Loneliness and depression reshape how people choose their entertainment media

People with higher levels of depression, anxiety, and loneliness make systematically different entertainment choices than their healthier peers, according to a new study that models media selection as a moment-by-moment decision process rather than a single static preference. Using a computational framework known as the drift diffusion model, researchers linked participants’ mental health scores to the way evidence about affective media features accumulates in the mind during a choice. The work, published in the journal Communication and Change, replicates an earlier finding that people generally prefer negatively valenced entertainment, and adds a sharper twist: lonelier individuals lean into that negative preference even more strongly, while those with moderate to severe depression or anxiety symptoms shift toward low-arousal content.

The study was led by Valerie Klein, Xuanjun Gong, Allison Eden, and Richard Huskey, and it set out to address a gap in a large body of research on mood and media. Most prior work on how transient affective states drive media selection, and how selected media in turn shape subsequent moods, has been conducted on healthy adults. That leaves open the question of how chronic affective states—such as the sustained low mood of depression or the persistent unease of anxiety—shape what people choose to watch. The team treated mental health status as exactly such a chronic affective state and asked whether it moderates the well-documented pull of negative and arousing content.

To capture the decision process itself, the researchers used the drift diffusion model, a computational tool borrowed from cognitive psychology that decomposes a binary choice into its underlying components. In this framework, drift rate reflects how quickly evidence favoring one option accumulates over another, effectively indexing preference. The decision boundary captures how cautious a decision maker is, with smaller boundaries indicating faster, less deliberate choices. Decision bias reflects a pre-existing leaning toward one option before evidence accumulates, and non-decision time captures perceptual and motor processing outside the evidence accumulation process. By estimating these parameters with Bayesian methods, the team could see not just what people chose, but how the machinery of choice differed across mental health profiles.

A total of 765 participants consented to the online study, of whom 663 completed it; after data cleaning procedures, the final sample was 313. The researchers noted that this level of attrition and filtering is not uncommon for online behavioral experiments using the drift diffusion model. Importantly, filtered and retained participants did not statistically differ on their PHQ-9 depression scores, GAD-7 anxiety scores, or UCLA-LS-20 loneliness scores, suggesting the filtering did not systematically bias the sample on the key variables of interest. The sample was predominantly female and largely composed of undergraduate students, and participants reported slightly higher depression scores than a comparison dataset from the Healthy Minds Network, a difference the authors attribute in part to the gender composition of the sample.

The task itself was a two-choice media selection paradigm. On each trial, participants saw two movie summaries that differed in experimentally manipulated affective properties: one more positive or negative in valence, one more or less arousing. A manipulation check confirmed the stimuli worked as intended. Self-reported valence for positively valenced summaries averaged 3.239 compared with 2.142 for negatively valenced ones, and high-arousal summaries elicited higher arousal ratings than low-arousal ones, with both differences highly significant. The authors acknowledge that this simplified two-option design prioritizes internal validity over the richness of everyday media environments, which involve many competing options and viewing contexts.

The core results concern how mental health scores interacted with valence and arousal in the drift rate. Replicating the earlier study by Gong and colleagues, participants overall showed preferential evidence accumulation for negatively valenced and high-arousal media—a direct counter to the valence hypothesis of Mood Management Theory, which holds that people select positive content to intervene in noxious moods. But the interactions told a more nuanced story. As depression scores on the PHQ-9 increased, preference shifted toward low-arousal media. The same pattern held for anxiety: participants with few or no symptoms favored high-arousal content, but as scores climbed toward moderate to severe levels, the preference crossed zero and reversed toward low-arousal media. For both measures, at roughly one standard deviation above the sample mean—PHQ-9 of 14.3 and GAD-7 of 13.1—preferences converged at chance before flipping.

Loneliness behaved differently. Higher scores on the UCLA-LS-20 loneliness scale were associated with a stronger preference for negatively valenced media, building on an overall negative preference that all participants shared. Notably, the mental health effects emerged specifically in drift rate, the parameter indexing preferential evidence accumulation, with no credible corresponding effects on decision boundary, decision bias, or non-decision time. Because the models were estimated with Bayesian inference, these null results are interpretable: the posterior distributions for the other parameters were tightly centered around zero. This parameter specificity suggests that mental health shapes media selection by altering how affective information is sequentially sampled during a decision, rather than by changing general response strategies or perceptual-motor constraints—a level of mechanistic detail the authors argue would be obscured by static choice models.

One exception involved anxiety and decision bias. All participants showed a pre-decision bias toward positively valenced media, consistent with Mood Management Theory, but this bias weakened as anxiety scores increased, meaning more anxious individuals were less oriented toward positive content even before actively evaluating their options. The authors suggest this may reflect habitual, rather than value-based, selection, citing work showing that decision bias tracks how frequently people make similar choices in life. They also caution that their study investigates the decision function while only inferring the value function—the stage where media options are assigned subjective value—and that future work with reinforcement learning models is needed to separate these stages.

Interpreting the findings, the authors argue they broadly support a stress and coping perspective on media selection. People with higher anxiety may be choosing low-arousal media in a way that could soothe their symptoms, perhaps seeking comfort or using media to cope by comparison or distraction. The similar pattern for depression is unsurprising given the strong correlation between PHQ-9 and GAD-7 scores in this sample, which the authors note is consistent with well-documented comorbidity between anxiety and depression. Yet the authors flag a possible downside: seeking low-arousal content when anxious or depressed may be adaptive emotional coping that helps users feel better in the moment while doing little to address underlying issues, potentially limiting media’s capacity to change mood states. For loneliness, the interpretation is even more open. Rather than acting as a social surrogate, lonely individuals may be monitoring their social environment through negative content—an instrumental use suggested by Uses and Gratifications theory—or displaying a maladaptive pattern; the authors state that more research is needed to determine which.

The authors are careful about causal claims. Although the affective characteristics of the media were experimentally manipulated, the mental health indicators were only measured, limiting causal inference. They note that existing evidence suggests effects of any single media exposure on subsequent mental health are, at best, small and sensitive to platform, content, and context, while depression, anxiety, and loneliness tend to be relatively stable within individuals over time—a pattern consistent with mental health shaping media selection, though reciprocal and cumulative effects remain plausible. Looking forward, the team suggests the findings could inform interventions that target the decision process itself, for example by structuring recommendation systems or choice environments to disrupt maladaptive feedback loops between mood and media use, rather than simply prescribing particular content. They also call for replication in more demographically representative samples across multiple institutions, and for extending the approach to naturalistic settings where real-world media selection unfolds.

Subject of Research: How depression, anxiety, and loneliness relate to media selection dynamics modeled with the drift diffusion framework

Article Title: Using the drift diffusion model to examine relationships between mental health and media selection dynamics

Article References: Klein, V., Gong, X., Eden, A., & Huskey, R. (2026). Using the drift diffusion model to examine relationships between mental health and media selection dynamics. Communication and Change, 2(1), Article 20. https://doi.org/10.1007/s44382-026-00040-5

Image Credits: AI Generated

DOI: 10.1007/s44382-026-00040-5

Keywords: media selection, drift diffusion model, depression, anxiety, loneliness, mood management theory, decision making, arousal, valence, mental health, computational modeling, entertainment media