Many scholars discuss the sexist ways women are depicted in and around video games (Beasley & Standley, 2002; Behm-Morawitz & Mastro, 2009; Carvalho & Cappelli, 2018; Cote, 2018; McArthur, 2018; Mou & Peng, 2009). This includes quantitative and qualitative explorations of biased, stereotypical, reductive, and potentially harmful visual and narrative depictions of women in games.

Many scholars examine sexist depictions by focusing on the frequency of women characters and the qualities of their visual appearance within videogames and associated paratexts (Burgess et al., 2007; Dill & Thill, 2007; Gardner & Tanenbaum, 2018; Gestos et al., 2018; Miller & Summers, 2014; Shaw & Friesem, 2016). Although scholars observe an increase in the amount of women characters present in videogames (Lynch et al., 2016), men are still much more frequently main characters compared to women when present (Gardner & Tanenbaum, 2018; Maclean, 2016). Other scholars observe how women are still frequently depicted with unrealistic bodies and exaggerated sexualization (Downs & Smith, 2010; Glaubke et al., 2001; Jansz & Martis, 2007; Tompkins & Martins, 2021), complicating the quality of increased representation.

Scholars also examine narratives, roles, and dialogues of women in video games (Butt & Dunne, 2017; Hart, 2017; Gray et al., 2017; Kennedy, 2002; MacCallum-Stewart, 2014; Masso, 2011; Perreault et al., 2018; Santonicollo et al., 2023; Sherman, 1997; Ugolotti et al., 2020). Many of these studies suggest women characters often give insignificant or demeaning roles in relation to main (men) characters (Friedberg, 2015; Masso, 2011). For example, Heritage describes how men in the popular Witcher series more often have active roles with more power (e.g., kings or military leaders) while women more often serve as advisors and onlookers (2022).

Despite these negative trends, several scholars also examine more contemporary characters who challenge stereotypical depictions (Allen, 2023; Lankoski, 2004; Lucas, 2019; Solska, 2022). For example, Solska identifies a sample of three games that present women characters in narratively focal roles in which they interact with other characters and environments instead of only being submissive and passive (2019). These studies encourage us to consider what and how other aspects of characters may be sexist, and how to begin to confront the issue beyond characters’ surface appearance.

A wide variety of scholars identify how norms can dramatically shape how people experience social, professional, and recreational communities such as those that surround games (Auerbach, 2014; Campbell, 2014; Cote, 2017; Edidin, 2014; ESA, 2023; Fox & Tang, 2014; Gonzalez et al, 2014; Gray et al., 2017; Hart, 2017; Koh & Das, 2017; Lenhart et al., 2008; Leonhardt, 2021; Mou & Peng, 2009; Ochsner, 2017; Reza et al, 2022; Shliakhovchuk & Muñoz García, 2020; Tompkins et al., 2020; Weber, 2012; Williams, 2003), and can contribute to the stereotypes discussed in the previous section. We organize our research with this theme that describes norms, attitudes, and structural forces that deter women from accessing the digital game medium, its production, and associated communities to the fullest.

Several studies suggest the predominantly male game industry perpetuates a sexist workplace culture as well as negative or harmful portrayals of women (Gray et al. 2017; Mou & Peng, 2008; Tompkins & Martins, 2021. Gaming has too often been considered ‘an activity created by men and for men’ (Carvalho & Cappelli, 2018), with many women professionals testifying to workplace discrimination and/or harassment (Weber, 2012). Sexism in the gaming industry makes it harder for women to succeed professionally and facilitates ‘the sexualization of female game characters’ (Tompkins et al., 2020). These studies help support the persistent claim that increasing equity and inclusion, and hiring more women, are necessary in the design process to address the sexistnormative industry environments that perpetuate sexist representations (Campbell, 2018; Doerrer & Campain Live, n.d.; Elderkin, 2023; Kennedy, 2018; 2002; Passmore et al., 2017; Qin et al., 2009; Reza et al., 2022; Solska, 2022; Wood & Wood, 2008).

Perceived market factors can drive sexist representations of women in games and reproduce past narratives as a risk management strategy (Stermer & Burkley, 2015; Ugolotti et al., 2020). Tompkins and Martins describe demands for a ‘success formula’ (e.g., Figure 1) to guarantee return of investments based on repeating past (men-dominant) successes (2021). However, this approach heavily depends on assumptions about the ‘real world’ that do not always align with the actual population of players (Duggan, 2015a; 2015b; Lenhart et al., 2008; Lenhart et al., 2015; Passmore et al., 2017; Shaw, 2009; 2014).

Player-based gaming communities are also historically discriminatory towards women. There is a common refrain that ‘girls cannot play, are mechanically unskilled, and owe their achievements to a male counterpart who has helped them get there’ (Koh & Das, 2017). Scholars identify how women players are regularly subject to harassment and sexist environments (Aghazadeh et al., 2018; Bezio, 2018, Campbell, 2014: Condis, 2015; Edidin, 2014; Massanari, 2017; Mortensen, 2018; O’Leary, 2012; Wingfield, 2014). with some suggesting lax consequences and scrutiny of such norms contributing to their proliferation (Fletcher, 2012). Fox and Tang identify how endorsing broader masculine norms is a predictor of sexist attitudes about women playing video games (2014). While Cote reveals that women players are often capable of managing online game-related harassments, it is an inequitable burden that they need to adopt coping strategies at all simply to participate in the medium (2017).

Despite the wealth of studies identifying poor representation in games described in section 2.1 and beyond, there are surprisingly few studies examining how players and people surrounding the medium themselves perceive and interpret representation. Beyond the general influence on formal and informal norms discussed in section 2.2, negative visual and narrative portrayals can be linked to harmful outlooks on life (Stermer & Burkley, 2015). We organize studies that analyse the influence of the representation on people and community perceptions (emphasizing harmful effects) or that analyse how people or communities themselves perceive or interpret that representation.

Studies show how representation in video games can influence player attitudes and perceptions about gender (Bowey, 2017; Liu, 2018; Mou & Peng, 2009; Shaw, 2014; Stermer & Burkley, 2015; Yao et al., 2010). Frequent engagement with sexist games links to sexist attitudes in players and negatively impacts women’s well-being beyond games (Fox & Bailenson, 2009; Gabbiadini et al., 2016; Scharrer, 2004; Summers & Miller, 2014). In a meta-analysis, Gestos et al. demonstrates how increases in objectification in games can diminish women players’ self-efficacy and well-being while promoting sexist views in male players (2018).

Though fewer in number, some studies examine how players or potential players themselves perceive, interpret, and respond to representation in games (Desai et al., 2017; Stermer & Burkley, 2015; Gardner & Hacker, 2022; Tomkinson et al., 2020). Kondrat found that 76.9% of participants in their study recognize negative stereotypes in women characters in games, generally, and 31.62% identify sexual objectification specifically, with a majority of participants desiring less stereotypical characters (2015). Other scholars identify how the interpretations of character representation, and their influence can vary between different player demographic identities (Gardner & Corvite, 2023; Liu, 2018; Nguyen et al., 2020; Reza et al., 2022). Our study further explores community perceptions focusing on women character narratives rather than focusing just on researcher-centred content analysis of physical appearances.

Our primary analysis is of participant impressions rather than the videos themselves. That is, rather than us defining or assigning sufficiency to—for instance—introduction duration and asking participants how strongly they agreed with us, our key data are participants’ own interpretations of sufficient understanding, within the scope of provided clips. We chose a survey during our preliminary stages to reach larger and more diverse audiences and to better understand the state of the medium. Our survey was organized into three sections on a Google Form. The first section asked participants about demographic characteristics including age, primary language, gender, race, gaming habits and aspects they felt make characters more relatable. The second section showed participants one-minute video game clips introducing six sampled characters outlined below in section 3.2. Following each clip, we asked participants to respond to four statements about their impressions of each character/introduction based on time, dialogue, how strongly they felt the characters resemble any women they know in their life, and their perception of the character’s importance to the game’s narrative. These questions utilized Likert scales to evaluate how strongly participants agreed or disagreed with statements about our sample of characters and their introductions. The third section asked questions about participant thoughts on the roles of the sampled characters and any other remaining general thoughts or feelings.

In November 2023, we selected the three best-selling ‘story rich’, single-player games with women protagonists and the three best-selling games with men protagonists, where women characters were introduced (with dialogue) within the first two hours of gameplay. We sampled from Steam because it is the largest digital games retailer, accounting for 61% of all full game revenue from new releases in 2023 (Fragen, 2024). We did this by going to their digital store and selecting ‘top sellers’, filtering by the tags above, and identifying the best-selling games by units sold on Steam. We chose to look at recent mainstream games from 2023 to capture perceptions about a relatively current snapshot of the state of the medium. We narrowed down our game selection to games meta-tagged with story-rich and single-player as these generally suggest heavier character narratives and/or story-driven goals. We excluded games that had deuteragonists or customizable main characters, non-human characters, and games without clear dialogue for women characters in the first two hours of the game. We made these exclusions to maintain a clear focus on the representation and narrative introduction of women characters.

We chose to focus on one-minute clips to be respectful of participants’ time by providing shorter length videos, only capturing first scenes that introduce the characters. For games with women protagonists, we found clips of their introductions/first dialogues using commentary-free gameplay walkthroughs, or ‘let’s play’, on YouTube. Similarly, we also identified clips for the first appearances/introductions of supporting women characters with dialogue within the first two hours by first watching walkthroughs to confirm on YouTube.

We wanted to observe potential differences in perceptions between games led by women protagonists and games with women but led by men protagonists. We chose sub-samples of three of each game-type and six games total to reach a minimum number of examples to make comparisons while also trying to be considerate of participants’ time and the length of the survey.

To be eligible, the only requirement was for participants to be at least 18 years old and speak English—because the survey was in English—regardless of their gaming experience. However, we were especially interested in the perspectives of women and anyone who feels historically underrepresented in media.

Our team digitally and physically distributed the survey in three waves. The first two waves involved social media platforms such as Discord, email lists, Facebook, Instagram, LinkedIn, and Slack. Our team members each represent unique interdisciplinary and demographic backgrounds and—at time of data collection—they were based at four different institutions in California, Indiana, and Texas. We began with personal networks and moved on to forums and community groups related to gaming—either public groups online or associated with student groups on our respective campuses. To further increase participant diversity, we made additional efforts to include communities centred on women and gaming such as the ‘women in gaming’ discord or inclusive gaming network and the game HERs. The third wave involved posting flyers and recruiting faculty at our respective institutions to physically solicit additional possible participants. Because our sampling relied heavily on university groups centred in North America, participation was biased toward college-based/aged communities, likely residing in North America.

We began by using descriptive and comparative statistics to identify basic trends in closed-ended responses by all participants and between participant groups (e.g., race, gender, or frequency of play). We then identified, applied, and iterated on emergent descriptive qualitative codes for open- ended responses.

We ran three waves of Welch’s t-tests focused on responses to the second section of the survey. Responses to the Likert scales were converted to integers 1 to 5 to calculate averages and make comparisons, with 1 representing strongly disagree, 5 representing strongly agree, and 3 representing a neutral response.

The first wave consisted of high-level t-tests within and between groups based on identified race and gender. Specifically, we compared responses between participants who identified as White and those who did not, and between those who identified as men and those who did not. We chose to start with these separations to see if we could observe any differences in perception between those who are historically well-represented in games and media and those who are not. That is, white men are often more represented than other folks in games (Gardner & Tanenbaum, 2018; Passmore et al., 2017; Williams et al., 2009), and in the game industry (AGA, 2023). We calculated each participant’s mean response to all six copies of the repeating questions in the second section of the survey and tested them between groups to identify any significant differences. We compared responses by these groups across all six games, across men-led games, across women-led games, and for each individual game. One participant who preferred not to state their gender was only included in analysis based on identified race and of total participants.

The second wave involved the same comparisons across games and sub-samples of games, but tested responses based on intersectional demographic identifiers combining two variables such as race, gender, and frequency of stated gaming activity (e.g., every day, a few times a week, a few times a month, rarely, never). We did, however, still primarily with a focus of comparing the ‘default’ against everyone else. For example, we compared responses between white-men and not white men, white-women, and not white women, and those who identified as white and as playing a particular frequency and those who did not identify as white who played that frequency.

We conducted the third wave after qualitative review of open-ended responses, specifically participant comments relating to the cultural context of characters (discussed more below). In this final round of tests, we compared responses to the introductions in Half-Life: Alyx, Red Dead Redemption 2, and The Invincible, realizing based on responses that they were Western (North America and Europe) developed games, and Asian-developed games: Lies of P, Sekiro, and Witch Spring R.

We created and applied emergent, descriptive codes to responses to the open-ended questions in the final section (Saldaña, 2015). Each author independently read and coded open-ended responses to independently identify themes. The authors then came together to discuss, aggregate, and iterate on thematic codes, and consider how these themes and participant quotes related to quantitative data.

Our quantitative results showed that all participants, on average, were more likely to know at least one woman similar to the characters that appeared in games developed in a Western context compared to those developed in an Asian context. Relatedly, we observed multiple participants not only acknowledging the role of a developer’s culture in open-ended responses, but also identifying tropes and stereotypes that may rely on cultural reference such as one participant’s use of ‘kawaii’, a Japanese trope for ‘lovable’ or ‘cute’.

When asked how strongly they agreed or disagreed with whether characters were similar to at least one woman they knew in real life, participants, on average, were more likely to agree for characters developed in a Western context compared to those developed in an Asian context (p=0.0190). Technically, participants neither agreed nor disagreed, on average, that they knew at least one woman similar to the characters created by Western developers (x̄=3.16), However, they disagreed, on average, with knowing at least one woman similar to those in games created by Asian developers (x̄=2.24).

We also observed a variety of open-ended responses that directly recognized cultural components of video game developers as well as their impact on woman character development. For example, three participants explicitly mentioned Japanese culture and context for development/consumption when describing Witch Spring R (‘Japanese-made’; ‘J-game for a J- Audience’; ‘Her creators might have hailed from Japan’). Another described this character saying she is ‘cute and her speech being kawaii.’

We use cultural context to describe the customs, norms, values, and language surrounding a community. Our findings indicate participants implicitly and explicitly reacted to cultural influences in the sampled games, notably, when they live in the same cultural context as the game’s developers. For developers, these results should indicate how their experiences may influence content, and how it may resonate with different players.

We observed that participants—especially in aggregate—often explained their perceptions of characterizations and character roles in sampled introductions using oppositional descriptors. We coded 62 comments out of 110 open-ended responses (56.36%) with one or more of the following codes: Strong vs Weak, Main vs Support, and Important vs Basic. Strong vs Weak Characterization and Main vs Support were our two most common codes applied overall. Moreover, participants used descriptors from multiple oppositional axes, perhaps describing a character as weakly characterized, in a supportive role, and basic.

We used strong vs. weak characterization to describe participant comments about the general depth and complexity of game characters. Participants did not always use the words ‘strong’ or ‘weak’, but referring in some way to general complexity, detail, or lack of either were cause for applying the code (e.g., ‘very strong characters and personalities’, ‘headstrong’, ‘[not for use of strong, but strong impression of character]’, ‘pretty vague’, ‘very little detail’, ‘lack depth’, ‘understated or flat’). Both similarly and conversely, we applied Main vs Support to participants using explicitly contextual or game-mechanical terms to describe characters’ relative roles (e.g., main character, support character; ‘helpers or one-off’, ‘assistive to the main character’). We applied Important vs. Basic to participant comments discussing a character’s perceived overall relevance to the narrative (e.g., primary cast member or once-met NPC; “helpers or one-off” again).

We should also note that while we saw comments on both sides of these oppositions, participants more frequently described characters in our sample as weak, supportive, and/or basic. For example, from the 38 responses coded with ‘strong vs weak characterization’, 32 contained comments about weaker characterizations while only 11 contained comments about stronger characterizations; we had five responses that discussed both ends at the same time. We received multiple similarly worded comments relating women characters to the more negative sides of oppositions and men characters to the more positive sides:

It seems like for most of the female characters presented here serve a role to aid the protagonist/player in some way, making them important to plot, but don’t serve any further purpose. Then there are the female characters that we actually play as and see their perspective. They seem to exert more personality to a degree.’; ‘It seems like most female characters were side characters that help the main protagonist. Though some clips showed strong female leads which I believe to be was nice to see’; ‘They all seem to be like not thoroughly represented to me.’; ‘Most of the time seemed to be side characters with little depth beyond what is required to give a mission’; ‘Apart from a few exceptions most of the presented roles seem to be assistive to the main character.’

Although participants occasionally mentioned some ‘strong’ women characters, an overwhelming amount of responses discussed that all or most women characters portrayed in these games seemed to have shallow characterization and assistive roles.

Overall, we observed that all participants, on average, felt neither the time nor dialogue of the sampled introductions were sufficient or insufficient enough to understand sampled characters. In fact, we saw near consensus across all six games—with several p-values above .95—with participants neither agreeing nor disagreeing (averaging 3.11 and 3.24, for time and dialogue respectively) that time or dialogue of the introductions were sufficient to even begin to understand these characters. We did not find a significant difference between responses to these questions based on demographics, individual games, across all six games, or whether games were women- led or not. Although neutrality itself is not a finding, in this case it presents an inconclusive result when attempting to analyse participant perceptions.

Despite the neutrality displayed, on average, in our closed-ended questions, many open-ended responses discuss the seemingly incompleteness and length of introductions or the depth and quality of the dialogue. Several participant comments coded with time—to refer to explicit references to the concept—expressed that the introductions felt ‘understandably short’, ‘to the point’, ‘rushed’, ‘unexplained’, or ‘did not have much introduction time.’ Similarly, many comments coded with dialogue—for the same reason as time—described it as ‘pretty vague’ or clearly stated that ‘none [of the characters] had sufficient enough dialogue.’ The inconclusive contrast between expressive open-ended responses and neutral quantitative responses suggests there may have either/both been underlying issues with our survey construction, or the content of the clips we used themselves, discussed below

Our findings reinforce the notion that the perceptions and assumptions of developers are embedded in the games they make—and that players and non-players alike are often aware of this influence, if only implicitly. We did not in any way indicate the development context of games in our survey, nor did we intentionally sample three western-developed games and three Asian- developed games. Yet, participants observed differences in the characters and narratives between these groups of games and found they were more familiar with women characters developed in their own likely cultural context. These findings highlight how important it is to consider cultural contexts when developing or analysing character designs and narratives—especially when they are introduced or creating their first impressions—or when examining how potential players perceive those characters.

Our study suggests that if game makers want to develop characters that resemble women from more diverse cultural contexts, they need to include more diverse perspectives in the development process–specifically in roles with significant influence on the creation of character narratives or their narratives.

The use or production of culturally-linked stereotypes and tropes may also influence how potential players—and those around them—perceive game characters. Our analysis highlights how sexist representations previously analysed by many scholars described in sections 2.1 or 2.3, such as Summer and Miller’s emphasis on oversexualization of women in game-related media compared to men, may be more closely tied to regional, cultural contexts than previous scholars have acknowledged (2014). Better understanding how developer and player perceptions influence—or are impacted by—characters is important for mitigating the perpetuation of stereotypical conceptualizations that reinforce negative representations of certain groups in games.

Our three-axes visualization allows future researchers to pivot from Likert scales to ask different, more specific, sorts of questions as part of surveys, interviews, or focus groups. How participants, players, or potential players might chart their perceptions allows us to ask more precise questions about, for example, ‘how’ and ‘why’ they may position certain characters, or how/why they position them relative to each other—potentially inviting deeper, multi-dimensional insights even in relatively brief interactions.

Such frequent use of specific oppositional descriptors also urges researchers to question why we are seeing this trend in the first place. Perhaps participants genuinely perceive men and women as oppositional descriptors, or perhaps women characters are developed and presented to fit within the three axes. Either way, our visualization can help encourage research questions that reflect on these oppositions, their origins and beg a deeper understanding of game consumer’s perception.

Participants leaning toward describing presented portrayals of women characters on the weaker, supporting, basic ends of identified axes suggests that game developers should be mindful of their perceptions on gender when developing characters. As discussed in the previous section, it is likely that perceptions of game developers are embedded in their creations. Hence, bringing more women’s perspectives to the workspace may help towards robust character development and overall success of the game if game developers aim to improve women representation in video games.

The inconclusiveness we observed on the ‘sufficiency’ of time and dialogue in introductions may stem from the durations of clips, the content of clips, or how we described/presented/asked about clips/introductions. Because we were presenting participants with 6 clips of footage, we attempted to make clip durations digestible. It is tempting to consider adjusting clip durations— perhaps making them longer—as an immediate fix. Our focus on clip quantity may have influenced the neutral responses, on average, perhaps due to participants hurrying through or due to clips actually being too short but participants being uncertain how to respond. However, because participants perceived them as neither sufficient nor insufficient, simply changing or increasing the duration of clips could remain ineffective in identifying participant perceptions. Alternatively, the nature of the in-game introductions our clips contain may be the issue. In this case, participants’ neutral consensus could be communicating, on average, that time and dialogue simply did not positively or negatively affect their understanding of the characters. The inconclusiveness, then, may be driven by the quality of dialogue, narrative, or overall characterization. As an additional note, we also must acknowledge in hindsight that participants may have understood the use of ‘introduction’ in two ways: the character’s in-game treatment or the clip itself embedded in the survey.

Modifying question structure and asking additional questions to clarify how time and dialogue affect participant perceptions and better acknowledging how the quality of character introductions can affect initial understandings of the character can help similar studies avoid similar inconclusive results. Rephrasing questions can reduce potential misinterpretations of ‘introduction’. Asking high-level questions about how participants consider the role of time and dialogue to influence their experiences, or about the quality of sampled introductions can provide better understanding of how they perceive those introductions. However, it bears mention that the issue may simply be issues with the inherent quality of those introductions themselves and a need for developers to produce introductions that allow participants to develop initial understandings of characters more effectively.