A Scoping Review of Prosocial and Antisocial Media Content and Effects in Narrative Entertainment Research Across Five Decades

Conceptual Ambiguity Surrounding Prosocial and Antisocial Media Content and Effects

Definitional clarity is critical for the advancement of science (Chaffee, 1991). Good definitions should specify a term’s essential features and identify the term’s boundary conditions (Chaffee, 1991; Miller & Nicholson, 1976). As Miller and Nicholson (1976) noted, when shared understanding of a concept’s essence and boundary conditions increases, this heightened precision not only improves the concept’s utility, but can “provide the impetus for a major conceptual breakthrough” (p. 124).

Yet the essence and boundary conditions of prosocial and antisocial media content and effects are not well-demarcated in early or more recent literature studying entertainment media and its effects. Tamborini and colleagues (2024) suggested that most extant definitions of prosocial and antisocial have listed examples of content and effects rather than “delineating their essence” (p. 1) by describing the concepts’ indicators and identifying their boundary conditions (but see Coyne & Smith, 2015). For example, the earliest study of prosocial media content in the population of studies identified by the present work defines prosocial by listing, intentionally leaving the list open-ended when it describes prosocial as “programs that present cooperation, altruism, self-control, and achievement orientations, to mention only a few possibilities” (Friedrich et al., 1973, p. 1). As another example, scholars using a widely cited definition describe prosocial media content as that which depicts a “voluntary behavior intended to benefit another,” (Eisenberg & Miller, 1987; Eisenberg & Fabes, 1998), with some scholars including self-interested motivations within the boundary conditions of this definition (Padilla-Walker et al., 2013, p. 395) and others treating self-interested motivations as mutually exclusive with “prosocial” (Clark & Giacomantonio, 2013).

Popular conceptions of antisocial media content suffer from similar ambiguous boundary conditions and defining by listing, with some defining antisocial acts as those that broadly go “against the aims or norms of society, therefore all antisocial acts are classified as unjustified,” (Potter & Ware, 1987, p. 667), and more recent work conceptualizing antisocial media simply according to features of “violence, bullying, and stereotypes” (Maitland et al., 2018, p. 310). As noted by Tamborini et al. (2024), defining by listing alone can lead scholars to study the same concept but refer to fundamentally different things (for instance, one scholar might study violence and one might study stereotype depictions both under the umbrella of “antisocial content”). List-based definitions likely contribute more to knowledge fragmentation than they do to incremental advances toward a shared understanding of a phenomenon (see Miller & Nicholson, 1976). In line with Hempel (1952) and Chaffee (1991), Tamborini et al. (2024) stated that relying solely on listing to define pro/antisocial content/effects is detrimental for two reasons.

First, listing without a fundamental idea of a concept’s essential qualities fails to denote that concept’s boundary conditions. Without conceptual boundaries, distinguishing a prosocial act from an antisocial act – or distinguishing either from a nonsocial act – is difficult at best and impossible at worst. For instance, Roloff and Greenberg’s (1979) definition of antisocial content includes deceit, which can resolve conflict but also threaten one’s relationships with others. Compare this to Comstock & Strzyzewski’s (1990) definition of prosocial content as depicting conflict resolution. The conditions under which conflict resolution is considered prosocial versus antisocial are unclear without the conceptual delineation that boundary conditions provide. Indeed, previous investigations found that people can simultaneously enact prosocial and antisocial behaviors, leading to uncertainty about whether prosocial and antisocial behaviors are mutually exclusive (Erreygers et al., 2017; Gentile et al., 2009).

Second, readers can make idiosyncratic interpretations of a construct when a definition based on listing does not identify how the listed items are meaningfully related. For example, den Hamer et al. (2017) defined prosocial media content as that depicting “helping, sharing, [and] cooperation” (p. 290). Although these terms seem related at face value, the list is not finite nor is a meaningful connection between the terms specified, increasing the likelihood that scholars might infer different meaningful connections. For example, one scholar might think the word’s shared meaning was a function of group coordination, while another might think that shared meaning was based on giving aid. The two scholars are likely to interpret findings from research using a scale designed to measure effects of this content very differently. Further, in efforts to expand the scale, we might expect the first scholar to add items like “cohesion,” “unity,” or “connectedness,” whereas the second might add items focused on like “assisting,” “supporting,” and “comforting.” What began with intentions to measure a similar concept would have then diverged into two entirely different measures that assess responses to entirely different concepts (e.g., group solidarity versus compassion, respectively). Yet because of scholars’ intentions, these measures may be labeled as assessing “prosocial” attitudes, even if they now focus on entirely different concepts. Future scholars interested in assessing the influence of prosocial media on prosocial effects may then be surprised to observe different magnitudes of effects depending on which scale they select, given that both purport to measure the same thing (also see jingle-jangle fallacy in construct validity research; Kelley, 1927). In each case, failure to specify the meaningful connection among the words used in a definition (i.e., the concept’s essence and boundaries) impedes efforts to advance understandings of pro/antisocial media’s influence.

Perhaps the method of scientific advancement most hindered by the lack of a construct’s shared meaning is the use of meta-analysis to synthesize findings across studies in an area of research. Meta-analysis is specifically designed to integrate the findings of research on a focused topic in order to uncover patterns obscured by extraneous factors. Its ability to reveal these patterns is largely dependent on the homogeneity of the studies included (Carpenter, 2020). The failure to account for heterogeneity causes analytical problems often described using the metaphor of “comparing apples to oranges” (Carpenter, 2020). This drawback can be seen in meta-analyses examining extant research on pro/antisocial media’s influence.

Although previous meta-analyses have attempted to synthesize the findings of research on prosocial and antisocial media (e.g., Sanborn & Harris, 2019), the prospect that these concepts lack a shared meaning across research in this area raises concern about those attempts (see Pfattheicher et al., 2022). Indeed, several meta-analyses investigating prosocial and antisocial media effects report small aggregate effect sizes (Coyne et al., 2018; Mares & Woodard, 2005). Of course, it could be that the true magnitude of effect in this area of research is small and/or these effect size estimates represent great individual-level variability in effects. However, the potential that studies included in these meta-analyses lacked a shared conception of prosocial and antisocial media suggests that heterogeneity of concepts may also be attenuating the demonstrated effects.

Given the ambiguity surrounding prosocial and antisocial definitions, we argue that a scoping review is necessary to help “map” extant literature and determine how researchers in this area have defined these concepts (see also Coyne et al., 2018). We turn to the MIME (Tamborini, 2013) for a theoretical “compass” to guide this endeavor, and attempt to determine whether extant work has defined prosocial/antisocial media content and effects according to the definitions outlined by Tamborini et al. (2024). An investigation into how past scholars have defined prosocial and antisocial has two advantages. First, it would allow scholars to determine whether past meta-analyses were hindered by studies using different conceptions of prosocial and antisocial media (i.e., whether these studies compared apples and oranges). Second, it would allow scholars to correct for problems in past meta-analyses created by studies using different conceptions of prosocial and antisocial media. Once identified, a meta-analysis could control for the difference in these conceptions and eliminate their ability to obscure existing patterns across studies.

Adding Clarity to Existing Definitions of Prosocial & Antisocial Media Content & Effects

In general, the MIME (Tamborini, 2013) describes a reciprocal relationship between media and audiences, wherein exposure to media emphasizing specific moral values can lead audiences to hold those values in higher regard than others. In turn, this can shape media appraisal and subsequent consumption to drive the production of similar value-laden content.

The MIME adopts its understanding of morality from MFT (Haidt & Joseph, 2007), which defines moral values as innate, evolutionary-developed sensitivities toward right/wrong across five social domains: care (versus harm), fairness (versus unfairness), ingroup loyalty (versus betrayal), respect for authority (versus disrespect), and purity/sanctity (versus impurity). Care refers to concern and regard for others’ well-being. Fairness refers to a desire to be equitable, just, and impartial. Ingroup loyalty refers to favoritism toward members of one’s group and bias against non-members. Respect for authority refers to reverence for benevolent, legitimate authority figures, institutions, and traditions. Last, purity refers to avoidance and prevention of bodily and/or social contamination.

MFT argues that all people favor the upholding of these values and abhor their violation, implying that humans have hardwired sensitivities to these values, that, when activated, spark intuitive emotional responses. Notably, humans are also thought to exhibit more or less sensitivity to these values depending on environmental factors such as culture (e.g., Graham et al., 2009). Exemplars of these values manifest across a variety of media, such as television and social media content across the world (Gonzalez et al., 2024; Lewis & Mitchell, 2014; Prabhu et al., 2020). Previous research demonstrated that observing exemplars of these values upheld in media content produces positive affect, whereas observing these values’ violation generates negative affect (Lewis et al., 2014).

We adopted the MIME as the guiding theoretical framework for the present study because it offers a systematic approach to classifying both media content and effects based on a comprehensive scheme of moral values. Bridging insights from moral psychology (e.g., MFT) with established models of media effects (e.g., social cognitive theory, exemplification theory), the MIME is uniquely suited to our goal of examining how consistently prosocial/antisocial media content and effects are defined in extant literature. The MIME was specifically developed to explain how exposure to morally laden media exemplars affects audiences’ moral sensitivity, attitudes, and behaviors. Crucially, it builds on MFT to offer a comprehensive taxonomy of moral values that can be used to identify and compare the moral values both embedded in media content and underlying audiences’ responses to that content. Because morality is central to most scholars’ conceptions of prosocial/antisocial media content and effects (e.g., see example definitions above), efforts to assess how these terms are defined requires a framework that accounts for both morality and media effects processes. The MIME fulfills this requirement, making it particularly well suited to guiding the present study’s investigation of conceptual and operational definitions in narrative entertainment research.

Drawing on the MIME’s conceptualization of moral values, recent work attempted to develop a theoretically-driven definition for prosocial and antisocial media content and effects based on whether content/effects exemplify the upholding or violation of one or more moral values. Specifically, Tamborini et al. (2024) defined prosocial in terms of content/outcomes that exemplify the upholding of one or more moral values, and antisocial as content/outcomes exemplifying the violation of one or more moral values. Tamborini et al. (2024) note that in complex narratives where multiple moral values may conflict, it is the upholding or violation of the overridingly salient moral value (i.e., the value emphasized as most important by the narrative) that determines whether content is labeled as prosocial or antisocial. This concept of overriding salience provides a theoretical mechanism for resolving definitional conflicts in the literature, such as those we identify between Roloff and Greenberg’s (1979) and Comstock and Strzyzewski’s (1990) definitions above. In such cases, it would have been difficult to determine whether actions taken to resolve conflict between different motivations were prosocial versus antisocial without the conceptual delineation that Tamborini et al.’s definition provides.

Importantly, Tamborini et al.’s (2024) definitions delineate the essence of pro/antisocial media content/effects by indicating that prosocial/antisocial involves the upholding/violation (respectively) of a set of moral values outlined by the MIME/MFT and suggest clear boundary conditions regarding what constitutes pro/antisocial content/effects as the upholding/violation of these values, and no other values. They argued that, due to the universal nature of the MIME and MFT’s moral values, scholars may inherently construe prosocial and antisocial as associated with the motivation for upholding and violating moral values, respectively.

To test this claim, Tamborini et al. (2024) presented mass communication scholars and the general public with a list of behaviors in which a character was motivated to uphold or violate a moral value and asked them to indicate whether these behaviors were prosocial, antisocial, or neither. Their study demonstrated that scholars and members of the public implicitly define prosocial and antisocial in a manner consistent with the upholding and violation of moral values, respectively, and demonstrated that behaviors motivated by drives outside MFT’s moral values (i.e., egoistic, or self-focused, motivations including desires for autonomy, competence, relatedness, hedonism, power, and security), were not similarly categorized as prosocial or antisocial. Based on these findings, the authors recommended that scholars, who already implicitly agree with these definitions, work to explicitly use these definitions in their research to help synthesize future work investigating pro/antisocial media and effects. Although this solution would undoubtedly help clarify future work in this area, the task of synthesizing existing published research on pro/antisocial content and effects remains.

The present study explores the extent to which the definitions of prosocial and antisocial media content and effects in extant narrative entertainment research can be categorized according to the definitional scheme proposed by Tamborini and colleagues and therefore scaffolded onto extant theory. We focus on narrative entertainment not only to remain consistent with and extend Tamborini et al.’s (2024) work, but also for theoretical and historical reasons independent of that specific study. Narrative entertainment media have long occupied a central role in scholarly and public debates about media’s influence, from Plato’s Republic to contemporary concerns over television, film, music, and video games. These media forms are distinctive in their frequent and explicit portrayal of characters’ inner thoughts, motivations, and the moral consequences of their actions across coherent story arcs. Such motivational transparency is foundational to moral storytelling and central to seminal theories of moral development through media (e.g., social cognitive theory). Tamborini et al.’s definitional scheme depends on this feature, and they explicitly caution that it “may not be functional for all disciplines,” noting that narrative media is unique in revealing characters’ motivations and that awareness of such motivations shapes moral judgments and audience dispositions (see also Tamborini et al., 2021; Zillmann, 2013). Because the stimuli, measures, and theoretical logic underpinning the scheme were developed specifically for narrative entertainment, applying it to media forms in which motivations are underdeveloped or absent (e.g., some news, social media) would risk theoretical overextension.

In the present work, we attempted to examine the extent to which researchers’ definitions of prosocial and antisocial media content in narrative entertainment align with MFT’s and the MIME’s framework, thereby assessing whether extant literature’s definitional practices can be meaningfully organized within a theoretically coherent structure rooted in character motivation and moral values. Specifically, we asked:

If previous conceptions of pro/antisocial media content/effects can be categorized according to MFT and the MIME, then it stands to reason that the manner in which scholars operationally define these terms could be categorized by this scheme as well. Thus, we asked:

Finally, we examined the extent to which previous experiments in this area operationalize prosocial and antisocial media content and effects as laden with the same moral values. Without theoretical guidance specifying the essence of these terms, we question whether unclear boundary conditions may be responsible for attenuating experimental results in studies that investigate the effects of prosocial or antisocial content.

For instance, if prosocial media is conceptually defined as “helping, sharing, and cooperating,” then operationalizing prosocial content as a television clip emphasizing sharing and operationalizing prosocial effects with a behavioral measure of helping would appear logically consistent. However, the MIME would distinguish acts that are motivated by equality/equity concerns as those that exemplify fairness from acts that are motivated by helping as those that exemplify the moral value of care. According to the MIME, exposure to content that exemplifies one moral value (e.g., fairness) can increase the likelihood of outcomes related to that value (e.g., fairness). However, exposure to an exemplar of one moral value (e.g., fairness) is unlikely to affect outcomes related to other moral values (e.g., care; Tamborini, 2013; see findings by Hahn et al., 2022). Put simply, to our knowledge, no theoretical or empirical evidence suggests that exposure to media content emphasizing fairness would predict an increase in care outcomes (see Hahn et al., 2022).

With this mind, the present study offers a unique opportunity to examine the extent to which published experimental research on pro/antisocial media effects manipulates the same moral values in its stimuli as it measures in its outcomes. In other words, we investigate the extent to which the operational definitions of pro/antisocial media and effects in experiments are related in terms of the moral values they exemplify. We focus this analysis on experiments only, rather than observational or survey studies, because experiments involve deliberate, researcher-controlled operationalizations of prosocial/antisocial media content and outcomes. Focusing only on experiments enabled us to assess the degree of definitional alignment between the moral values that the researchers explicitly or implicitly operationally define as prosocial/antisocial in content and effects, rather than values that emerge from participants’ self-selected content exposure.

Search Strategy

In consultation with a social science librarian, we constructed a search string to identify articles written in English that contained the terms “prosocial” or “antisocial” in the title or abstract, and the co-occurrence of one of the following terms in the title or abstract: media, film, television, movie, video game, or music. We conducted our search across three scientific databases in summer of 2020: PsycINFO, Communication & Mass Media Complete, and Medline. No other limiters were used for the initial search string. This process identified n = 1,015 journal articles. Although the search was conducted in 2020, we note that the years of publication in the list of studies investigated here spans through 2021 due to several 2020 online-only articles being assigned to a journal issue published in 2021. An additional n = 555 articles were identified by (a) compiling articles from previous meta-analyses focused on pro/antisocial media, and (b) searching articles written by scholars known to conduct research in this area. After combining these records and removing duplicates, a total of n = 1,248 articles were examined in a more detailed screening procedure.

Screening Procedure

Next, two trained researchers who hold PhDs in media psychology manually screened and coded all n = 1,248 articles in a two-phase process. First, coders screened the articles’ titles and abstracts to ensure (a) the article was focused on entertainment media and (b) the full text was accessible. Articles focusing on media used for social purposes (e.g., Twitter, Facebook, etc.) were excluded at this point. Both coders applied the screening criteria to a random sample n = 225 (18.03%) of articles. Intercoder agreement was α_Krippendorff = 0.81 (90.22%) which exceeded the acceptable threshold of α_Krippendorff = 0.70 and 80% (Neuendorf, 2010). Disagreements were resolved through discussion with the study’s lead author. The remaining n = 1,023 articles were divided between the two coders and independently screened according to the criteria described above, resulting in the exclusion of n = 686 irrelevant articles and n = 20 unpublished studies. Articles were excluded due to irrelevance if they failed to meet all criteria specified by our preregistration for relevance. To be relevant for inspection in our study, each article had to have: (1) been an academic journal article, (2) been written in English, (3) featured either the term “prosocial” or “antisocial” in its title or abstract, and (4) focused on entertainment media (i.e., media used for entertainment, but not social, purposes, which includes television, movies, books, video games, YouTube, music, and excludes social media such as Twitter, Facebook, Instagram, Snapchat, etc.).

Next, in the second phase, coders screened the full-text of the remaining n = 542 articles for eligibility. In this phase, the same inclusion criteria were applied to the articles’ full-text, in addition to the requirement that relevant articles needed to feature pro/antisocial content or effects as a key variable or concept in the article. Both coders and the study’s lead author examined and discussed all n = 542 articles to determine their eligibility for inclusion based on the criteria above. This process resulted in the exclusion of n = 182 irrelevant articles (per the criteria above), n = 23 inaccessible articles, and n = 39 more duplicates. After screening the abstracts and full articles for relevance, our final population consisted of n = 298 journal articles reporting the results of n = 346 studies. At this point in our procedure, we switched from using articles as the unit of analysis to examining individual studies reported in the articles. We included all n = 346 studies in the scoping review (see Figure 1 for PRISMA-ScR flowchart). A list of all studies examined in the present work is available on OSF: https://osf.io/eu7gw/

OPEN IN VIEWER

Figure 1.

Flowchart of systematic review.

Definition Coding

Following study selection, the same coders extracted several categories of manifest information from each article, including the study type (experiment, survey, content analysis, other), primary age group of the sample (17 or younger, 18 or older, both), medium investigated (video, video game, written text, audio-only, other), total N (for empirical studies with human subjects), total K (for meta-analyses only; see Table 1).

OPEN IN VIEWER

Table 1.

Descriptive Information for All Studies Included in Review.

	Overall		Prosocial		Antisocial		Both
	kstudies	Mn (SDn) A	kstudies	Mn (SDn)	kstudies	Mn (SDn)	kstudies	Mn (SDn)
Medium B
Video (e.g., TV/film)	200	499.89 (1,388.39)	125	384.71 (574.60)	48	835.63 (2,851.31)	25	247.00 (195.42)
Video games	123	685.85 (2,121.75)	86	768.27 (2,459.29)	28	367.75 (406.72)	18	201.19 (223.00)
Written (e.g., books)	5	330.33 (111.55)	4	330.33 (111.55)	–	–	–	–
Audio (e.g., music)	45	117.34 (114.36)	29	130.96 (134.47)	6	246.00 (164.05)	5	246.00 (164.05)
Other	16	742.90 (502.84)	12	810.29 (544.04)	6	1,205.67 (449.00)	4	1,306.00 (585.48)
Population age C
Under 18	140	751.16 (2,389.28)	98	683.52 (2,284.10)	29	1,061.09 (3,150.75	17	356.38 (478.71)
18+	141	272.88 (632.57)	93	286.87 (632.57)	25	206.91 (244.07)	18	166.06 (130.56)
Both	62	640.81 (741.20)	37	665.56 (673.14)	22	410.33 (278.69)	13	469.50 (266.77)
Study type
Experiment	180	315.23 (1,515.73)	125	334.76 (1,799.03)	30	212.83 (319.77)	24	208.54 (340.14)
Survey	78	946.32 (1,913.31)	54	826.13 (1,073.79)	20	1,222.26 (3,355.03)	11	474.20 (280.48)
Content analysis	23 D	461.00 (1,112.65)	14	249.29 (391.85)	14	279.79 (350.91)	7	205.86 (431.82)
Non-empirical	47	–	26	–	9	–	5	–
Meta-analysis	8	Mk: 119.00 (SDk1 123.01)	4	141.75 (161.73)	0	–	0	–
Other	12	557.88 (874.59)	6	727.50 (1,099.51)	4	495.25 (650.50)	2	263.00 (145.66)
Publication year
1973–1979	23	129.21 (166.75)	18	131.88 (181.77)	5	93.00 (54.94	5	93.00 (54.94)
1980–1989	37	116.67 (110.49)	26	111.69 (116.66)	10	226.75 (172.34	8	226.67 (211.07)
1990–1999	42	307.85 (373.94)	25	408.87 (403.29)	12	213.57 (254.35)	6	424.33 (273.96)
2000–2009	47	363.10 (622.51)	21	307.44 (580.23)	14	275.88 (259.88)	5	303.75 (97.09)
2010–2019	190	650.50 (2,048.13)	132	627.91 (1,988.30)	34	912.28 (2,746.99)	24	313.38 (409.94)
2020–2021	7	388.74 (727.00)	6	92.00 (36.45)	1	152.00 (0.00)	1	152.00 (0.00)

Note. Overall column includes information for every study in our final list of articles, including those with no definitions. To appear in the prosocial/antisocial/both columns, studies (or the article they are reported in) must have featured at least one definition indicating whether the study focused on prosocial/antisocial/both. The Both column is not mutually exclusive to the Prosocial or Antisocial columns. Meta-analyses and non-empirical studies did not contribute to the estimation of mean and standard deviation for sample sizes. SD = 0.00 indicates that only one study had an empirical sample or that all studies featured the same sample size. Studies that defined neither pro- nor antisocial content and/or effects do not appear in a separate column.

A: For all study types besides content analyses, this represents the average number of participants in the study sample. Studies that did not report a number of participants did not contribute to these estimates (k = 263 studies reported sample size). For content analyses, this represents the average cases/units analyzed. In other categories (Medium, Age, Year), content analyses, meta-analyses, non-empirical work, and studies categorized as “other” contribute to the k number of studies but not the average number and standard deviation of sample sizes.

B: Each study could feature multiple media types (e.g., if a study analyzed music and video games, it would feature in both categories).

C: Five studies did not specify primary age group or provide statistics describing the sample’s age, and were excluded from descriptive analyses of population ages featured in studies in our sample.

D: One experiment content analyzed its stimuli as part of its manuscript. This article contained no definitions, and thus did not contribute to the estimation of sample size.

Moral Coding

Next, we developed a coding procedure for categorizing the content of each definition according to which moral value the definition exemplified. Specifically, we drew from the moral foundations dictionary (Graham et al., 2009), moral foundations dictionary 2.0 (Frimer et al., 2017), and extended moral foundations dictionary (Hopp et al., 2021) to create a master list of known morally laden words/phrases for each moral value. Each dictionary was created to extract morally-laden content from text. Notably, each dictionary also distinguishes between words that uphold a given moral foundation (termed virtue words) and words that violate it (termed vice words). Thus, by combining these three dictionaries, the master list of words was the most comprehensive list of MFT-based moral words currently obtainable.

Despite the comprehensiveness of this master list of morally laden words, early practice coding attempts suggested that when definitions were unable to be categorized according to MFT’s scheme, it was usually because they used broader, higher-order moral terms to describe what pro/antisocial is, such as “altruism” or “aggression” (both of which connote certain MFT domains but do not neatly fall into one void of context and do not feature in any of the dictionaries). To overcome this issue and attempt to categorize as many definitions as possible in our investigation, we inductively created a coding category for ‘general im/morality’ based on the uncodable definitions in the practice coding round that was aimed at capturing definitions focused on broader moral concepts that, without context, cannot be neatly categorized into one of MFT’s domains. This list included virtue terms (and all possible variations of those terms identified by truncated search), including ‘cooperation,’ ‘altruism,’ ‘benefit others,’ ‘donate,’ and ‘moral,’ as well as vice terms (and all possible variations) of ‘aggression, ‘evil,’ ‘revenge.’ Altogether, our master list contained 1868 unique words.

The two coders then coded each definition according to whether it contained words from the combined master list that represented care, fairness, loyalty, respect for authority, purity, and/or general morality. Coders were able to code as many moral values as were present and identified whether the moral value was upheld or violated in the definition. For example, units defining pro/antisocial as “behaviors that help others” were coded as upholding care, while those defining pro/antisocial as “behaviors that hurt others” were coded as care violations.

Importantly, we used human coders, rather than an automated coding procedure, because human judgment was required to (1) identify the difference between context units (phrases containing each definition) and content units (the definitions themselves), and (2) reference a context unit to make the content unit interpretable (e.g., a context unit for antisocial might read, “prosocial content depicts helping and antisocial content is the opposite” but we only wanted to code the content unit that is relevant to antisocial [in italics]). In the example here, an automated procedure would not have been able to classify the coding unit (i.e., an antisocial definition) without taking into account the context unit (i.e., the prosocial definition), whereas a human could reference the context unit to determine meaning for antisocial’s content unit (i.e., antisocial is defined as the opposite of prosocial [or helping]).

We randomly selected n = 57 definitions for coders to practice coding using this method. After coding the practice items together, we randomly selected 19.53% of the definitions in our sample (n = 100 definitions from n = 49 studies) to be coded individually by each coder for intercoder reliability assessment. Intercoder agreement was acceptable (Brennan & Prediger’s κ_care = .93, κ_fairness = 1.00, κ_loyalty = .94, κ_authority = .96, κ_purity = .99, κ_{general moral} = .94). After achieving acceptable intercoder agreement, we divided the remaining n = 355 definitions between the two coders for solo-coding.

Classifying Prosocial and Antisocial Media Content and Effects Using the MIME’s Scheme

The present study’s findings emphasize the utility of defining pro/antisocial media content and effects as the upholding or violation of the MIME and MFT’s foundational moral values. The vast majority of definitions offered for prosocial and antisocial in this population of literature were able to be categorized according to this definition, which was proffered by Tamborini et al. (2024). Although it might be tempting to interpret our ability to categorize most definitions as definitional coherence across pro/antisocial media literature, we note caution here given the heterogeneity in studies’ definitions of prosocial/antisocial content and effects across the five values we investigated. As discussed earlier, treating a multidimensional concept as though is unidimensional can impede scientific advancement when different scholars use the same words (i.e., prosocial/antisocial) to mean different things (i.e., acts of compassion, fairness, or loyalty). That said, this finding provides additional support for the claim that scholars implicitly define prosocial and antisocial using the MIME and MFT’s foundational values. The present synthesis also helps to add theoretical coherence to the large body of literature investigating pro/antisocial terms. This type of definitional coherence is necessary to reduce knowledge fragmentation, improve the utility of research findings, foster the potential for conceptual breakthrough across studies and contexts (Miller & Nicholson, 1976), and ensure homogeneity among pro/antisocial media studies included in meta-analyses (Carpenter, 2020).

Although a small proportion of definitions were unable to be categorized within each definition type according to this scheme (17% of all definitions), examination revealed that these cases did not meet the criteria for what Chaffee (1991) or Miller and Nicholson (1976) specify constitutes a good definition (i.e., clear essential features and boundary conditions). As such, no scheme could adequately categorize these definitions. Specifically, the inability to categorize these definitions resulted from two primary shortcomings: one dealing with criteria for a good conceptual definition and the other for a good operational definition.

Conceptually, some definitions did not identify the essential features or boundary conditions of prosocial and antisocial, making it difficult to determine what types of behaviors were included in each category. For instance, one example of an antisocial-content conceptual definition stated, “One or more [characters] exhibit prosocial behavior while . . . others display antisocial or inappropriate behavior,” which did not specify the essence of antisocial or inappropriate behavior, making the ability to identify those types of behaviors challenging. Additionally, an example of an antisocial-effects conceptual definition stated, “. . . [content in one genre] will be perceived as more likely to inspire antisocial or maladaptive behaviors than [content in another genre],” which did not note whether antisocial was distinguished from maladaptive or specify the essence of either term.

The second shortcoming occurred when an operational definition referred to specific media content or measures, but did not delineate what occurred in the content or what items were in the measure. For instance, an operational definition of prosocial content that was uncodable stated, “prosocial content was a scene from Mister Rogers’ Neighborhood,” which did not describe the content features or specific acts that categorized content as prosocial. Additionally, an operational definition of prosocial effects that was uncodable stated, “prosocial behavior [was assessed] using . . . the Strengths and Difficulties Questionnaire,” which similarly did not describe content features or outcomes that categorized a participant’s subsequent responses as prosocial. Although more detailed descriptions of scales may be available elsewhere (e.g., many complete questionnaires are published in articles that originally proposed and/or validated the scale), the present investigation limited the context unit of our coding to the individual article, and thus these cases nonetheless lacked detail necessary to assign a code.

Coupled with the finding that 17% of the definitions in this population of studies were uncodable because they did not meet the criteria for a good definition, the fact that 26.01% of the population offered no overt definition of the prosocial or antisocial terms at all further highlights the definitional problems with research in this area and points to the need for greater conceptual and operational clarity. The current approach provides a strong theoretical foundation for future research attempting to synthesize past work in this area or new studies examining prosocial/antisocial media content and effects.

The primary theoretical contribution of this study lies in its application of the MIME to systematically categorize how "prosocial" and "antisocial" have been conceptualized and operationalized across entertainment research, revealing that these terms are primarily defined through the MIME’s/MFT’s dimensions, and that often these terms are not defined at all. By mapping definitions across nearly five decades of research, the study provides a unifying theoretical lens that explains inconsistencies in past prosocial/antisocial findings and offers a coherent framework for future investigations, potentially resolving the definitional ambiguity that has limited the field's ability to detect robust patterns in media effects research.

Moral Values Exemplified in Definitions of Prosocial and Antisocial Media and Effects

Across the population of studies investigated here, pro/antisocial were conceptually and operationally defined most frequently in terms of care and, to a lesser extent, conceptions of general im/morality. First, given that care considerations are particularly valued among Western societies (Graham et al., 2009), and that most studies in this literature were from scholars at Western universities, it is not surprising that investigations of media content and effects define pro/antisocial accordingly. This is consistent with MIME scholars who have demonstrated that care is the most prominent moral value in entertainment content (e.g., Aley et al., 2021). Second, the frequency with which im/morality has been defined in general terms appears to be driven by the facts that (a) many studies conceptually defined prosocial content/effects in line with Eisenberg and Fabes (1998), who conceptualized prosocial behaviors as those that benefit another, and (b) many studies conceptually defined antisocial simply as “aggression.” Importantly, MIME logic adds critical nuance beyond general conceptions of im/moral behavior by focusing on a set of comprehensive, specific moral values. Future researchers could leverage this nuance to assess media’s ability to prompt more precise attitudes/behaviors than “benefit” or “aggression.”

Our results also point to a smaller but still substantial focus on values related to fairness, loyalty, respect for authority, and purity. In particular, the FLAP analysis depicted in Table 3 shows a mix of moral foundations throughout definitions, which demonstrates that prosocial/antisocial have, historically, been defined with considerable conceptual heterogeneity. The array of moral values investigated by previous studies again points to the benefit of adopting the definition for pro/antisocial content and effects offered by Tamborini et al. (2024). Our review showed that many studies defined content or effects simply as pro/antisocial. While these studies seemed to examine the same constructs, nuanced inspection revealed that a considerable number of cases were examining the distinct dimensions of fairness, loyalty, authority, or purity. This nuance has important implications for scholars wishing to assess media’s role in promoting or discouraging separate values, such as equality/equity, inclusion, deference to social institutions or traditions, or social contamination. In particular, parents or practitioners who use media to instill specific values in children may find value in this more nuanced understanding of media content and its potential effects.

Implications for Meta-Analyses in Research on Prosocial and Antisocial Effects

Our study’s demonstration of substantial heterogeneity in the moral values investigated by prior research on pro/antisocial media content/effects also has important implications for scholars examining effect size averages across this literature. Meta-analysis guidelines emphasize that the inclusion of heterogenous studies under the assumption that they are homogenous can doom attempts at assessing a body of work’s average effect size (i.e., comparing apples to oranges; Carpenter, 2020). The findings of the present study suggest that meta-analyses in this area are subject to this bias. In essence, the present work suggests that prosocial and antisocial could be considered higher-order constructs that, although potentially useful in everyday conversations about media content and effects, may lack necessary specificity in research examining these constructs.

Further complicating the issue, the results of the present investigation revealed that, in most experimental studies, the moral values manipulated were mismatched to those measured. These mismatches may obscure effects in individual studies. Based on MIME logic, we might expect that content exemplifying one moral value should increase the importance audiences place on that specific value, but not other values (Tamborini, 2013; see findings by Hahn et al., 2022). For example, if a researcher manipulated care in media content and predicted it would affect people’s care-related attitudes and behaviors, but then measured loyalty-related attitudes and behaviors, this measure of the expected outcome would be invalid and likely mask existing effects.

Together, results suggesting great heterogeneity in the moral values investigated by prior research, and that most experimental studies mismatch the moral values manipulated and measured, reveal problems in existing meta-analysis research on pro/antisocial media content and effects (also see Pfattheicher et al., 2022). These problems suggest that meta-analyses investigating the higher-order constructs of prosocial and antisocial may be pooling the effects of conceptually heterogenous concepts (i.e., comparing apples to oranges), potentially hiding the existence of robust effects linked to patterns with important social implications. Building on the findings of the present study, future researchers should attempt more precise meta-analyses focused on the different moral dimensions comprising prosocial and antisocial effects research (e.g., studies focusing on care compared to fairness). Additionally, future researchers should endeavor to determine whether studies investigating operationally similar concepts (e.g., care media influencing care effects) have produced evidence of stronger effects compared to those which investigate operationally distinct concepts (e.g., care media influencing purity effects).

Further, MIME-based logic points to several potential moderators that future researchers may wish to examine in meta-analyses of pro/antisocial content and effects research. These moderators might include effect size differences associated with the importance audiences place on the five separate moral values, or variability in media content and effects linked to these values in different cultures or age groups. The MIME suggests the existence of different morality subcultures, across which we should expect exposure to morally-laden content and the effects of said exposure to vary (Eden & Tamborini, 2017; Tamborini et al., 2012). For example, we might expect American conservatives and liberals to select and respond differently to content emphasizing values such as care (in which liberals place great importance) or respect for authority (in which conservatives place great importance; Graham et al., 2009). Further, we might expect citizens of the US and India to select and respond differently to media content associated with values of care versus authority (e.g., Prabhu et al., 2020). Similarly, young child audiences who are often reminded to defer to adults may be more sensitive to authority-laden media content’s effects than older adolescent audiences who are encouraged to think for themselves (e.g., Aley et al., 2021; Cingel & Krcmar, 2020). Future meta-analytic research should investigate these possibilities in pursuit of a more precise understanding of media’s influence on different audiences.

Limitations

The study’s first limitation concerns our use of the words in three separate moral foundations dictionaries (MFDs) as a coding scheme for classifying the moral values inherent to prosocial and antisocial definitions. Despite the excellent efforts to develop these dictionaries, their combined list of words is not exhaustive to the moral values described by the MIME and MFT. Thus, it is possible our use of these dictionaries meant that our measurement of the moral values was not sensitive enough to detect subtle references to moral values in various definition types. Even so, we were able to categorize over 80% of all definitions in this study, and the vast majority of those unable to be categorized failed to specify essential features of the terms they were attempting to define (as described in the discussion section). Nevertheless, future researchers attempting a procedure similar to ours may wish to adopt a more latent approach to coding, similar to previous MIME studies (e.g., Prabhu et al., 2020).

Second, in the second step of screening (wherein we scanned the articles’ full texts to determine their inclusion eligibility), we opted for a procedure in which both coders and the lead author examined and discussed the application of the inclusion criteria to each of the n = 542 articles. The decision to rely on this “expert discussion” procedure instead of independent coding was made to minimize the chances that relevant studies were excluded from our investigation. Although this procedure potentially led to the inclusion of studies that do not feature a substantive focus on pro/antisocial content/effects, we note that these studies would have been excluded in the subsequent coding phase for not offering clear definitions of these terms.

Third, this study focused exclusively on extracting definitions from articles focused on entertainment that were written in English published prior to 2020. Given that the salience of moral values has been shown to differ across cultures, and that media content varies across cultures in ways that mirror these different moral values (Prabhu et al., 2020), it is possible that the inclusion of non-English articles or articles published after 2021 in this investigation would have highlighted even more divergence in the moral values that make up the essence of prosocial/antisocial entertainment media content and effects. Additionally, extending the current study’s procedure to investigate prosocial/antisocial definitions in other media forms outside of entertainment (e.g., news, social media) would assess the generalizability of our findings across media contexts and potentially reveal interesting differences in how prosocial and antisocial concepts are operationalized across different media platforms and genres. Future researchers should investigate this possibility by expanding their search to include articles investigating prosocial/antisocial in other media genres, written in languages other than English, and published after 2021.

Finally, as with any systematic review, this study is limited by the breadth and depth of its search strategy and results. Because we can never be certain we have obtained every record of a study in this field of literature, we encourage future researchers to attempt to replicate the present work using other search terms/strings. Although our focus here was intentionally limited to narrative entertainment for the theoretical and methodological reasons described earlier, the MIME’s moral-value framework could also be applied to other forms of narrative media when character motivations are made sufficiently clear. Examples might include documentaries, long-form news features, or social media storytelling. Examining whether the definitional patterns observed in narrative entertainment research also appear in these contexts would help assess the framework’s generalizability, while also identifying genre-specific tendencies in how prosocial and antisocial are defined.