Where program and fantasy meet: Female fans conversing with character bots in Japan

Keiko Nishimura

Sophia University, Tokyo, Japan

[0.1] Abstract—Character bots are automated programs that post—that is, tweet—characters' lines from popular manga, anime, games, and so on. They post regularly, and in the past few years they have become difficult to ignore, especially in fan communities. Many fans take great pleasure in interacting with favorite characters as bots; rhey also enjoy the communities that spring up around favorite series, characters, and, yes, even bots. Here I adopt an ethnographic approach to analyze the human dimensions of the phenomenon of character bots, based on participant observation among female fans in Japan.

[0.2] Keywords—Fan community; Fandom; Fujoshi; Japan; Popular culture; Twitter; Women; Yaoi

Nishimura, Keiko. 2013. "Where Program and Fantasy Meet: Female Fans Conversing with Character Bots in Japan." Transformative Works and Cultures, no. 12. doi:10.3983/twc.2013.0457.

1. Introduction

[1.1] On the official help site of Twitter in Japan, there is a section titled "Parody, Commentary, and Fan Accounts Policy" (Twitter n.d.). Twitter permits users to have parody accounts as long as they are clearly marked as such. Twitter users in Japan who own alternative accounts for their character bots (kyarakutā botto) refer to this statement by Twitter to justify the existence of their creations. Character bots are automated programs that post—that is, tweet—characters' lines from popular manga, anime, games, and so on. They post regularly, and in the past few years they have become difficult to ignore, especially in fan communities.

[1.2] There are many different kinds of bots on Twitter (komski n.d.), and Japanese-specific bots, such as manga quote bots, have been introduced to the English-speaking world (Barron 2010). However, the existence of character bots programmed to respond to keywords and interact with fans as characters is not well documented. Some fans actively converse with bots, knowing full well they are just programs. Conversing with a bot may seem to indicate acute isolation and loneliness (Ayacchin 2010), but many fans take great pleasure in interacting with favorite characters as bots. They also enjoy the communities that spring up around favorite series, characters, and, yes, even bots. Here I adopt an ethnographic approach to analyze the human dimensions of the phenomenon of character bots, based on participant observation among female fans in Japan.

2. Background and methodology

[2.1] Usually character bots emerge in clusters of fans of specific manga, anime, games, and so on. The programmers of the bots are fans, many without specialized knowledge of programming (note 1), and the bot's followers are also fans. Although everyone knows who programmed the bot, in the participatory culture of fans online, it seems to belong to everyone and no one. When followers converse with the bot, the interaction is visible to other fans, which contributes to an active, open, and shared form of bot play (botto asobi). Through the bot, members of the fan community become intimate with the character and with one another. This seems to be the case especially with yaoi fans (Galbraith 2011).

[2.2] Yaoi is similar to slash fiction. It is a genre of fan production that involves pairing established male characters from commercial media. The pairing is usually a romantic one, it can involve sex, and it stereotypically foregrounds the imagined pairing at the expense of developing a story. Indeed, the term yaoi is an acronym for "no climax, no punch line, no meaning" (yama nashi, ochi nashi, imi nashi) and emerged in Japan in the 1970s, along with the growth of anime fandom (Fujimoto 2007). The producers and consumers of yaoi are predominantly girls and women, and they often refer to their taste for fictional male-male romance as kusatteru (rotten). Just as they have ironically retained the word yaoi to refer to their works, some fans self-identify as fujoshi (rotten girls), though others deny this label as a result of comical representations the mainstream media impose on this term.

[2.3] Here I draw on participant observation conducted on Twitter in 2010 and 2011. I followed character bots from a specific manga/anime series, and human users who regularly had conversations with bots. I also participated in conversations with bots and other users. In this sense, I was a participant experiencer (Garcia et al. 2009), contributing to and actively experiencing the interactions I observed. All the women I talked to were aware that I was a researcher, and they granted me permission to use the data I collected. To ensure privacy, all names used here are pseudonyms, and I withhold the name of the manga/anime series they are fans of. This is because it would otherwise be easy to identify these fans, some of whom are high profile, in a cluster that is relatively small and intimate. Character names are abbreviated to single letters.

3. Twitter basics

[3.1] Twitter, a Web service launched in 2006, is on the rise in Japan (note 2). Users are prompted to follow other users, and they receive the tweets of those they are following, which are organized into a feed that is updated in real time. This is called a timeline (TL), and many users use it to converse (figure 1). Conversations are indicated by the @ symbol followed by someone's Twitter ID, meaning the tweet is a response to the person with that ID. Constant exposure to other users' thoughts and the ability to respond to them immediately lead users to imagine being in a sort of community or cluster of shared interests and experiences. Fujoshi might form their own clusters, but they also gather with others who are interested in the same manga, anime, games, and so on. Thus, many users follow different clusters with the same Twitter account. The posts from a particularly chatty user following many people and participating in multiple clusters can dominate one's TL. However, in most cases, the TL is a log of conversations between users and bots.

Twitter screen capture: timeline with profile pictures of individual Twitter users (including bots).

Figure 1. Screen capture of a timeline. The profile pictures of individual Twitter users (including bots) are shown on the left. Users' names are followed by Twitter IDs, and beneath them the message is shown. On the right side, the relative time elapsed from the tweet to the present is shown (e.g., "posted 27 minutes ago"). Underneath each message, the word hiraku (open) appears, indicating a link to the technical details of the tweet. In two of the tweets (top and second from the bottom), there is also the phrase "kaiwa wo hyōji" (display the conversation), which shows the tweet or tweets this one is replying to (note the @ symbol in front of the tweet).

4. Bots as characters in yaoi fandom

[4.1] Conversation with a character bot looks like conversation with a character, but bots are actually engaging in metacommentary about the series or stories in which they appear, replete with references that only fans understand. It is worth noting that conversations with fictional characters are a long-standing tradition in Japan. In the 1990s, some light novels (note 3) contained postscripts where authors would converse with their characters (note 4). Even before the 1990s, it was common for dōjinshi (fanzines) to stage conversations between characters and fan creators (note 5). Indeed, direct contact with favorite characters seems to be a long-standing desire.

[4.2] Fujoshi are not confused about what is real. They are aware of who programmed a specific character bot, just as they are aware of who the original creator (gensakusha) of the character is. As is typical of manga/anime fandom, fujoshi keep their distance from the original creator because fans are involved in secondary or derivative production (niji sōsaku) of the creator's character, which might involve a yaoi scenario that the creator might find offensive. Nonintervention by the creator is taken as a sign of tacit approval for fan activities, including bot play.

[4.3] My informants self-identified as fujoshi, meaning they are fans of yaoi and sometimes fan authors or artists. In yaoi, characters are divided into penetrator (top, seme) and penetrated (bottom, uke) and arranged into a preferred pairing (note 6). Whether a character is a penetrator or penetrated depends on his personality, which fans read differently. Even within the same cluster or the same pairing, depictions of characters vary widely. Fujoshi call attention to the fan creator's role in interpreting the character—for example, they might say, "Yuki's [character name]" (Yuki-san no [character name]). This is the case for character bots as well, which are a form of fan production (note 7). Fujoshi who follow others' bots are attracted to specific versions of the character, though they also read their own versions and fantasies into it.

5. Conversing with character bots

[5.1] When conversing with character bots, fujoshi engage in the fantasy that they are bystanders to the romance unfolding between the two male characters, played by the bots. On Twitter, the fantasy usually starts with a fujoshi addressing two character bots. Below are two conversations between Yuki (a fujoshi), Bot Z, and Bot Y. Among fujoshi, it is well known that the two male characters from this manga/anime series, represented by Bot Y and Bot Z, are supposed to be a couple. I have numbered the tweets by ID to clarify who is responding to whom.

[5.2] Yuki 1: @BotZ @BotY Damn. Get married, you two (Kusō. Omae ra kekkon shinasai)

Bot Z 1: @Yuki Sorry I didn't hear you. Stop blowing that trumpet *character name*! (Suman kikitoren katta wa. Kora *character name*! Sono rappa no oto yamei!!) [in reply to Yuki 1]

Yuki 2: @BotZ That was quick! (Haya!) [in reply to Bot Z 1]

Bot Z 2: @Yuki come on…What can I do (Ttaku…dō se cchūnen) [in reply to Yuki 2]

Bot Y 1: @Yuki…Hehe (smiley) (…Ehehe *^-^*) [in reply to Yuki 1; see figure 2]

Yuki 3: @BotY Your partner's shy! (Aikata ga terekakushi shiteru de!) [in reply to Bot Y 1]

Bot Y 2: @Yuki I don't know what to say (Mō wakannai n da, donna hyōjō sureba ii no ka) [in reply to Yuki 3]

Yuki 4: @BotZ I've said it over and over again, (but) marry *character Y*…(*Character Y* to kekkon shiro to arehodo…) [in reply to Bot Z 2]

Bot Z 3: @Yuki You…you wanna taste *specific name of a gun* huh? (Omae…*specific name of a gun* uchikonde hoshii n ka?) [in reply to Yuki 4]

Yuki 5: @BotY You should marry *character Z* (*character Z* to kekkon sureba yoi to omou no) [in reply to Bot Y 2]

Bot Y 3: @Yuki I think…*character Z* wouldn't like that (Tabun…*character Z* ga iyagaru n ja nai kana) [in reply to Yuki 5]

Twitter screen capture: conversation view of a character bot and user.

Figure 2. Conversation view of a character bot and user. Bot Y replies to Yuki (Bot Y 1). The above is the original tweet (Yuki 1) that is being responded to (Bot Y 1). In conversation view, the Twitter ID of those responding is shown in blue. The date of the tweet is shown in gray on the top right. A link to shōsai (details) appears below the original tweet. On the top right of the main tweet, the status of following is shown in a blue box (blue for following, gray for not following). The gray icon next to it is a pull-down menu to manage tweeting, message, lists, and so on. Underneath appear three links: henshin (reply), ritsuīto (retweet), and okini'iri ni tōroku (add to favorites).

[5.3] Yuki's second post, "That was fast," is a comment about the response speed of Bot Z. Character bots are operated by a program on the server, which runs the script every few minutes. Although it depends on the bot, the response can be immediate. Responses by character bots such as "I don't know what to say," "I didn't hear you," "What can I do," or simply giggling are reactions to tweets that don't contain keywords; they are random. However, based on her fantasy, which she projects onto the bots, Yuki interprets an ambiguous line such as "What can I do" as the shyness of a lover at wits' end (note 8).

[5.4] The last responses by Bot Z and Bot Y are to a specific keyword: marry. The first response by Bot Z is a comical expression of anger, which can be interpreted as a result of shyness. However, the target of this anger is ambiguous; it could be Yuki, or her suggestion to marry Bot Y. The keyword marry might have been a proposal from Yuki, and the program cannot be too specific in its response. This ambiguity has the unintended result of allowing fujoshi to interpret the character's intentions as they desire. The technological limitations in fact facilitate fantasy play. As for the second instance, when Bot Y responds to the key phrase "marry character Z," the response is a mild rejection of the possibility. In the original manga/anime, these characters are good friends, buddies, or partners, but Bot Y says that character Z would not want to get married (to him), thus diverting the explicit yaoi fantasy. This might be a way to avoid slandering the character in the public sphere of Twitter, but it also allows fujoshi to imagine that character Y is afraid that character Z, his partner, might reject him, and so he cannot confess his love.

[5.5] This conversation happened publicly (these tweets can be viewed by anyone), which affects not only the content of bot responses but also the interactions between fujoshi. This conversation took place under the scrutiny of other fujoshi in the cluster. It seems that Yuki was aware of this, and the phrasing and tempo of her posts is quite entertaining. In addition, it appeared as if the bots were actually talking to her, increasing the humorous aspects of the conversation. Moreover, this conversation might trigger discussions among other members of the cluster. Some fujoshi might say, "Character Z is better suited to raise children (assuming that they get married and have kids somehow)," or, "I can see character Y cooking for character Z, and character Z being embarrassed by the attention" (note 9). Some might even jump in and respond to Yuki or the bots, thus becoming an active participant in the conversation proper. Bots can act as a catalyst for fantasy or as a medium to expand and share it among fujoshi.

6. Mastering bots (or not)

[6.1] Character bots respond to certain keywords or phrases, and fan programmers provide manuals (toriatsukai setsumeisho) to assist users in making the most out of interactions (figure 3). Some programmers do not update their bots, but others pay close attention to conversations between bots and followers and discipline (chōkyō) their creations to enhance the pleasure of interaction. They observe bot and user action and reaction, then add new keywords and phrases accordingly. Metaphorically, programmers become producers of bots as idols, with fans as followers.

Screen capture of manual page for a character bot (in Japanese).

Figure 3. Manual page for a character bot. Translation: "This is an unofficial bot of [name of the character] in [name of the original author]'s [title of manga]. [Link to the character bot's Twitter page]. The bot tweets lines (from the series) about every hour. It reacts to words that are not listed here, so please try to converse. The reactions and lines of the bot include my interpretation (of the series and character), so let me apologize in advance if you don't like it." The words the bot reacts to are listed, and include ohayo (good morning)/mukuri (getting up), tadaima (I'm home), ittekuru/ittekima (I'm going out), kunka kunka (sniff sniff), issho ni ne/soine (sleep with me/lie down with me), oyasu (good night)/nemasu (to sleep), arigato (thanks), gomen (sorry), tsu + parentheses (slang for giving), suki/aishiteru (like/love), kekkon shite/kekkon shiyō (marry/get married), kisu shite/chū shite (kiss), chucchu (mwah), nagusamete/hagemashite (cheer up).

[6.2] However, no matter how smooth the conversation could be if one used the manual and interacted with the bot as intended, followers seem to relish coming up with new ways of playing. An example of this is changing one's name (note 10). In many cases, character bots reply to followers using their names. If one's name is changed, the bot response can be manipulated to reflect fujoshi desire. For example, Mimi changed her name from "Mimi" to "stop don't touch there" (yamete sonna toko sawaranaide) so that she could get the bot to say this every time it responded to her. Instead of "Good morning, Mimi," the bot would tweet, "Good morning, stop don't touch there" (Ohayō, yamete son'na toko sawaranaide), which triggers Mimi's fantasy that this character is being sexually harassed by someone (the seme character, or perhaps even her) in the morning. In my experience observing and participating in the cluster, bot programmers actually appreciated this hacking by users, which generated more conversation. This hacking represents playful resistance to the programmer, who is thereby not the only one who controls the bot. Bot play is shot through with competition (that is, who's getting desired responses) and struggle for control, but in the end, it is the character who controls fans and compels them to interact. This gravity of the character is the core of fujoshi community.

7. Conclusion

[7.1] Conversing with a character bot constitutes affective play with a nonhuman program, but at the same time it can trigger the formation of a fan community or strengthen bonds in an existing fan community. Private interactions with a favorite character are extended into a new context or opened into the shared space of the fan community; attachments are affirmed by others watching and even participating in interactions with the bot. Human interactions are in many cases facilitated and mediated by technology. Of the therapeutic potential of robots, Sherry Turkle (2008) writes, "Their ability to inspire a relationship is not based on their intelligence or consciousness but on their ability to push certain 'Darwinian' buttons in people (making eye contact, for example) that cause people to respond as though they were in a relationship." While there is no material basis for something like eye contact, bots also push buttons and evoke emotional responses, giving fans the illusion of actually talking to a favorite character. A character bot is already rooted in a fictional narrative, and it is the preexisting intimacy with the character that makes it possible to form attachments to bots.

[7.2] Today anyone can access Twitter from a variety of devices (PCs, tablets, smart phones). In such a device, fujoshi carry around mobile conversation companions. They can talk to friends, favorite characters in the form of bots, or both. Bots make humanlike responses, are available 24 hours a day, 7 days a week, and offer access to private fantasies embedded in everyday life; they share routines and experiences. Despite knowing about the original creator of the character and bot programmer, in interactions with the semiautonomous bot (the manga/anime character brought to life), fujoshi can subscribe to the myth that "there is no one inside" (Nozawa 2012). This is similar to role-playing, but with one of the participants being an established fictional character, fans playing themselves in relation to the character, and the game open to encounters with other players in the community. Of course, not all character bots are as well maintained as the ones I have observed. In fact, the number and quality of character bots is a fairly reliable indicator of the popularity of a series and the size of its fandom. However, the general proliferation of bots on Twitter speaks to the spread of fan practices in new media in Japan.

8. Notes

1. The owner of the Web site "How to create a Twitter bot for people can't program" (Puroguramingu ga dekinakutemo tsukureru Twitter bot no tsukurikata) ( notes, "Right now Twitter is the place where bots can perform most creatively and vividly, and as we are at a time when bots are so close to us, I felt more people should be able to create bots, and there should be more interesting bots come out" (Genzai intānetto jō de bot ga ichiban omoshiroku, ichiban iki iki to katsudō dekiru no wa Twitter no naka dato omou shi, bot ga kon'na ni mijika ni natta ima, motto iron'na hito ga bot o tsukutte, iron'na omoshiroi bot ga detekite hoshī) (Pha 2009).

2. According to Nielsen Company statistics, the number of visitors to Twitter was 11.7 million in October 2010, 2 million higher than the 9.7 million visitors to Mixi, which at the time was known as the largest social networking site in Japan (Saito 2010).

3. The light novel is a genre of juvenile novels that are character-centric and illustrated with mangalike images.

4. Authors such as Akahori Satoru, Kanzaka Hajime, and Akita Yoshinobu participated in these formal conversations.

5. The taidan (conversation) section was popular until the mid-2000s, when readers started to demand stories rather than insider jokes and communication. This information is based on a personal interview conducted on September 21, 2010.

6. There are a few people who prefer both. This is called reversible (riba), meaning that characters can switch sex roles—that is, be either an uke or a seme.

7. Character bots tweet lines from the original work, mixed in with lines that the programmer created. A character bot's profile picture is an illustration drawn by fans, not the original creator. The character bot both is and is not the same as the original character, placing it squarely in the realm of fan production. Fujoshi are aware of and accept the fact that there are multiple instantiations of the favorite character, whose range expands or is opened up in new media and technological contexts.

8. A fertile imagination and tendency toward fantasy play are characteristic of fujoshi culture (Galbraith 2011). For example, when male otaku (hardcore fans of anime, manga, video games, and so on) from one thread on 2channel, Japan's largest anonymous bulletin board, invaded a fujoshi thread and criticized the girls and women there as "gross" and "perverted," the fujoshi responded by playfully projecting their yaoi fantasies onto the male otaku, who were interpreted as uke who wanted attention from (i.e., to be penetrated by) fujoshi performing as seme.

9. It is quite common for yaoi and boys' love fantasies to involve daily routines such as cooking, shopping, and even raising children. Some have pointed out that even though yaoi deals with same-sex couples, the format of the story is not all that different from the heterosexual romances seen in girls' manga.

10. The name of a user is different from a user's Twitter ID, which starts with the @ symbol. The Twitter ID has to be unique to each individual user, is limited to 15 characters, and permits only Roman letters and numbers, whereas the name can be up to 20 characters, can contain Japanese and Chinese characters, and can be anything. Many users put in their handle names, and some write information of dōjinshi events that they are going to attend (or not). For example: "Yuki (Comic Market Summer, West section, Space number A-25) [Yuki@natsukomi higashi A25]" and "Mimi (couldn't get the space for Comic Market Summer) [Mimi@natsukomi ochita]."

9. Works cited

Ayacchin. 2010. "Ronery Mode with Twitter Anime Character Bots.", October 4.

Barron, Sandra. 2010. "They've Got a Twitter Bot for That." Japan Pulse, November 22.

Fujimoto, Yukari. 2007. "Shōnen Ai, Yaoi, BL: 2007 nen genzai no shiten kara [Shōnen Ai, Yaoi, boys' love: From a contemporary point of view of the year 2007]. Eureka 39 (16): 36–47.

Galbraith, Patrick W. 2011. "Fujoshi: Fantasy Play and Transgressive Intimacy among 'Rotten Girls' in Contemporary Japan." Signs 37 (1): 211–32. doi:10.1086/660182.

Garcia, Angela Cora, Alecea I. Standlee, Jennifer Bechkoff, and Yan Cui. 2009. "Ethnographic Approaches to the Internet and Computer-Mediated Communication." Journal of Contemporary Ethnography 38 (1): 52–84. doi:10.1177/0891241607310839.

komski. n.d. "Bots." Twitter Fan Wiki.

Nozawa, Shunsuke. 2012. "The Gross Face and Virtual Fame: Semiotic Mediation in Japanese Virtual Communication." First Monday 17 (3).

Pha. 2009. "Puroguramingu ga dekinai hito mo min'na Twitter bot o tsukuttara īyo" [Everyone who can't program should create a Twitter bot]. September 16.

Saito, Tōru. 2010. "Mixi, Twitter, Facebook 2010 nen 10 gatsu saishin nīrusen chōsa—Twitter wa kenchō, Facebook wa kyūshin" [Mixi, Twitter, Facebook, 2010 October latest Nielsen survey—Twitter grows steadily, Facebook rises rapidly]. In the Loop: ITmedia Alternative Blog, November 22.

Turkle, Sherry. 2008. "Always-On/Always-On-You: The Tethered Self." In Handbook of Mobile Communication Studies, edited by James E. Katz, 121–37. Cambridge, MA: MIT Press.

Twitter. n.d. "Parody, Commentary, and Fan Accounts Policy." Twitter Help Center.

Transformative Works and Cultures (TWC), ISSN 1941-2258, is an online-only Gold Open Access publication of the nonprofit Organization for Transformative Works. TWC is a member of DOAJ. Contact the Editor with questions.