Introduction

Inflammation is a part of the immune system that evolved to combat infections and heal from injuries. When the body is infected or injured, the immune system triggers inflammatory responses to prevent further spread of the infection (e.g., swelling at the infection site). In addition to these events at the infection site, inflammatory signals also travel to the brain, triggering a cascade of psychological processes in response. Notably, inflammation is now considered to play an important role in how people think, feel, and behave1,2; with effects that include altering mood3, person perception4, feelings of social connection5, social judgment6, and sensitivity to reward7.

One key research area has focused on how inflammation affects social behavior1,2,8. When the immune system is triggered, it releases pro-inflammatory signaling molecules that can result in “sickness behaviors” such as fatigue, anhedonia, and weakness1,8. From an evolutionary perspective, these behavioral tendencies not only help the body recuperate, but also protect the individual from further harms and infecting others. Consistent with this idea, initial research on animals showed that inflammation promotes social avoidance1,8,9. Similarly, studies that experimentally induced inflammation in humans showed that inflammation led to greater feelings of social disconnection5, as well as greater neural activity in response to socially threatening images4 and receiving negative social feedback10. These studies contributed to the initial belief that inflammation promotes social avoidance, as part of a hallmark sickness behavior1,8.

However, recent work has begun to suggest that the effects of inflammation on social behavior may be more nuanced than uniform social avoidance2,10,11,12,13. Specifically, inflammation at times can increase social approach behaviors and neural sensitivity to positive social stimuli10,11. For instance, compared with those who received a placebo, people who were injected with a compound that triggers an inflammatory response reported a stronger desire to be around a close other and showed stronger neural response to pictures of close others in a motivation-related brain region11. In another study using similar methods, inflammation was associated with a greater willingness to receive social support from a caring individual (e.g., nurse) relative to non-caring individuals14. Moreover, flu vaccination, which acutely raises inflammation levels, led people to approach a support figure faster on a laboratory approach/avoidance task12. These findings are consistent with the idea that approaching others who can provide social support and care may be adaptive15,16, especially when in a potentially vulnerable state (i.e., inflammation). They also indicate that inflammation can, in fact, increase social approach behaviors under certain contexts—for example, when the context presents an opportunity to interact with close others or receive social support. Thus, one key question surrounding the social effects of inflammation is understanding the various conditions under which inflammation facilitates or undermines affiliative behavior.

Building on prior work on social effects of inflammation1,2,13 and inspired by recent findings highlighting that different contexts (e.g., whom people are interacting with, what their needs are) can modulate how inflammation affects social behavior12,16, the goal of the present research was to examine how inflammation is associated with two common—yet varying in contexts—daily social behaviors, namely, interacting with others via face-to-face (FtF) and social media use (SMU). Specifically, we test the novel idea that inflammation would be associated with more SMU over FtF. Extending this, we also test whether the relation between inflammation and social behavior would vary across individuals with different personality traits.

Social affordances of social media

Although social media platforms have diverse features and are constantly evolving, they are largely perceived as “social” platforms17,18,19. For instance, common lay definitions of social media often emphasize how social media “provide new opportunities for finding, observing, and interacting efficiently with others across time and space20.” Indeed, key features shared across social media platforms, such as removal of geographical/temporal barriers and ease of disseminating self-relevant information, help facilitate self-disclosure and social interactions, which are key to maintaining and strengthening interpersonal relationships21,22,23. Moreover, the mediated (and sometimes asynchronous) nature of communication on social media enables people to strategically present themselves or be more deliberate in communicating their needs24. As a result, people can receive social support and fulfill their social needs through social media25,26. Unsurprisingly, social media platforms are often perceived and treated by users as social, relational, and communication tools19,20,26. This seems to be the case despite the fact that individuals spend much of their time on social media without directly communicating with others27,28.

Taking this notion further, some scholars contend that fulfilling affiliative needs is a robust, central motive propelling individuals to use social media17,19,29. For instance, one study showed that social interaction was one of the primary motives for using Instagram30. Similarly, another study discovered that one of the main reasons for using Instagram was related to social purposes (e.g., relationship maintenance), in particular surveillance and gathering knowledge about others31. Other researchers have identified seeking new relationships and cultivating existing relationships as two primary motives for Facebook use32,33. Collectively, these studies suggest that SMU is an appealing option for fulfilling affiliative needs (e.g., relationship maintenance, exchanging social support).

Inflammation and SMU

While SMU and FtF can both fulfill affiliative needs, we propose that individuals experiencing higher levels of inflammation may be more likely to rely on SMU over FtF. First, inflammation increases fatigue, as demonstrated by experimental studies in which inflammation is raised through either vaccination34 or administration of a bacteria-mimicking compound35. Thus, individuals experiencing higher levels of inflammation may be more motivated to fulfill their affiliative needs in energy-efficient ways. Prior work has shown that compared with other types of social interactions (e.g., FtF, voice calls), social interactions on social media require less energy36. Thus, SMU may be an appealing choice for individuals with heightened inflammation to meet their affiliative needs. This notion is consistent with the perspective from the Communicate Bond Belong theory37, which posits that conserving one’s energy is a major factor in people’s decisions about how to interact with others (i.e., interaction mode).

Second, individuals experiencing higher levels of inflammation (even when symptoms are mild and not explicitly recognizable) are likely to be motivated to protect themselves from potential physical threats (e.g., pathogen) and psychological threats (e.g., social exclusion). Indeed, inflammation has traditionally been thought to trigger sickness behaviors such as fatigue or weakness, in part to protect the body against potential infection risks1,8. Similarly, inflammation can activate the behavioral immune system38, a motivational system that increases vigilance toward potential pathogen cues and promotes avoidance-oriented behaviors39. When the behavioral immune system is activated, people become more vigilant against cues of possible contamination and more motivated to engage in behaviors that minimize exposure to others40. For example, people higher in disease concern were more likely to classify ambiguous, unfamiliar targets as threatening6,38,41 and showed higher endorsement of engaging in social distancing behaviors42. Thus, individuals experiencing higher inflammation may find the mediated nature of interacting with others on social media more appealing because it allows them to fulfill their affiliative needs while minimizing the likelihood of encountering pathogen cues. Furthermore, recent research has begun to show that inflammation may also motivate people to be protective against psychological threats such as negative feedback or social rejection10,43,44. If so, individuals experiencing higher levels of inflammation should prefer to fulfill their affiliative needs on social media not only because the social media environment is perceived to have fewer explicit rejection cues compared with FtF45,46 but it allows them to curate messages and have more control over with whom and how to communicate17,47,48. Thus, inflammation should lead to a stronger preference for SMU to FtF.

In an initial support to this idea, one study showed that C-reactive protein (CRP), a biomarker of systemic inflammation in the blood, predicted more SMU and using social media to interact with others (e.g., sending direct messages)49. However, because this prior work did not examine how inflammation is associated with other modes of social interaction (e.g., FtF), it remains unclear whether social media serves as a unique context that enables individuals with higher levels of inflammation to fulfill affiliative needs. Critically, we sought to refine and extend prior work by testing whether inflammation would predict more SMU over FtF. To provide additional supporting evidence, we also examine whether inflammation would be associated with using social media for social interaction (SMUSI) over FtF.

Social behaviors of individuals high in introversion or neuroticism under inflammation

A growing body of evidence suggests that the motivation for using social media may differ based on the individual’s personality traits and goals31,50. Consistent with this idea, the link between inflammation and SMU over FtF may be stronger for individuals higher in introversion or neuroticism.

According to the Five Factor Model51, introversion is a trait characterized by preference for low stimulation environments and focus on one’s internal thoughts and feelings. Compared with extroverts who are sociable and outgoing, introverts tend to be quiet and reserved52. Importantly, a key distinction between introverts and extroverts is how they derive energy from social interactions53: In general, whereas extroverts are energized by most social interactions, introverts often find them depleting. For example, introverts report having to exert more energy during FtF and needing time alone to recuperate after54,55. They prefer interacting with a smaller (vs. larger) group of people and maintain smaller social networks55,56,57, presumably because this helps them conserve energy. In contrast, extroverts not only engage in more frequent FtF58 but find them to be more rewarding59. Because they are more socially skilled and comfortable interacting with others60,61, extroverts find FtF energizing rather than energy depleting62,63.

The above findings suggest that introverts and extroverts may differ in how they typically fulfill their affiliative needs. However, we further propose that their difference will be more pronounced under higher levels of inflammation. Specifically, introverts who usually find FtF energy depleting may find SMU and SMUSI as an appealing option to fulfill their affiliative needs. Critically, this preference should be greater with higher levels of inflammation, which is associated with more fatigue64. In contrast, because extroverts do not find interacting with others—whether on social media or FtF—to be energy depleting, they may not show a strong preference between SMU/SMUSI and FtF even when experiencing higher levels of inflammation.

Neuroticism is a personality trait characterized by emotional instability51. People high in trait neuroticism often have high anxiety and difficulties in FtF65,66. They report more negative FtF with others67,68 and experience lower relationship satisfaction69. According to the social compensation hypothesis70,71,72, individuals who lack high social resources or skills offline seek to compensate by investing their resources into online relationships (e.g., social media). Importantly, this tendency may be more pronounced among people high on internalizing symptoms such as neuroticism or rejection sensitivity73,74, in part because online environments contain fewer explicit rejection cues45,46.

Given that inflammation motivates people to avoid situations that could be psychologically threatening10,43, these findings suggest that individuals with varying levels of neuroticism may also differ in how they fulfill affiliative needs under higher levels of inflammation. Specifically, individuals high in neuroticism may find SMU and SMUSI more appealing than FtF because the social media environment may have fewer rejection cues and allow them to engage in selective self-presentation75. Importantly, this preference should be stronger with higher levels of inflammation, which can make people more vigilant for physical or psychological threats43,44. In contrast, for individuals low in neuroticism, such affordances may not be as salient of a factor in how they fulfill their affiliative needs. Thus, they may not show a strong preference between SMU/SMUSI and FtF even when experiencing higher levels of inflammation.

Present research

Despite recent advances in understanding the social effects of inflammation, important gaps remain. First, although a growing amount of evidence suggests that the impact of inflammation on social behavior is context-dependent (e.g., interacting with a close other vs. a stranger), little is known about additional conditions under which inflammation promotes social approach or avoidance behavior. Building on prior work49, we propose interaction modality (e.g., face-to-face, social media) as one understudied context that can potentially modulate the link between inflammation and social behavior. Specifically, we argue that social media is an especially appealing environment to fulfill affiliative needs for people experiencing higher levels of inflammation.

Second, most studies to date have examined proxy measures of social experience in the laboratory rather than actual daily social behaviors13. For example, prior studies have measured performance on computer-based social behavior tasks12, self-reported desire to approach a close other11, feelings of social disconnection5, or neural response to social stimuli10. Related, these studies have used experimental manipulation of inflammation (e.g., flu vaccination, endotoxin administration), which provides strong causal evidence. However, less is known about how naturally occurring, systemic, inflammation relates to daily social behaviors. This gap is notable because inflammation can be affected by common daily experiences such as diet, sleep, exercise, or stress76,77, and emerging perspectives suggest that elevations in inflammation can impact daily life more broadly than previously recognized78. Thus, to address this gap, the present research examined how inflammation relates to two common daily social behaviors: social media use (SMU) and face-to-face interaction (FtF). Examining SMU and FtF together allows us to control, to some extent, for general social motivation within individuals and provides more information about one’s daily social behaviors (see an approach championing assessing multiple modes of interaction in a study79,80). Moreover, the focus on SMU and FtF helps isolate daily social behaviors from broader social desire or subjective feelings of social connection, thereby extending prior work on the social effects of inflammation.

The present research had two primary goals. First, we tested whether people experiencing higher levels of inflammation would engage in more social media (SMU) over face-to-face interaction (FtF). Based on prior work indicating how inflammation can enhance social affiliative motivation under some contexts10,11 and our theoretical framework linking inflammation to psychological states such as fatigue and vigilance against physical and psychological threats, we hypothesized that inflammation would be associated with more objective SMU over FtF (H1a). Following prior work81,82, we defined SMU as time spent on social media, objectively recorded via participants’ smartphone. However, given that people can use social media for purposes other than social interaction (e.g., reading the news), we also examined whether inflammation would be associated with more self-reported social media use for social interaction (SMUSI) over FtF specifically. We defined SMUSI as amount of social media use for social interaction (e.g., directly communicating with others through functions such as direct messaging), based on relevant prior work27,83,84. Because measuring SMUSI objectively was not available to our knowledge, we assessed it via self-report, consistent with prior work83. By the same logic as in H1a, we predicted that inflammation wouldbe associated with more SMUSI over FtF (H1b). Second, we explored whether these associations would vary across individuals with different personality traits—namely, introversion and neuroticism. Given that people high in introversion find FtF more energy-depleting and their tendency to prefer lower stimulation environments, we hypothesized that introversion would moderate the relation between inflammation and objective SMU over FtF (H2a), as well as self-reported SMUSI over FtF (H2b). Specifically, we expected the relations to be stronger among individuals higher in introversion. Finally, because people high in neuroticism tend to be more sensitive to psychological threat cues and thus may prefer environments with fewer explicit rejection cues (i.e., social media vs. FtF), we predicted that neuroticism would also moderate the links between inflammation and objective SMU over FtF (H3a) and self-reported SMUSI over FtF (H3b). Specifically, we expected the relations to be stronger among individuals higher in neuroticism.

Method

Participants and procedure

One hundred and seventy-one college students (102 females; Mage = 19.24, SDage = 2.68) from the Ohio State University participated in this study for partial course credit in their Introductory Psychology course. Participants first completed background questionnaires assessing factors such as sociodemographic information, personality, social media use, face-to-face interactions, and various social and health behaviors on Qulatrics in the laboratory. Once participants completed the questionnaires, they were taken to another room where a trained research assistant collected their blood samples. To collect blood, each participant’s finger was swabbed with alcohol and pricked with an 18-gauge lancet, following prior work85. We then collected the blood drops on a protein saver card. Participants were assured that they could opt out of the blood collection without losing compensation or dropping out of the study. Fourteen participants (8.19%) opted out of blood sample collection, and 3 participants (1.75%) provided samples that were judged to have an insufficient amount of blood for assay. Thus, we did not assay the blood from the latter three participants. Because inflammation (from the blood) data from these participants were not available, we excluded these 17 participants from all analyses. Thus, the final sample size for this study was 154 (93 females; Mage = 19.28, SDage = 2.81).

This study and its research protocol were approved by the Institutional Review Board at The Ohio State University (Protocol 2018H0452). All procedures were performed in accordance with relevant guidelines and regulations, including the Declaration of Helsinki. Informed consent was obtained from all participants prior to their participation.

Measures

Objective social media use (SMU)

Using the Screen Time application on their iPhone (the iOS operating system), participants retrieved information on how much time they spent on each of the five social media platforms (i.e., TikTok, Snapchat, Instagram, Twitter/X, and Facebook) during the week for which they visited our laboratory to provide their blood samples. Because the Screen Time app records SMU for the entire week from one Sunday to the next Sunday, we contacted participants during the week immediately following their blood collection so that they can provide their SMU for the entire week that overlapped with when the blood was collected. We assessed SMU across these five platforms because they are considered to be the most popular social media platforms among college students86. This approach is also consistent with recent recommendation to collect social media use data across multiple platforms20,87. We summed the weekly averages across the five social media platforms to create a composite weekly social media use (SMU) variable (M = 870.00 min, SD = 653.01 min). At the time of our data collection, objective SMU information was not reliably accessible for non-iPhone users, making it difficult to create consistent and reliable research materials for non-iPhone users. Thus, we decided to only collect SMU data from iPhone users. Therefore, SMU data were not available from 17 non-iPhone users.

Social media use for social interaction (SMUSI)

Participants indicated the extent to which they used social media (i.e., TikTok, Snapchat, Instagram, Twitter/X, and Facebook, and other social media combined in general) for social interactions and direct communication during the week. We defined for them social interactions as “exchanging messages, tagging, reacting and commenting on others’ posts on social media sites,” and direct communication as “sharing pictures/videos and updates, reacting and commenting on others’ page, or exchanging direct/private messages.” These items were based on relevant prior work27,83,84. Participants used a 5-point scale (1 = not at all, 5 = very much) to indicate their response to the two items (α = 0.91, M = 3.18, SD = 1.17). Just as with the SMU measure, to obtain SMUSI for the week of blood collection, we contacted participants the week immediately following their blood collection so that they can provide their SMUSI that overlapped with when they provided their blood samples. Because SMUSI measures were obtained via self-report, all participants (including non-iPhone users) were able to provide their data (n = 154).

Face-to-face social interactions (FtF)

Participants indicated the extent to which they engaged in a variety of face-to-face social interactions during the week using a 5-point scale (1 = not at all, 5 = very often/all the time). To capture a variety of in-person social interaction settings, we slightly modified a measure from prior work88 to better fit our college student sample contexts. Our modified activities were: “get together with friends to hang out in-person, go to a party or other social events (e.g., concerts, sports events) with friends in-person, watch a movie or TV show with friends in-person, have a meal with friends or family together in-person, have a deep face-to-face conversation with someone in-person, volunteer to help others in-person, and attend religious services in-person” (α = 0.80, M = 2.63, SD = 0.81). This FtF measure was negatively associated with introversion in our study, which is consistent with prior work58. Similar to SMU and SMUSI, the FtF measure was obtained during the week immediately following their blood collection.

CRP

We assayed CRP from dried blood spots based on prior work85. Specifically, we swabbed each participant’s finger with alcohol and pricked it with an 18-gauge needle (Owen Mumford Unistick 3). We collected the blood drops on a Whatman 903 Protein Saver Card and left them to dry at room temperature for 24 h. Then, we punched the samples (3 mm Biopsy punch) and stored them in microcentrifuge tubes at − 80 °C until they were assayed. To assay, we thawed a single 3 mm punch and added 200 µl of buffer (Phosphate Buffered Saline with 0.1% Tween 20) for overnight incubation at 4 °C while shaking at 60 rpm. We then diluted this eluate 1:10 and assayed CRP the following morning using Meso Scale Delivery Vplex Plus kits [K151STG]. All samples were within the linear range of the standard curve and therefore they were included in the analyses. The intraassay CV was 3.8% and the interassay CV was 11.04% (M = 0.99 mg/L, SD = 1.75 mg/L).

Introversion and neuroticism

Participants indicated the extent to which various pairs of traits apply to them from the Ten-Item Personality Inventory89 using a 5-point scale (1 = strongly disagree, 5 = strongly agree). Introversion was measured with “extraverted, enthusiastic” (reverse-coded) and “reserved, quiet” (α = 0.85, M = 2.99, SD = 1.26). Neuroticism was measured with “anxious, easily upset” and “calm, emotionally stable” (reverse-coded) (α = 0.71, M = 2.67, SD = 1.01).

Covariates

Based on prior work49, we controlled for extraneous factors that can influence social behaviors. For our sociodemographic covariates, we measured gender, age, family income, and highest level of education completed by mother and father (1 = some high school, 5 = graduate school). We also controlled for depressive symptoms using the CES-D90 (α = 0.87, M = 1.98, SD = 0.59) and body mass index (BMI; M = 23.59, SD = 4.47), which we derived from participants’ self-reported height and weight.

Analytical strategy

Because our outcome of interest was the extent to which participants engaged in SMU over FtF and in SMUSI over FtF, we created two variables to test our hypotheses: 1) SMU over FtF by first standardizing each participant’s objective SMU (in minutes) and FtF (5-point scale) scores and subtracting the FtF variable from the SMU variable so that higher scores indicate more objective SMU over FtF (M = − 0.096, SD = 1.44) and 2) SMUSI over FtF by standardizing each participant’s SMUSI and FtF scores (as they were measured on different scale labels) and subtracting the FtF variable from the SMUSI variable so that higher scores indicate more self-reported SMUSI over FtF (M = − 0.005, SD = 1.18). Based on theoretical grounds that inflammation should have linear effects on dependent measures, CRP was used as a continouous variable without excluding values over 10 mg/L, as is sometimes done when CRP is a dependent variable91. We used these variables in all analyses.

In testing our hypotheses, we followed prior work49 by sequentially controlling for an increasing number of covariates: (1) no covariates (Model 1), (2) socio-demographic factors (Model 2), (3) BMI (Model 3), and (4) depressive symptoms (Model 4). This was done to provide greater details on the nature of the effect. We controlled for sociodemographic factors because they can influence social media usage92. Although BMI was not controlled for in prior work, we included it as a covariate given its association with sedentary behaviors, which may influence SMU, SMUSI, and FtF. Finally, given the possibility that depression can influence social media use93,94 and inflammation95, we controlled for depressive symptoms. We controlled for these extraneous variables sequentially (vs. simultaneously) to provide more details about how the inclusion of each covariate influenced the association between CRP and SMU/SMUSI.

We first explored how our outcome variables (i.e., SMU over FtF, SMUSI over FtF) were associated with key variables by conducting Pearson correlation using SPSS version 30. We then tested our main hypotheses by conducting a series of multiple regression (H1a and H1b) and moderation analyses (H2a-H3b) using Model 1 of the PROCESS macro (version 4.2) for SPSS in SPSS. To assess the robustness of our main analyses, we also used Structural Equation Modeling (SEM; lavaan, version 0.6–18) to test whether CRP was more strongly associated with SMU and SMUSI than with FtF. As detailed in the Supplemental Materials, the results were consistent across the different analyses.

Results

Table 1 presents zero-order correlations among all key variables. Notably, SMU over FtF was positively correlated with introversion (r = 0.31, p < 0.001), neuroticism (r = 0.33, p < 0.001), and depressive symptoms (r = 0.46, p < 0.001), indicating that people who scored higher on introversion, neuroticism, and depressive symptoms engaged in more SMU over FtF. Similarly, SMUSI over FtF was positively associated with introversion (r = 0.21, p < 0.001), neuroticism (r = 0.15, p = 0.07), and depressive symptoms (r = 0.34, p < 0.001), indicating that people who scored higher on these variables engaged in more SMUSI over FtF. These findings are largely consistent with the idea that individuals high in introversion and individuals high in neuroticism may use more social media70 and how depression can lead to more social media use93.

Table 1 Zero-order correlations for key variables.
Full size table

H1a

CRP will be positively associated with objective SMU over FtF.

We conducted a series of multiple regression analyses with CRP as a predictor of SMU over FtF (i.e., SMU minus FtF). CRP was positively associated with SMU over FtF in Model 1 (β = 0.19, p = 0.033), Model 2 (β = 0.20, p = 0.028), and Model 3 (β = 0.23, p = 0.019). Although directionally consistent, CRP was not a statistically significant predictor in Model 4 (β = 0.17, p = 0.070). See Table 2 for details.

Table 2 Coefficients from linear regression models predicting SMU over FtF.
Full size table

H1b

CRP will be positively associated with self-reported SMUSI over FtF.

We conducted regression analyses with CRP as a predictor of SMUSI over FtF (i.e., SMUSI minus FtF). Although CRP was positively associated with SMU over FtF in Model 1 (β = 0.18, p = 0.036) and Model 2 (β = 0.19, p = 0.035), it was not a statistically significant predictor in Model 3 (β = 0.17, p = 0.069), and Model 4 (β = 0.11, p = 0.20). See Table 3 for details.

Table 3 Coefficients from linear regression models predicting SMUSI over FtF.
Full size table

H2a

Introversion will moderate the link between CRP and objective SMU over FtF.

Moderation analyses yielded a CRP x introversion interaction predicting objective SMU over FtF in three models, Model 1 (F(1, 124) = 5.33, p = 0.053), Model 2 (F(1, 114) = 4.51, p = 0.036), Model 3 (F(1, 113) = 5.00, p = 0.027), Model 4 (F(1, 112) = 6.05, p = 0.016). To better understand the broader pattern of our results, we further probed the interactions. Simple slopes analyses indicated that for individuals high in introversion (1 SD above the mean), CRP was positively associated with SMU over FtF, Model 1 (b = 0.36, p = 0.003, 95% CI = [0.12, 0.59]), Model 2 (b = 0.38, p = 0.002, 95% CI = [0.14, 0.61]), Model 3 (b = 0.42, p = 0.001, 95% CI = [0.17, 0.66]), Model 4 (b = 0.37, p = 0.002, 95% CI = [0.14, 0.61]); for individuals low in introversion (1 SD below the mean), CRP did not predict SMU over FtF (ps = 0.87, 0.98, 0.87, 0.73). See Fig. 1 depicting Model 4 results.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

CRP and SMU over FtF for individuals low in introversion (− 1 SD from mean) and individuals high in introversion (+ 1 SD from mean) with all covariates (Model 4). Note *p = .002. 95% CIs for individuals low in introversion: [− .25, .17] and individuals high in introversion: [.14, .61].

H2b

Introversion will moderate the link between CRP and self-reported SMUSI over FtF.

Moderation analyses yielded a CRP x introversion interaction predicting SMUSI over FtF: Model 1 (F(1, 138) = 2.90, p = 0.091), Model 2 (F(1, 126) = 3.75, p = 0.055), Model 3 (F(1, 124) = 3.80, p = 0.054), Model 4 (F(1, 123) = 4.80, p = 0.030). Although the interaction was non-significant depending on different analytical approaches, based on theoretical grounds we further explored the interactions: For individuals high in introversion (1 SD above the mean), CRP was positively associated with SMUSI over FtF, Model 1 (b = 0.26, p = 0.008, 95% CI = [0.07, 0.45]), Model 2 (b = 0.29, p = 0.004, 95% CI = [0.09, 0.48]), Model 3 (b = 0.26, p = 0.008, 95% CI = [0.07, 0.45]), Model 4 (b = 0.23, p = 0.01, 95% CI = [0.05, 0.41]); for individuals low in introversion (1 SD below the mean), CRP was not associated with SMUSI over FtF (ps = 0.81, 0.94, 0.88, 0.49). See Fig. 2 depicting Model 4 results.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

CRP and SMUSI over FtF for individuals low in introversion (− 1 SD from mean) and individuals high in introversion (+ 1 SD from mean) with all covariates (Model 4). Note *p = .01. 95% CIs for individuals low in introversion: [− .23, .11] and individuals high in introversion: [.05, .41].

H3a

Neuroticism will moderate the link between CRP and objective SMU over FtF.

Moderation analyses yielded a significant CRP x neuroticism interaction predicting objective SMU over FtF in three models: Model 1 (F(1, 124) = 3.68, p = 0.057), Model 2 (F(1, 114) = 4.38, p = 0.039), Model 3 (F(1, 113) = 5.29, p = 0.023), Model 4 (F(1, 112) = 5.29, p = 0.026). Simple slopes analyses indicated that for individuals high in neuroticism (1 SD above the mean), CRP was positively associated with SMU over FtF, Model 1 (b = 0.30, p = 0.008, 95% CI = [0.08, 0.52]), Model 2 (b = 0.34, p = 0.004, 95% CI = [0.11, 0.56]), Model 3 (b = 0.39, p = 0.001, 95% CI = [0.16, 0.63]), Model 4 (b = 0.35, p = 0.003, 95% CI = [0.12, 0.57]); for individuals low in neuroticism (1 SD below the mean), CRP was not associated with SMU over FtF (ps = 0.69, 0.69, 0.78, 0.59). See Fig. 3 depicting Model 4 results.

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

CRP and SMU over FtF for individuals low in neuroticism (− 1 SD from mean) and individuals high in neuroticism (+ 1 SD from mean) with all covariates (Model 4). Note *p = .003. 95% CIs for individuals low in neuroticism: [− .29, .17] and individuals high in neuroticism: [.12, .57].

H3b

Neuroticism will moderate the link between CRP and self-reported SMUSI over FtF.

The CRP x neuroticism interaction significantly predicted SMUSI over FtF in two models: Model 1 (F(1, 138) = 3.47, p = 0.065), Model 2 (F(1, 126) = 4.12, p = 0.044), Model 3 (F(1, 124) = 3.75, p = 0.055), Model 4 (F(1, 123) = 3.97, p = 0.048). Simple slopes analyses indicated that for individuals high in neuroticism (1 SD above the mean), CRP was positively associated with SMUSI over FtF, Model 1 (b = 0.24, p = 0.009, 95% CI = [0.06, 0.41]), Model 2 (b = 0.26, p = 0.005, 95% CI = [0.08, 0.44]), Model 3 (b = 0.23, p = 0.01, 95% CI = [0.05, 0.42]), Model 4 (b = 0.20, p = 0.02, 95% CI = [0.03, 0.38]); for individuals low in neuroticism (1 SD below the mean), as predicted, CRP was not associated with SMUSI over FtF (ps = 0.70, 0.66, 0.60, 0.42). See Fig. 4 depicting Model 4 results.

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

CRP and SMUSI over FtF for individuals low in neuroticism (− 1 SD from mean) and individuals high in neuroticism (+ 1 SD from mean) with all covariates (Model 4). Note *p < .01. 95% CIs for individuals low in neuroticism: [− .11, .11] and individuals high in neuroticism: [.05, .24].

Discussion

The present research examined how inflammation relates to daily social behaviors, namely, SMU, SMUSI, and FtF. Based on emerging research on inflammation and social behavior11,13 and extant work on interaction modes and their affordances24,37, we predicted and showed that inflammation was associated with greater SMU and SMUSI over FtF. Although these effects became weaker with some covariates (e.g., depressive symptoms) included in the model, the pattern of results remained the same across all models. Importantly, we found that these patterns were stronger among individuals higher in introversion or neuroticism.

Our findings make several novel contributions. First, by identifying inflammation as a correlate of greater SMU/SMUSI over FtF, we provide insight into when and how people engage in daily social behaviors. Because interaction modality can have implications for relational and personal well-being17,96,97, understanding the conditions under which people interact with others on social media versus face-to-face is important. Second, we contribute to the growing body of research on inflammation and social behavior. While most prior studies have focused on demonstrating the effects of experimentally manipulated levels of inflammation on laboratory-based proxies of social behavior11,12, our findings show that naturally occurring inflammation is related to actual daily social behaviors. Moreover, inflammation’s link to greater SMU/SMUSI over FtF corroborates the view that the impact of inflammation on social behavior is context-dependent10,12. Finally, our CRP x personality traits moderation findings provide a biological account for why certain individuals may turn to social media over FtF. Here, we contribute to research on the social compensation hypothesis71, which to date has predominantly focused on personality traits or sociality (e.g., introversion, rejection sensitivity) to explain why people may prefer online to offline interactions. We add to this literature by showing that a biological marker of inflammation can contribute to social compensation effects.

Broader implications

Social media’s role as a social technology

Our findings suggest that social media can play an intriguing role as a social technology, especially for individuals who, for whatever reasons, cannot or prefer not to engage in FtF. Because humans have a fundamental need for social connections98, such individuals still must find ways to fulfill their affiliative needs. Indeed, research shows that socially vulnerable individuals (e.g., low self-esteem, high rejection sensitive) desire high-quality social interactions and relationships just as much as others do99. Similarly, it is not that individuals with high inflammation do not want social interactions, but rather, they seem to have a particular preference regarding whom they interact with (e.g., supportive figures11) or, as we suggest, how they interact with others (SMU/SMUSI over FtF). We propose that social media may be a particularly appealing space for these individuals as it has lower energy demands, infection risks, or social rejection cues compared with FtF. One interesting future direction is to explore whether our findings would generalize to other social technology, for example, AI companion apps such as ChatGPT, Character.AI, or Replika. Although empirical evidence is needed, these technologies also offer opportunities to fulfill affiliative needs in energy efficient ways with virtually no infection or social rejection risks. Thus, we would suspect that our findings would generalize to many other technologically mediated social contexts.

Nevertheless, whether SMU or SMUSI is an effective method for these individuals to address their affiliative needs remains to be an important empirical question71,72. On the one hand, studies show that social media can help individuals who struggle to cultivate social relationships face-to-face connect with others online72,100. On the other, such individuals also report higher rates of being cyberbullied and harassed online101. However, we suspect that higher SMU/SMUSI over FtF may be detrimental to well-being, especially when SMU/SMUSI displace FtF102. Despite the potential benefits of SMU/SMUSI, studies show that FtF, compared with other interaction modalities, is still the most effective way to achieve feelings of social connectedness and meet various relational needs17,36. Similarly, although we cannot rule out bidirectional interpretations, in our study SMU over FtF and SMUSI over FtF were both positively correlated with depressive symptoms. Thus, while more research is needed to understand when and how SMU/SMUSI can effectively fulfill affiliative needs, future work should also consider a broader array of daily social behaviors (e.g., FtF, SMU, texting, parasocial interactions) rather than examining any one behavior in isolation.

Inflammation and social behavior

We note that inflammation levels in the daily lives of young people like those in our study are appreciably lower than inflammation levels present in people with infections (e.g., flu, COVID) and chronic illnesses (e.g., autoimmune or cardiovascular disease). Thus, it is likely that our participants were not necessarily feeling sick or aware of their inflamed state. In this way, our finding that inflammation, even when mild with no drastic symptoms, is associated with SMU/SMUSI over FtF is consistent with evidence that “low-grade” inflammation (i.e. elevated inflammation in the absence of an infection) can have psychological effects similar to those from acute inflammation in response to infection, for example43,78. Because inflammation can be elevated by a variety of daily experiences such as stress, poor sleep quality or diet, or low exercise76,77, its impact on daily social behaviors is likely to occur frequently and for many individuals including those who are healthy. Thus, future research should aim to better understand how inflammation is related to a wider variety of daily social behaviors.

Another future direction is identifying the upstream biological pathways that may be influencing the inflammation seen here. One potential system to investigate in future research is the parasympathetic nervous system (PNS). Stimulation of the PNS inhibits the production of proinflammatory cytokines103,104 and a non-invasive index of greater PNS activity (High Frequency Heart Rate Variability; HRV) is associated with lower circulating levels of CRP and pro-inflammatory cytokines105. A rich body of theoretical work106 has postulated a social engagement system governed by the PNS. In support of a role for greater PNS activity facilitating social approach, greater HRV is associated with greater social support seeking as a coping response107 and greater felt affiliation with an ingroup and greater behavior supporting the ingroup108. A related measure of greater PNS activity has been associated with greater probability of being with others rather than alone on an event recall diary109. Conversely, a dispositional tendency for social avoidance (trait shyness) was strongly associated with lower levels of HRV110. Thus, PNS activity does appear to be related to social approach, and whether higher PNS activity is associated with the degree to which one engages in SMU over FtF and mediates the inflammatory effects seen here will be an important area for future research.

Limitations and future directions

The present research has some limitations. First, the correlational nature of this study does not allow us to make strong causal or directional inferences regarding our findings. For example, we cannot conclude that greater SMU over FtF does not lead to higher inflammation82. These results do provide a valuable platform justifying more challenging (and expensive) studies that would address limitations such as intensive longitudinal designs that measure inflammation, SMU, SMUSI, and FtF multiple times or experiments that manipulate inflammation levels (e.g., administering a molecule that mimics a bacterial infection). Second, although we have hypothesized that inflammation would be associated with more SMU/SMUSI over FtF through several mechanisms such as increased energy conservation motivation or vigilance against physical or psychological threats, our findings do not provide direct evidence for these mechanisms. Similarly, although we have posited that much of the appeal of SMU is its ability to fulfill affiliative needs, people may use social media for other reasons (e.g., reading the news). Thus, future research should more directly examine the motivations behind using social media (e.g., seeking social support) and different types of SMU (e.g., direct messaging, posting selfies, scrolling). Relatedly, future studies may also examine whom (e.g., strong vs. weak ties) people seek to interact with when they are under higher levels of inflammation. Finally, future research should seek to improve measurement of social behaviors. For example, it is possible that some of our participants had imposed Screen Time limit on their phone during the week of the data collection, which can directly influence SMU and SMUSI. Given the growing number of apps and features designed to help users limit their social media or smartphone use, future studies should assess in their study whether participants are using such tools to restrict their usage. Moreover, because self-report measures of SMU frequency estimates can be imprecise111, future studies should consider developing more objective measures of SMUSI (e.g., using data donation or passive sensing method to capture time spent on direct messaging) and FtF (e.g., Electorincally Activated Recorder, which unobtrusively records conversations naturalistically112). Related, our study was limited in that we collected SMU data from five social media apps, which may not capture the full range of apps participants use. This reflects a broader challenge in social media research: researchers must decide in advance how many apps to assess prior to data collection while participants are likely to be using more apps than can be reasonably included in a survey. Although our approach in collecting usage across five apps is an improvement over prior studies that focus on a single app (e.g., Facebook), future research may benefit from more intensive methodologies such as passive sensing, which can unobtrusively track smartphone use across all apps over time. Such approaches would allow researchers to retrieve comprehensive usage information after data collection and test more targeted hypotheses—for example, examining within-person associations between inflammation and the use of messaging apps such as WhatsApp or Telegram.

Conclusion

The present research examined how inflammation is associated with daily social behaviors. Extending prior work, inflammation was associated with more SMU and SMUSI over FtF. Importantly, this pattern was more pronounced among individuals with higher introversion or neuroticism. These results not only corroborate the idea that the link between inflammation and social behavior may be context-dependent but highlight the unique role of social media as a social technology for certain individuals. Given the increasing awareness of mind–body connections and the importance of daily social lives, more research investigating the relation between biological processes and social behavior is needed.