‘Stupid Girls’ vs. ‘Real Men’: Identity Construction and Social Media Polarization over Revenge Porn

not conforming to societal views of common-sensical sexual behavior or praised for their strength in standing social judgment.


Introduction
This paper explores an increasingly prominent phenomenon: revenge pornography.Revenge pornography (henceforth "revenge porn") refers to "the nonconsensual sharing of intimate images" (Bond andTyrrell 2021, 2166), videos, or other intimate material, as a form of retribution.Although revenge porn does not occur exclusively via digital sites, since private intimate material can be shown offline, perpetrators usually resort to online internet sharing to exponentialize the repercussions of the crime.Crimes of revenge porn have devastating consequences for its victims, as they disrupt their daily lives, confidence and self-esteem, preventing them from leading a normal life.Several sources show that around 75-90% of revenge porn victims are women (UK Safer Internet Centre 2021; Channel 4 News 2022), statistics which pinpoint that this crime is mostly gendered.Revenge porn is particularly stigmatizing, for many victims are blamed and slut-shamed by their environment (Franklin 2014;Mckinlay and Lavis 2020) and the consequences of this crime are long-term: victims encounter several obstacles in having the intimate material removed from public sites and social media, as it is usually re-uploaded multiple times, leaving victims with little to no control over the aftermath.Due to the notoriety of these crimes, different campaigns and movements have attempted to raise public awareness via mass and social media.Nevertheless, these have sometimes raised controversy over the legitimacy of victims' and perpetrators' actions and claims.
In this paper, I explore revenge porn as constructed in the discourse of social media.Particularly, I examine the social networking platform YouTube and analyze users' comments and replies on videos featuring cases and victims of revenge porn.Previous studies on social media discourse have shown that users engage in misogynistic commentary when addressing topics of violence against women (henceforth VAW) (Bou-Franch and Garcés-Conejos Blitvich 2016;Poland 2016).Likewise, users can also engage in activism through the collective use of hashtags as a way of unifying their voices to stand against VAW (Jones et al. 2022).The co-existence of conflicting ideologies on social media tends to give way to discursive polarization in users' interactions.This study aims to explore how this conflict unfolds when addressing the crime of revenge porn.
Although discourses of VAW on social media have been largely investigated, the discourse around revenge porn crimes remains under-researched (but see Hall and Hearn 2019).Similarly, research on polarization has mainly focused on political discourse, overlooking issues of VAW and, consequently, revenge porn.This study attempts to overcome this lacuna by looking into the discourse practices that emerge in YouTube users' comments about revenge porn.Particularly, the analysis aims to bring to light how YouTube users represent and evaluate crimes of revenge porn, its victims, and its perpetrators.Accordingly, the following research questions intend to guide the study: 1. How is the discourse around revenge porn linguistically characterized in YouTube users' comments?What linguistic patterns are used to discursively construct the crime of revenge porn and its involved participants?2. How are evaluative resources used to blame or acquit revenge porn victims and perpetrators?How are gendered social identities negotiated?This paper adopts a Corpus-Assisted Discourse Studies (CADS) approach to identify major themes and overarching linguistic patterns in the corpus.In particular, frequency, collocation and keyness analyses are drawn.Additionally, a qualitative analysis of concordance lines delves into the discursive behavior of keywords and collocates to further explain evaluative and grammatical patterns.

Digital violence against women and feminist activism on social media
Social media have become sites of communicative democratization due to the relatively easy access to online output and interaction.Traditionally, information sharing and reception were mostly controlled by news media and institutions.More recently, digitally mediated communication has allowed the general population to take part in the negotiation of discourse practices.This has been possible with the advent of social media, which KhosraviNik (2017, 582) defines as "the intersection of mass and interpersonal communication", giving way to a "new paradigm of communication".Although digitally mediated communication has developed for more than two decades, within the last ten years our daily communicative practices have increasingly become encoded in the online world.Digital practices have thus become an extension of our offline reality, which includes the production and reproduction of ideologies "as part of a larger social world" (Garcés-Conejos Blitvich and Bou-Franch 2019, 4;Jones, Chik, and Hafner 2015).
Like in the offline world, this ideological work may pursue the reproduction and/or challenge of power imbalances and domination, as individuals strive to put forward representations that favor them and contest those that disfavor them in the power struggle (Fairclough 2001;Reisigl and Wodak 2001;Van Dijk 2015).Computer-mediated interaction is naturally conflictual and thus might reproduce and perpetuate social inequalities.This remains true with issues of VAW, as debates over gender inequality, misogyny and sexism often take place on social media.For instance, Hardaker and McGlashan (2016) explore the language of (sexual) aggression on Twitter1 .Drawing on corpus linguistics tools, they find that users semantically associated rape with threats, with women being the targets of these threats.Rape threats are viewed as a weapon to control women's discourse online.Conversely, discourses around rape and men "involved the construction of ''real'' masculinity as one that categorically excludes the use of threatening or violent behaviour towards women."(91).
Discourses of VAW, albeit reproduced in everyday offline discursive practices, are particularly prevalent on social media.Generally, social media users incur hate speech with greater disinhibition than in offline contexts, "express[ing] opinions online that they would never voice if they knew that those opinions could be attributed to them offline" (Hardaker and MacGlashan 2016, 82).In this sense, social media communicative properties like detachment enable users to convey violent and hateful ideologies, taking extreme positions and expressing them in a seemingly inconsequential way.Anonymity is another significant factor in the perpetuation of such ideologies in that users (re)present themselves and others without the restrictions of social norms, causing "dysfunctional social behaviours" (Ghaffari 2020, 161).
Recent years have seen, however, a shift in which overt sexist and misogynistic language is gradually being considered unacceptable, and genderbased inequalities against women are being contested (Anderson and Cermele 2014).In this line, Palomino-Manjón (2022) examines Twitter users' discourse on the hashtag #WhyIDidntReport, which emerged as a response to Donald Trump's questioning of Dr. Christine Blasey Ford's decision to not report her case of sexual assault when it occurred.In her analysis, she finds that the hashtag, used by victim-survivors of sexual violence, has an affiliative function whereby they denounce the scarcity of social response to end sexual violence and further condemn institutions for reinforcing rape culture.Moreover, victim-survivors use the hashtag to share their story and emotional suffering, enhancing empathy and offering alternative readings of victim experiences that challenge hegemonic conceptions of female victimhood.Similarly, Jones et al. (2022) focus on the hashtag #NotAllMen and encounter comparable results, as users affiliating with this hashtag express "a predominantly anti-misogynist stance" (8).Born as a claim for men to not be homogenously considered as a threat to women, #NotAllMen is reclaimed by women, who wish to bring back the focus to women's experiences through, for example, the imperative 'stfu' or strong expletives.These allow women to resist oversimplistic denialist conceptions of sexism and misogyny and "reject the ideological expectations of hegemonic femininity" (8).These two studies illustrate how social media constitute online sites of struggle and negotiation of social representations and discourse practices, opening the way for social change.In this respect, social media affordances -collective interaction, anonymity, replicability and audience impact (boyd 2010; KhosraviNik 2017) -admit the co-existence of both dominant and resistance discourses in the (re) construction of social groups, identities and ideologies, and, more specifically, the perpetuation or challenge of discourses of misogyny and VAW.
Yet, despite feminist efforts to counteract gender-based inequality and patriarchal conceptions and stereotypes over women and victims-survivors, "digital spaces remain fraught with gendered violence" (Mendes, Keller, andRingrose 2019, 1304).This has central implications for victims of revenge porn, who have their lives disrupted not only by the intentional distribution of intimate material by perpetrators but also by the unlawful consumption of that material online.Revenge porn victims are faced with judgment by their surroundings but also by online users who exert further symbolic violence on them.Analyzing how revenge porn is treated and discussed on social media (i.e.virtual realities) is of instrumental importance when considering the real consequences of discourses of VAW in the material world in terms of the potential scalability of violence (Blake et al. 2021;Koller 2011).
Against this background, this paper aims to contribute to the literature on VAW by focusing specifically on revenge porn, as an under-examined crime.Moreover, while a vast amount of research has focused on Twitter for investigating communication on social media, this study draws on the context of YouTube, a less explored platform.The following section describes YouTube and its affordances and explains why it is a suitable platform for the examination of polarization and conflict.

YouTube
YouTube is an online networking platform owned by Google, which is principally aimed at video sharing, hosting a wide variety of videos from all over the world.YouTube, like other social media platforms, feeds from users' uploads to expand its content.However, YouTube's impact is not only linked to its number of logged users; its impact measure is maximized in terms of views, as YouTube's audience does not need to be part of the social network to be able to watch YouTube videos or read their comments.
Studies on communication on YouTube have shown that the type of interaction that occurs in the platform is "multi-authored/multi-recipient", which creates massive discursive polylogues (Bou-Franch and Garcés-Conejos Blitvich 2016, 65) where complex communication is rendered possible through video sharing and textual participation of users who, despite being strangers, engage in discussions about all sorts of topics (Bou-Franch and Garcés-Conejos Blitvich 2014).YouTube users can participate in various modes: subscribing to online communities (i.e., channels), liking or disliking content through thumbs-up and thumbs-down icons and interacting with others in the comments.In this context, linguistic propositions are continuously recontextualized and reshaped, conferring social media users with great power in negotiating social and discursive practices.
As explained earlier, social media characteristics such as anonymity, reduced social and contextual cues and asynchronous interactions may establish the threshold for conflict and linguistic violence.YouTube is by default multimodal, meaning that it is slightly more enriched contextually than other platforms.Here users post comments in response to i) the video and/or its content, the trigger for social engagement, ii) other users' contributions, which are persistent for as long as the video remains posted and its producer does not delete it, and iii) the wider audience that accesses the polylogue (Bou-Franch andGarcés-Conejos Blitvich 2014, 2016).
Despite being contextually richer, the platform is not exempt from conflict.Bou-Franch andGarcés-Conejos Blitvich (2014, 2016) explain that YouTube is a 'deindividuated' environment, in which polarization may arise as a result of the accentuation of people's collective identities over their individual ones, attributing relevance to the in-group's beliefs and social practices and moving toward "a more extreme position in the direction already favored by the group" (Lee 2007, 385).
Regarding revenge porn, YouTube holds a large number of videos that share testimonies and advocate empathy toward victims in the form of interviews, selfnarratives, or journalistic reportages.These videos, although characterized by a generally positive attitude and aim, can be perceived as offenses to which users might feel the need to react, creating thus the basis for polarization.As such, polarization unfolds through the positive presentation of the in-group and the negative presentation of the qualities, actions and participants of the out-group (van Dijk 1998).Besides ideological divergence, polarization is grounded in axiological values; as argued by Filardo-Llamas and Morales-López (2021), polarization involves the encoding of emotion and the (re)production of evaluations of reality.Therefore, the analysis of polarization proposed here includes the examination of the representation and evaluation of in-groups and out-groups.

Data
The data stem from four videos uploaded to YouTube from 2017 to 2023 by officially verified news and informative media channels.The videos selected 'Stupid Girls' vs. 'Real Men': Identity Construction and Social Media Polarization... Issue 41, 2024, pages 129-149 featured real crimes of revenge porn and included the participation of former victims that spoke about their cases and experiences.The videos developed in a narrative and interview-like manner, in which background information was provided by the narrator, who also interviewed the victims.All the victims in these videos were women who belonged to different cultural backgrounds: two videos included victims from Western cultures, while the other two broadcasted the phenomenon in Eastern countries.Although the cultural context in which this sort of crime takes place is important for the understanding of the social consequences of revenge porn, the fact that revenge porn is nowadays allpervasive worldwide justifies the choice of a culturally varied sample.This broad representation of revenge porn in different cultural areas is deemed relevant in enhancing the richness of the sample and results.Additionally, YouTube is a global social network platform and can be accessed by users from multiple cultural backgrounds and geographical locations.Since the focus of this paper is on YouTube users' comments and not the videos themselves, the heterogeneity of the videos selected should not pose a problem.

Alicante Journal of English Studies,
After selecting the videos, all the comments posted to them were extracted.This was achieved with the web scraping tool YouTube Data Tools.Output included the text of the comment along with metadata comprising unique comment identifiers, commenter username, publication date, like count and reply count.Both top-level and nested comments were obtained to observe comments' potential dialogicality, that is, to determine how disagreements and polarization unfolded in users' interaction.A total of 14,802 comments were collected during November 2023 and January 2024, averaging 3,700 comments per video.Subsequent data processing discarded comments partially or completely in languages other than English as well as comments with special or unreadable characters.This procedure left 13,934 comments for the analysis, which amounted to approximately 411,000 words.
Regarding ethical considerations, the paper followed the 2019 guidelines of Association of Internet Researchers (AoIR), which approves of the use of social media comments provided they belong to the public domain, that is, they are freely accessible.In this regard, YouTube's 2023 terms of service state the following: By providing Content to the Service, you grant to YouTube a worldwide, nonexclusive, royalty-free, sublicensable and transferable license to use that Content […].You also grant each user of the Service a worldwide, non-exclusive, royaltyfree license to access your Content through the Service and to use such Content (including to reproduce, distribute, modify, transform, display, communicate to the public and represent you).
As such, YouTube users are informed of YouTube's policy on data transfer and content dissemination by other users of the service worldwide and agree to it when creating a YouTube account.
Additionally, the AoIR recommends the anonymization of names, usernames, images, localization and any other personal data, which has been considered in the presentation of results.

Methodological Approach
This paper follows a Corpus-Assisted Discourse Studies (CADS) approach (Gillings, Mautner, and Baker 2023;Partington, Duguid, and Taylor 2013) informed by (Feminist) Critical Discourse Studies (CDS) (Fairclough 2013;Lazar 2007;van Dijk 2015).Given the sexist nature of the crime hereby addressed, a critical reading of the results drawn from the analysis with corpus query tools contributes to the understanding of how ideological underpinnings operate as concerns revenge porn.
The corpus-assisted analysis was carried out with the query tool Sketch Engine (Kalgariff et al. 2014).First, the analysis retrieved data that described the most recurrent themes in the corpus.To this end, a frequency analysis was performed.Next, a keyness analysis was run, as it retrieves the unique lexical patterns (i.e., keywords and multi-word terms) that characterize the corpus.The Corpus of the English Web (enTenTen 2021) was employed as the reference corpus to collate the focus corpus with data of linguistic similarity.The minimum frequency was established at 50.Following up, a thorough analysis of the top 15 keywords and N-grams in context was carried out (KWIC analysis).This only included keywords or N-grams with a stable dispersion ratio, ascertained by the Average Reduced Frequency value calculated by Sketch Engine.Additionally, due to the gendered nature of revenge porn crimes and the saliency of these terms in the corpus (see section 4.1.),the adjective collocates of the lemmas girl, woman, man, guy as modifiers or predicates were explored using the 'Word Sketch' function to scrutinize the representation of victims and perpetrators, establishing the collocate range at 5 left and right.These tools were employed because lexical patterns and collocations can yield relevant results regarding writers' attitudes and ideologies (Palomino-Manjón 2022, 12).
The semi-automatic analysis was followed by a close qualitative analysis of keywords in context and collocational patterns to provide critical explanations of the linguistic patterns yielded.Focusing specifically on adjective collocations, the analysis of attitude and evaluation draws on Martin and White's (2005) Appraisal Theory model to examine how victims and perpetrators are appraised and how those appraisals contribute to changing overall axiological values regarding revenge porn.In line with this, the analysis of polarization was approached from a bottom-up perspective, following van Dijk's (1998, 267) notion of the 'ideological square' to observe how conflicting ideologies and groups are constructed in connection with the 'axiological square', since "polarization […] may be not only ideological […] but also affective […], related to feelings and evaluations" (Filardo-Llamas and Morales-López 2021, 3).

The Discursive Construction of Revenge Porn
As a point of departure, the Wordlist function was used to obtain a general overview of the themes most frequently addressed in the corpus.It was generated from lemmas instead of words, encompassing different derivations of the same entity in one form.For the analysis, only lexical lemmas were considered, which were later classified according to Part of Speech (PoS).Table 1 presents the most relevant content lemmas within the top 50 items of the frequency list.The frequency list in Table 1 displays only two lexical categories in its top 50, namely, nouns and verbs.Regarding nouns, YouTube users engage in discussions about (gendered) social actors 2 .As shall be discussed, commenters seem very concerned with actors and processes, that is, who does what and who is affected by what.At this stage, however, the Wordlist only reveals that female social actors are the most talked about (girl, woman) followed by "guy", which is a generic and less gender-specific term to refer to (usually unspecific) male social actors.This pattern seems to apply to verbs too, as those appearing in the top 50 are overall general.Commenters seem to be preoccupied with ascertaining attributions and qualities with the verb "to be" or with what social actors do or communicate ("do" & "say").
After contrasting these results with those from the N-grams analysis, interesting patterns emerge.The most frequent N-grams laid by Sketch Engine include pronouns plus negated auxiliaries such as "I don't", "you don't" or "she didn't" and grammatically negated verbs such as "don't send", "don't do", "don't know" or "don't understand".Looking closely, these patterns correspond to YouTube users' attempts to make sense of the motivations of victims in sending intimate material as well as relating these to their experience or general social rules, as shown in examples3 1, 2 and 3.
(1) I don't believe this.It seems staged.
(2) I don't understand why the girls have ruined lives (3) Don't send them if you don't want them to be shown tbh The N-gram "in the first place" was very salient ... in the focus corpus, being a multi-word term in the keyness analysis.Sketch Engine uses the 'simple maths' formula to ascertain the keyness score (KS hereafter) of an element in the focus corpus compared to the reference corpus.In this case, it showed that this multiword term was approximately 280 times more frequent than in the reference corpus.This phrase is particularly interesting, as users represent the illegal sharing of intimate material not only as a byproduct of revenge over a breakup, cheating or additional reasons but as the logical consequence of sending sexually loaded content.In this sense, if victims send nudes in the first place, they can only expect to have them distributed by their recipients (see examples 4 and 5).As examples 4 and 5 showcase, the action of illegally distributing intimate content by perpetrators is legitimized by way of the following structure: "activity verb 4 + in the first place" (e.g."send" or "share").Thus, in the first place performs the function of a discourse marker that introduces a logical cause-and-effect relationship between sending sexually explicit material and having it shared.This cause-effect discursive argumentation puts victims as the originators of the conflict, pinning the blame on them.In this way, users seem to subtly naturalize the crime and help obfuscate the role of perpetrators.Other uses of in the first place achieve covert perpetrator legitimation by condemning the crime, but deflecting blame and focusing on victims, as shown in examples 6 and 7: (6) While it wasn't right, you shouldn't of [sic] done it in the first place if you knew it was wrong.( 7) What do these sickos have to get revenge on in the first place??She shouldn't be taking nude pics in the first place.But it's not right either way.
Regarding single terms (i.e., keywords), Sketch Engine yields expected results inasmuch as some keywords revolve around the topic at hand.However, the KWIC analysis explains some of the usage patterns around these words.Table 2 displays the top 15 keywords selected for close analysis: As illustrated by Table 2, the three most frequent keywords that characterize the corpus are "nude", "sexting" and "blackmail".When analyzed in context, the word "nude" shows how the crime of revenge porn is conceived as merely distributing nude pictures.This constitutes an oversimplistic understanding of the crime, by which other types of intimate material (e.g.sexually loaded videos) fall out of the category of revenge porn, as shown in examples 8 and 9.In fact, the only comments which mentioned intimate videos as a revenge weapon were those who shared their story or had heard similar stories of revenge porn involving multimedia content.
(8) I believe it's called "revenge" because it often happens when someone feels the sender has wronged them in some way and leaks their nudes in an effort to get revenge.( 9) Oh so type of stuff is called " revenge porn"...guess u rlly do learn something new everyday  oke everyone that does revenge porn should be punished and educated on social norms and laws and sexuality happens a lot in my school.. guys exposing girls And this is why I'll never send nudes Both examples discuss and negotiate the meaning of revenge porn and associate it with sending nude pictures, but no other type of intimate material is mentioned.When looking at the collocations of revenge porn, this tendency was consistent, with the crime being defined by the sharing of pictures or photos, and, more marginally, videos.Oversimplistic and uninformed conceptions of revenge porn thus obscure the scope and consequences of the crime.
The KWIC analysis also reveals that sexting (KS: 410.6) is generally negatively evaluated.Some users condemned victims for sharing their pictures based on their single civil status; other users treated sex as taboo; a third group expressed no moral problems with sex but condemned sexting practices due to the potential risk of intimate material being leaked5 .Hence, victims are evaluated negatively with adjectives of Judgment of esteem -those which deal with normality - (Martin and White 2005, 52) that reproach them for not acting according to "common sense", as illustrated in examples 10 and 11: (10) you said it was not your first time sexting and sending nudes etc. Well i am not supporting the guy who did this,instead he should be jailed or punished more badly but don't you think sending nudes and semi nudes to anybody is just not a good idea.(11) The worst thing about girls is they thinks that to keep relationship longer they have to be in physical relationship with the boy or sexting this all are bullshit have some self confidence if someone loves u they will never force you or ask something like that.
In example 10, although the user condemns the actions of the perpetrator, by way of a disclaimer, he/she questions the victim's actions for not corresponding to common sense.Similarly, in example 11, the user describes inadequate behavior in relationships.Therefore, users' ideologies around love and sex are concealed through the notion of common sense, which is represented as being intrinsic and inherent to all human beings.This common sense around sex(ting) practices assumes people must not share intimate material because i) it is risky and ii) it is morally wrong.Thus, users' discourse around revenge porn seems to adhere to traditional value systems that view non-traditional sexual practices as deviant.As Fairclough (2001, 64) notes, "the effectiveness of ideology depends to a considerable degree on it being merged with this common-sense background to discourse and other forms of social action."Therefore, representing sexting as morally/attitudinally wrong and not sexting as common-sense behavior deems the reproduction of conservative ideologies around sexual practices very effective.
Finally, the analysis of blackmail (KS: 144.9) shows that users understand revenge porn not only as revenge, that is, a response to an offense for the sake of retribution, but also as blackmail.Blackmail entails that the blackmailer obtains something in return after extorsion, urging the blackmailed participant to comply with the blackmailer's demands.The understanding of revenge porn as blackmail suggests that users are aware of the coercive power exerted by perpetrators on victims, who, apart from seeking retribution, expect some sort of payment from the victim; see examples 12 and 13: (12) [In response to the question "what is revenge porn?"]It is the action of sharing/uploading/spreading intimate pictures/footage of someone for revenge or other malicious reasons (blackmail , ruining one's reputation, harassment), typically done by boys and men to their exes/sexual partners/ women and girls who rejected them etc (13) […] partners share their pictures with their own choice then it's ok but leaking or blackmailing them that they'll leak the pictures or videos out and just because he or she refused to have sex or do anything with their own choice and going against your will is wrong, that's what a real lover don't do In example 12, the user includes blackmail in his/her definition of revenge porn, along with other crimes, expanding thus the characterization of the crime and emphasizing the negative outcomes of the same, whereas 13 explains possible reasons behind perpetrators' motivations.This finding contrasts with the oversimplification found in the usage of nude, which shows that discussion over revenge porn is polarized.

Constructing Victims and Perpetrators: Allocation and Mystification of Blame
Although the previous section has already offered some strokes on the discourse around perpetrators and victims, this section presents the most relevant findings regarding the construction of the two main social actors sketched in the corpus (i.e.men, women) (cf.Table 1).The analysis of concordances revealed gender mainstreaming in the corpus, allocating the roles of victim and perpetrator to women and men respectively.This is not surprising considering that the videos selected only displayed female victims and that revenge porn is a gendered crime.For these reasons, a collocational analysis of the lemmas 6 girl, woman, man, guy has been performed.Due to space constraints, the analysis, performed with the Word Sketch function, solely focuses on adjective collocations as modifiers and predicates of the lemmas under scrutiny.These collocates are presented in Table 3 7 .Table 3.Most frequent adjectives collocating with gendered social actors 6 Boy has been excluded from the collocational analysis for two reasons: it was not a salient term in the frequency list and the analysis of the term did not yield relevant or new results.7 Collocations that were of an indexical nature by which users identified participants in the videos (e.g. the red scarf girl) were discarded.
Table 3 yields informative results about the predicational strategies used to characterize and evaluate social actors.Starting with girl and woman, the adjectives used to describe them are clearly contradictory in terms of semantic polarity.On the one hand, women and girls are positively appraised through positive judgments of social esteem ("strong", "brave") and affectual sympathy ("poor girl/woman) (Martin and White 2005).On the other hand, they are negatively evaluated through negative judgments of social esteem on the basis of their incapacity to perceive the potential risks of sharing intimate material with someone else ("stupid", "dumb") and through judgments of social sanction ("responsible"), whereby victims are held accountable for their actions.Examples 14, 15 and 16 illustrate these phenomena: (14) Nice to see both these ladies growth into strong women (15) Girls cannot understand the reason behind stubbornness of heart and sexual vigour of boys.They take full advantage of boys .Girls are DUMBEST.(16) they make guys do all this.they fool us in love, they cheat us later they dont tell world what is the real story.believe me, girls are equally responsible for such things As these examples attest, the axiological values attached to women are conflictual.Whereas some users praise women victims for having survived this crime, other victims continue to be derogated, disrespected and blamed for the crime -notice capitalization and superlative as means of emphasis in 15: "DUMBEST".
Moving on to guy and man, results indicate the existence of polarization as well.While judgments of women were mostly based on social esteem (i.e., on the normality of their behavior or their (in)capacity to comply with social norms), evaluations of male social actors fall rather within judgments of social sanction, with adjectives invoking truth, morality and ethics or their lack thereof, as expressed by "nice", "good", "bad" or "trustworthy" (Martin and White 2005, 52-53).However, in contrast with the results for female social actors, these evaluative adjectives were not used to ascribe qualities to male social actors but to place them in axiologically meaningful categories.In other words, adjectives evaluating guys and men as good or bad were not appraising perpetrators specifically.Instead, these qualities were attributed to different identitarian subgroups.
A close analysis of collocation patterns shows substantial ideological negotiation regarding male identities.These identities seem to be negotiated through processes of affiliation and disaffiliation in relation to values of morality, veracity and ethics.Examples 17, 18, 19 illustrate these phenomena.
(17) A good man never asks for the nudes ,first of all.(18) The boys who ask and send nudes are bad guys but she could have easily refused na.(19) they're [sic] are many girls who sext but the guys are trustworthy..they don't leak their pics even after a breakup...it all depends on the person you trust As illustrated in these examples, users establish rules of ethical behavior concerning revenge porn in connection with the categories "good/trustworthy men/ guys" and "bad men/guys".According to this categorization, good or trustworthy men/guys do not ask8 for nudes or leak them, while "bad guys" do.In example 21, the use of "person", which is a generic word, seems to signal ideological distance between the category of guys and that of revenge porn perpetrators.
Similarly, the adjective "real", which is the most frequent collocate modifying "man" points to the construction of a subgroup within the overall identity of men.The subgroup of "real men" seems to be endowed with strong morals and respect for women (see examples 22 and 23): (20)  Such men who do this are not real men .Just half ass men.With a ego (21) […] the guys who do these types of act they are not real Men because a real men don't break someone trust to get revenge.
It can be seen that the group of "real men" is implicitly presented positively, whereas the category outliers (i.e.revenge porn perpetrators) are represented and evaluated negatively (van Dijk 1998).Thus, a process of selective dissociation takes place (Bedolla 2003), whereby members of a group disaffiliate from the members and characteristics of that group, differentiating themselves in particular ideological aspects.In this case, example 22 specifies the type of qualities that fall out of the category of real men, namely, being egocentric.Example 23 expands on this by clearly stating that men who break someone's trust to get revenge do not belong to the group of real men.Such specific characterizations of real, good, and trustworthy men indicate that revenge porn perpetrators are excluded from this positively appraised category of men because they carry out unethical actions, which would tarnish real men's social image.'Stupid Girls' vs. 'Real Men': Identity Construction and Social Media Polarization... Issue 41, 2024, pages 129-149 In short, in users' representation of victims and perpetrators, gender roles are automatically ascribed.Thus, two dyadic groups are formed: victims-women and perpetrators-men.However, the analysis of the qualities attributed to each group has revealed that gender identities are complexly developed.Women are constructed as the main outgroup for their noncompliance with traditional views of sex (practices) and their incapacity to act according to 'common sense' upon potential risks.Men are represented as belonging to various differentiated subgroups through processes of axiological (dis)affiliation.These two subgroups -real, good and trustworthy men/guys vs. bad, egocentric and unreliable men/ guys -have their own rules, values and expected ethical behavior.

Conclusion
This paper has examined the discourse of social media users around revenge porn.Adopting a CADS approach combined with CDS, the study has analyzed a corpus of YouTube comments posted to four videos on revenge porn.
The first research question sought to examine the discursive construction of ideologies around the crime of revenge porn and its participants.Overall, results showed that users engage in polarized interaction which is characterized by ideological divergence in the understanding of revenge porn.Although perpetrators are overtly blamed for the act of distributing intimate material online, victims are blamed for producing or sending such content in the first place (Mckinlay and Lavis 2020).Indeed, the N-gram in the first place was key in creating discursive argumentation and cohesion for the reproduction of victim-blaming ideologies.The focus is therefore placed on victims rather than perpetrators.Moreover, findings have revealed that users (re)produce contradictory representations of revenge porn as a crime.While some users oversimplified the crime by reducing it to sending nudes, others expanded the conceptual category of revenge porn by comprising other crimes within it, like blackmail.In this sense, users seem to recognize the coercive power exerted on victims by perpetrators when soliciting and sharing private images/videos.Importantly, users engaged in ideological negotiations of sex and sexuality.In this respect, the linguistic patterns of the keyword sexting revealed that users perpetuate traditional ideologies that conceive of sex as belonging to the offline private sphere.Users linguistically emphasized the potential risks of sending sexually loaded material to another person through arguments of common sense.Women who engage in this behavior are consequently negatively appraised for non-observing common-sensical social norms.
The second research question covered the representations and evaluations of victims and perpetrators, and, by extension, women and men.The analysis of the adjectives collocating with the selected lemmas unveiled that victims/ women tend to be derogated in terms of social-esteem (e.g.stupid, dumb) for not following the rules of common sense when sharing intimate/sexual material, as suggested by the results above.Parallelly, victims are praised for their strength and bravery in facing public judgment and surviving the crime.Findings around guy and man indicate that the adjectival collocations retrieved categorize men into different identitarian subgroups according to their ethical values: good, trustworthy and real men/guys versus bad, egocentric, unreliable men/guys.Perpetrators are excluded from the category of good and real men because their actions and behavior do not match the group's moral standards (Hardaker and McGlashan 2016).In this respect, sharing and leaking intimate material is negatively evaluated and, consequently, so are perpetrators.Conversely, not asking or sharing intimate material is positively evaluated, an attitude which is attributed to those men -real men -upholding ethical values of goodness and trustworthiness.This axiological distance between the two groups contributes to a virtuous representation of men who do not leak intimate content and enhances in-group legitimation.In this sense, not committing a revenge porn crime is not perceived as regular ethical and civil behavior, but as something to be praised for.
Taken together, findings from this study suggest that YouTube users' understandings of revenge porn are complex and conflictual.Axiologically speaking, revenge porn crimes and perpetrators were negatively evaluated, but so were victims, although in rather convert ways.Ideologically, the construction of different identitarian subgroups of masculine identities suggests the legitimation of men who do not incur into intimate content sharing.However, users tended to downplay perpetrators' responsibility by equaling their behavior to that of victims (Hall and Hearn 2019).Perpetrators are not therefore completely delegitimized.Overall, it seems that revenge porn continues to be a controversial issue for the wider public where accountability boundaries are obscured.
Despite these findings, the scope of the paper is limited, for it mainly focuses on keywords and adjective collocations.Subsequent studies are encouraged to conduct a thorough analysis of verbal and nominal patterns.Moreover, although this study works with naturally occurring data from YouTube, it falls short in terms of the variety of social media data.This shortcoming could be offset by factoring in several datasets -videos text and visuals, interviews -allowing for triangulation of results (as done by Potts 2015).In addition, this study examines revenge porn in a general sense, overlooking cultural differences.Further studies that take culturally specific approaches are needed to understand how ideological underpinnings around revenge porn are worked out in particular sociopolitical

( 4 )
Why sent such pictures or videos in the first place?I really thing [sic] that if you guys were actually smart, such incidents would never occur.(5) […] I never get why girls share nudes in the first place.If you yourself couldn't keep your privacy, what makes you think someone else will.

Table 1 .
Top content/lexical lemmas within the first 50 items of the Wordlist

Table 2 .
List of (single-term) keywords examined in context 4 According toBiber et al.'s (1999)semantic categorization of verbs, activity verbs are those which designate an action performed (intentionally) by an agent.