses and literature reviews had three each, so this week’s evidence is heavy on lived experience and self-report, which is useful but not the same as causal proof.

Jenny Author-wise, this was an early-career-heavy week. Of 218 authors, 74 were first-time authors, meaning first-ever paper in the metadata, 94 were emerging, and 50 were experienced; that’s about three-quarters of the author pool not in the established bucket.

Paper Walkthrough

Paper 1 “Looking up” linked to feeling down: a meta-analysis of online upward social comparison and psychological maladjustment

Jenny Alright, let's get into the papers with “Looking up” linked to feeling down, by Yuqing Lei, Shirui Hu, Yidan Sun, and Lijun Zheng in Frontiers in Psychology. They ask a clean first question for this whole episode: when social media makes people feel worse, is the problem the feed itself, or the moment you measure your life against someone who looks better off?

Jenny Their answer is pretty consistent. Across ninety-four effect sizes from fifty-four independent samples, with thirty-six thousand five hundred eighty-three participants, online upward social comparison was tied to poorer psychological functioning, with an average correlation of zero point three three. Correlation means the two things move together; it doesn't prove one caused the other.

Davis How did they separate ordinary social media use from the specific act of comparing yourself upward, because scrolling past a vacation photo isn't the same as thinking, my life is smaller than theirs?

Jenny They focused on studies that measured upward comparison directly, not just hours online, and then used a meta-analysis, which is a study of studies, with a model that handles multiple results coming from the same paper. The strongest link wasn't broad depression; it was social-evaluative negative emotions at zero point four three eight, meaning shame, envy, or feeling judged next to other people. The big caution is that this is mostly association, and their exploratory checks didn't show stronger overall effects in longitudinal studies than in one-time snapshots.

Davis So the practical takeaway isn't only “use your phone less.” It's “change the comparison habit,” like muting aspirational accounts, hiding status cues, or teaching people to notice the moment a post turns into a ranking. And for our Comparison and Compulsion thread, that's a useful starting point: the harm seems tied less to looking, and more to looking up.

Paper 2 Mental Health Risk Detection From Social Media Text Data: A Scoping Review of the Machine Learning Research Landscape

Davis That point about not just counting screen time sets up this next paper nicely, because here the question is also measurement. Yiqing He, Yinning He, and Darong Liu have a PsyCh Journal review called Mental Health Risk Detection From Social Media Text Data, and it's about whether posts can be turned into early warning signals.

Davis Plainly, the field is moving fast, but it's messy. They reviewed one hundred thirty-six peer-reviewed studies from January twenty twenty-one to January twenty twenty-six, and most tried to detect depression, anxiety, or suicide and self-harm risk from social media text using machine learning, meaning software trained to spot patterns in words.

Jenny So are these models detecting mental illness, or mostly detecting online signals that researchers hope stand in for risk?

Davis Mostly the second, and the authors are pretty clear about that. This was a scoping review, which means they mapped the research landscape rather than pooling one final effect size, and they followed PRISMA scoping review guidelines while searching PubMed, Web of Science, and IEEE Xplore. The big limitation is labels: many studies used proxy indicators from user content, like posts or community membership, instead of survey-linked labels or clinically anchored labels, meaning a diagnosis or assessment tied to a real person.

Jenny That makes the practical takeaway narrower, but still useful. For the Health Advice Online thread, I wouldn't hear this as “an app can diagnose you from a sad post.” I'd hear it as “public-health teams might monitor population-level risk signals,” but only if they say exactly what the label means and don't pretend one hundred thirty-six uneven studies equal a clinic.

Paper 3 Narcissism and Problematic Internet Use: Differential Links to Internet Versus Social Media Addiction—A Multidimensional Meta-Analysis

Jenny That label problem carries right into this next one, because Narcissism and Problematic Internet Use is basically asking what happens when the label is “narcissism,” but the trait itself is doing several different things.

Jenny Zhou and colleagues meta-analyzed one hundred forty-three effects from seventy-four studies, with sixty-eight thousand seven hundred five participants, and the simple headline is that higher global narcissism is linked with more problematic internet use, with a correlation of point two six, meaning a modest but real positive relationship.

Jenny The sharper bit is that social media addiction had the stronger link, again about point two six, while general internet addiction was smaller, about point one six, so the “look at me, judge me, compare me” parts of the feed seem more relevant than just being online.

Davis So what changes when they stop treating narcissism as one big trait, and split it into the versions people actually mean in life, like grandiose, vulnerable, or antagonistic?

Jenny They used random-effects models, which means they assumed the studies were estimating related but not identical effects, and then looked at dimensions: grandiose narcissism, the confident status-seeking kind; vulnerable narcissism, the insecure and easily threatened kind; and antagonism, the hostile or exploitative interpersonal piece.

Jenny That’s where the measurement story gets interesting, because grandiose and vulnerable narcissism looked fairly similar for general problematic use and social media use, around point two to point two four, but in the smaller internet-addiction subgroup the estimate rose as vulnerable narcissism mattered more, and the authors warn there was substantial heterogeneity, so this is an overall positive pattern, not a law of nature.

Davis That puts this neatly in the Comparison and Compulsion thread: if a clinician or researcher only asks “are you narcissistic,” they may miss whether the risky loop is status chasing, feeling wounded, or picking fights online, and those are three very different intervention targets.