The Same Comparison Can Hurt Feelings and Change Behavior
Upward social comparison is the process of judging oneself against a person or group perceived as better off on some dimension. This chapter asks why that process can be associated with social-evaluative distress in one setting yet produce small changes in behaviour in another. The practical goal is not to decide whether comparison is simply good or bad, but to learn which comparison is taking place, what outcome is being measured, and what kind of evidence supports the claim.
The distinction matters because the two conclusions often appear to collide. A 2026 meta-analysis of online upward comparison combined 94 effect sizes from 54 independent samples, involving 36,583 people, and found an average association with poorer psychological functioning. A separate meta-analysis covered 79 randomized trials of comparison-based behaviour-change interventions involving more than 1.35 million participants and found small effects in the intended behavioural direction. One finding concerns how people experience a naturally occurring self-evaluative process; the other concerns what happens when comparison is deliberately structured as part of an intervention. They are not measurements of the same event.
A process, not a platform
It is tempting to say that social media makes people compare themselves with others. That statement is too broad to explain much. Time spent on a platform is an exposure measure, not a description of what a person is thinking or doing. The relevant process begins when someone selects a social standard, consciously or otherwise, and evaluates their own position against it on a particular dimension. If the standard seems superior—more attractive, successful, healthy, wealthy, productive, or socially accepted—the comparison is upward.
The comparison has three linked parts. First comes the standard: whom or what is being used as a reference point? Second comes the discrepancy: how large, important, and meaningful does the gap between the self and that standard seem? Third comes the reaction: does the person feel threatened, envious, encouraged, resigned, or motivated to adjust their behaviour? The same visible difference can produce different reactions because people bring different motives, cognitive resources, self-efficacy beliefs, and assumptions about whether the relevant dimension can change.
That last point is easy to miss. A superior standard can be information about a possible next step when the dimension feels malleable and the target seems reachable. It can instead become a status verdict when the standard appears impossibly distant, when the dimension feels fixed, or when the comparison says more about acceptance and worth than about a specific action. “Someone is ahead of me” does not by itself tell us whether the comparison will guide behaviour or damage self-evaluation. We need to know what “ahead” means and what the person believes they can do about it.
Online settings can make this process more frequent and more vivid. Feeds often present polished slices of achievement, appearance, and lifestyle rather than the fuller context of a person’s life. Images, rankings, likes, and follower counts can turn an ambiguous social difference into a visible and repeated signal. The Frontiers review presents these features as plausible reasons that online comparison may be especially salient, but its meta-analysis did not isolate the causal contribution of any one platform feature. The careful claim is that digital environments can provide conditions for repeated upward comparison—not that every platform use, or every user, has the same psychological effect.
A useful way to hold the distinction is to imagine two people seeing the same comparison. One sees a peer’s record of running and, because the goal is clear and attainable, treats it as a behavioural reference. Another sees a carefully edited display of career, appearance, and social life and reads it as evidence of personal inadequacy. The external stimulus overlaps; the dimension, perceived gap, motive, and available action do not. A theory of comparison must therefore explain reactions, not merely count exposures.
The comparison target is also relative, not globally superior. A person can be above someone else on income, below them on physical performance, and uncertain about both on social belonging. A group average may be an upward standard for one participant and a downward standard for another. That is why the label “upward” describes the relation on a chosen dimension, not a permanent ranking of people. It also explains why a message that seems encouraging to one audience may feel accusatory or irrelevant to another: the message changes the reference point, but not necessarily the person’s sense that the gap is meaningful or changeable.
Context can alter the meaning of the same gap before any emotion appears. During uncertainty or threat, people may look to others for information about what is normal, safe, or possible. In a behavioural intervention, that informational motive can be made explicit by showing a participant their own performance alongside a social standard. In an ordinary feed, the same act of looking may be less deliberate and may combine with appearance cues, status signals, or fear of judgment. The distinction is not between a “good” kind of person who uses comparison productively and a “bad” kind who does not. It is between different tasks, signals, and perceived options presented by the comparison.
This is why broad measures such as time online are poor substitutes for process measures. Two people can spend the same amount of time viewing the same platform while comparing themselves on different dimensions, or while not making a self-relevant comparison at all. Conversely, a brief encounter can matter if it targets a personally important gap. The research question becomes sharper when it names the standard, the dimension, the reaction, and the time horizon rather than treating exposure as the mechanism itself.
This is the central map for the subject. Upward comparison is shared, but the setting and the question differ. The online mental-health synthesis asks whether naturally occurring comparison is associated with how people are functioning. The intervention synthesis asks whether deliberately supplied comparison information changes a target behaviour relative to a control. Keeping those routes separate prevents a behavioural effect from being used to dismiss emotional costs, and prevents an observational association from being treated as proof that a carefully designed intervention cannot work.
Why the same comparison can point in two directions
The phrase “looking up and feeling down” captures one important possibility: a comparison can signal a threat to social standing, acceptance, competence, or appearance. The Frontiers review groups body dissatisfaction, loneliness, envy, and fear of negative evaluation under social-evaluative negative emotions. These are not interchangeable with depression or general anxiety. They are more specifically tied to how a person thinks they are being positioned in relation to others.
The review then places anxiety, depression, wellbeing, and self-esteem in a broader set of psychological indicators. Its proposed proximal–distal framework suggests that an immediate social-evaluative reaction may be closer to the comparison episode, while repeated self-devaluation could be related to more generalized or cumulative difficulties. This is a useful way to organize the evidence, but it is not a demonstrated chain of events. The meta-analysis compared average associations across outcome categories; it did not prove that envy comes first, depression follows, or that every episode travels through the same sequence.
The intervention literature changes the arrangement of the problem. In a social-comparison behaviour-change technique, or SC-BCT, an intervention draws attention to other people’s performance on a specified dimension so that participants can compare it with their own. A useful comparison generally contains information about the person’s own performance and a social standard. That is different from merely saying that a behaviour is popular, approved, or disapproved. The intervention may also include goals, reminders, knowledge, or repeated feedback that lets people compare with their own past performance.
This creates a more bounded question. Instead of asking, “What does looking at successful people do to a person’s wellbeing?” an intervention study may ask, “Does showing people their energy use alongside a social standard change energy use compared with assessment alone or another feedback method?” The target is narrower, the behaviour is observable, and the comparison can be tested against a control. A person may still react emotionally, but the primary outcome is whether the intended behaviour changes after the intervention.
That difference explains how the two bodies of evidence can coexist. Unstructured upward comparison often concerns personally important status dimensions whose gaps may feel difficult to change. A structured intervention can place comparison on a malleable dimension, define the desired direction, and pair the standard with actionable feedback. These are not guaranteed opposites: a behaviour intervention can still create pressure, and an everyday comparison can sometimes inspire useful action. But they create different conditions for the same basic process.
The question worth carrying forward
The first meta-analysis found a significant average association between online upward comparison and psychological maladjustment: the pooled correlation was 0.330, with a 95% confidence interval from 0.289 to 0.370. The number summarizes a distribution of effects across studies, not a personal score or a prediction that a particular comparison will cause harm. The second synthesis found small randomized effects for comparison-based interventions, including when compared with another active intervention. Those effects describe average behavioural change under tested conditions, not proof that comparison is emotionally safe or independently responsible for every improvement.
The right response is therefore not to choose one headline and reject the other. Ask four questions whenever a claim about comparison appears. What is the standard, and on what dimension is it superior? Is the gap being interpreted as a threat to the self or as information about an attainable action? What outcome is being measured—an immediate feeling, a broad indicator of psychological functioning, or an observable behaviour? Finally, is the claim based on an association between naturally occurring experiences or on a randomized intervention with a defined comparator?
Those questions keep upward social comparison at the centre while making room for its different consequences. They also set a standard for the chapters that follow: first trace how a perceived gap can become a feeling or an action; then examine the mental-health estimates and their limits; after that, test what deliberate comparison interventions actually achieve and which design choices may matter. A better judgment about looking up begins by refusing to confuse the person being compared with, the behaviour being targeted, and the evidence being used.