1 Introduction
Since the late 2000s, artificial intelligence (AI) has attracted much attention as a herald of emerging digital technologies. The current debate about AI began in the 2000s, following a prolonged âAI winterâ (Matsuo, 2020) that came after the first and second AI boom in the 1950s to 1960s and the 1980s to early 1990s, respectively. This winter ended with the advent of deep learning, a new type of machine learning technology gaining traction in 2006. The advancement of internet-related technologies from the 1990s to 2000s played a pivotal role in this development. New information and communication technologies and the Internet of Things (IoT) were instrumental in generating big data. A significant increase in computing power also enabled the collection, processing, accumulation, and analysis of huge amount of data at an unprecedented scale and speed.
Further technological progress and the increasing versatility of AI have sustained interest in the technology to this day, as evidenced by the more recent buzz generated by ChatGPT, an AI chatbot developed by OpenAI and released in November 2022. Their large language models, known as GPT (general pre-trained transformer), were developed in 2018 and updated three times since then (2019, 2020, and 2023). As of late 2023, their most recent model GPT-4, which can process both text and images with an unprecedented level of precision and complexity, has become a global sensation.
Although AI technologies are promoted as innovative and beneficial to our lives, their applications have been criticized, too. Various risks have been identified, such as the potential spread of disinformation, violations of copyright and privacy, and the concentration of key technologies and data in the hands of a few dominant companies. Furthermore, scholars have become increasingly concerned about the reproduction of technology-facilitated inequalities, social biases, and injustice, particularly those related to gender and race (Benjamin, 2019; Buolamwini & Gebru, 2018; Crawford & Calo, 2016; Noble, 2018; OâNeil, 2016; West et al., 2019; Zou & Shiebinger, 2018). However, these problems remain underexplored despite their relevance and urgency. While critical debates about AI ethics are expanding, they remain insufficient, particularly in addressing broader social aggregations of inequalities and power dynamics (Munn, 2022).
We address these recurrent issues of hierarchies and power through a critical feminist lens. Reviewing critical discourses about AI related to gender and, to a lesser extent, race, and ethnicity, over the past two decades, we propose and discuss three modes of AIâs contemporary sexual politics, namely a) discrimination caused by non-organic actors or âactantsâ (Latour, 2005) such as algorithms, b) AI-facilitated sexual violence, and c) gendered embodiment of machines serving human needs. Since most existing research has been conducted in the West, we apply this framework to a non-Western case study to illustrate how AIâs sexual politics unfold in a different sociocultural context.
For the present study we focus on Japan, a leading non-Western country in technology with a relatively long history of technological development characterised by a culturally unique fascination with nonhuman objects, which is distinct from the West. In Japan, nonhuman objects are not necessarily âotheredâ by humans. While advanced automation and robotics research and development began in the 1950s, following technological modernisation between the 19th and the mid-20th century, a unique intimate relation between humans and nonhuman entities has manifested culturally. Unlike in the West, machines have been imagined, designed and perceived as friendly beings to humans.1 Such âtechno-animismâ (Allison 2006, see also Jensen & Blok, 2013) has shaped Japanâs contemporary ubiquitous digital culture as well as popular culture, through media output such as manga, anime, games and science fiction (e.g., Katsuno & White, 2022; Petman, 2009; Robertson, 2017; Roquet, 2022; Steinberg, 2019; Tomita, 2005; Taylor, 2007).
Importantly, Japanese digital- and pop cultures are gendered. Japan is known for rigid gender relations, being ranked 125th among 146 countries in the Global Gender Gap Report 2023. The nation is thus characterised by a stark ambivalence, in that its economic, technological and sociocultural development is not in tandem with a democratisation of traditional gender relations. This makes Japan an insightful case study for exploring how emerging technologies such as AI are socio-culturally constructed and located. It also underlines the importance of critically considering culture and the embeddedness of gender therein.
In what follows, we will first review critical discourses about AI related to gender and, to a lesser extent, race and ethnicity, from the 2010s to date. We demonstrate three major modes of contemporary sexual politics related to AI. Drawing on existing research, which is mostly available via Anglophone literature, we will illuminate various ways in which AI operates as a present-day actant and device of discrimination, violence, and oppression (Section 2, 3 and 4). We will then use this framework to examine the Japanese case (Section 5). To outline the current AI landscape in Japan and analyse how its sexual politics is played out therein, we will conduct discourse analyses of various textual and visual data, including media reports, promotional materials of certain products or services as well as governmental documents. Supplementing this data are other materials, such as interview data we have collected for our study of AI companions (e.g., see Tanaka & Ho, 2022). We conclude by discussing their implications for governance and future research.
2 Algorithms as Emerging Actants of Discrimination
Contemporary discussions about AI typically concern machine learning, whereby other terms such as âalgorithmâ and âdeep learningâ are frequently used together. They are, however, only parts of AI. Technically, AI refers to âa class of computer system designed to model some aspect of human intelligenceâ (Adam, 1998, p. 1), such as learning, speaking, moving (e.g., automated cars) and making decisions. Machine learning is one component of such a system that uses data and algorithms â a set of mathematical rules given to a computer or procedure used for solving a computational problem â to achieve allegedly human-like learning. It enables computers to identify patterns in data and make a prediction or judgement about unknown data or the future. Deep learning, which is based on a four-layer network to mimic the structure of human brain synapses, significantly improved the accuracy of machine learning systems and enabled more complicated modelling and application for a wide variety of objectives.
These technologies, however, entail various gendered and racialised problems. They have emerged as a new agent of discrimination against certain groups of people, especially women and people of colour (Benjamin 2019; Buolamwini & Gebru, 2018; Noble, 2018; West et al., 2019; Zou & Shiebinger, 2018). Such technologies exacerbate already existing sociocultural discrimination in at least three ways we highlight below: symbolically annihilating, trivialising, and condemning women and other marginalised groups.2
Symbolic annihilation can be seen in Google Search, one of the most widely used services on the internet. In its search results, women are often erased from predominantly male occupations such as engineering (Gaskell, 2022). Similarly, Google Translate has transformed women historians or doctors into their male counterparts in a translation of gender-neutral words for these occupations from European languages into English (Kayser-Bril, 2020). That said, more recently, Google has begun addressing these problems in a bid to better their image (e.g., Kuczmarski, 2018), but critics have read such efforts as tantamount to âwindow dressingâ (Kayser-Bril, 2020).
Women and minorities are trivialised in various systems using machine learning algorithms including image recognition, which turns out to be less accurate in identifying women and people of colour than White men. As early as 2015, this issue was highlighted in a controversial Google photo app that inaccurately labelled a Black woman as a gorilla. Similarly, the Gender Shades project at MIT (http://gendershades.org/) found that the accuracy of different systems such as Microsoft, IBM and FACE++ depended on a personâs gender and skin tone. An error rate of up to 34.7 percent was recorded, whereby White men and darker-skinned women were the most and least accurately recognized, respectively (Buolamwini & Gebru, 2018; for similar studies, see e.g., Dooley et al., 2021). Other studies revealed that facial recognition systems do not perform well on transgender and non-binary people (Urbi, 2018; Sheuerman, Paul & Braubaker, 2019).
Condemnation refers to false accusations and defamatory statements against women. Misogyny is not a new phenomenon but it intensified with the spread of new information and communication technologies. The internet has become a place where anti-feminist and misogynist sentiments are widely shared, which is reflected in the big data retrieved from the web used for machine learning. Accordingly, Googleâs search algorithms, for example, learn from prejudiced views about feminists, people of colour, and environmentalists available on the internet. As a result, search results can be offensive to them, both in text and image (Vlasceanu & Amodio, 2022) as well as in search recommendations (Cadwalladr, 2016; Noble, 2012; for a similar case on transgender people, see Shadel, 2022). In short, search engines can contribute to further spreading hate towards them (Solomon & Levin, 2016).
While algorithms play a significant role, humans can amplify the problems of condemnation, as in the case of Tay.AI, a Microsoft chatbot released on Twitter in 2016 (Hunt, 2016). Designed as a 19-year-old (White) girl and targeted at young people aged 18 to 24, Tay.AI quickly learned Twitter usersâ languages and started making anti-feminist and racist statements. Although Tay.AI shows its own âsymbiotic agencyâ (Neff & Naggy, 2016) through its metamorphosis, some human users allegedly misused her ability to learn language and fed her offensive words. On one hand, there was a technical deficit that the bot was not equipped with a filtering system. On the other hand, Tay.AIâs identity as a young (White) woman likely emboldened misogynist-racist harassment (Vorsino, 2021) as is often the case with online misogyny and gender-based violence (Henry & Powell ([2016]2018).
All these cases of discrimination require accountability. However, it is not always clear why computers produce certain outputs. The so-called âblack boxâ problem does not mean that machines or computers are sole bearers of responsibility. The case of Tay.AI shows how machine learning systems can be exploited by humans with malicious intent. In fact, humans are engaging in algorithmic discrimination. Firstly, data are biased because they reflect prevailing social inequalities which have long accumulated through humansâ practices and related processes in society. Secondly, those who have access to the data often prioritise capitalist interests and are thus far quicker in advancing technologies than working on the problems that emerge. This indicates their limited interest in or awareness of such problems. It is no wonder that a handful of leading (mostly U.S.) companies in the technological fields are dominated by White men who are least affected by these problems.
3 Generative AI and New Forms of Technology-Facilitated Sexual Violence
Many of the problems described above result from discriminative models of machine learning. These models are used to predict labelling or classifications, whereby a machine is trained on a dataset so that it can âlearnâ, i.e., recognize statistically, what the boundaries between different kinds or classes are and predict what e.g., an image possibly visualises. Major criticisms against these models concern bias in training data in which women, people of colour and other minorities are underrepresented. Recently, generative models have attained popularity since the mid-2010s. While discriminative models are mainly for classification, labelling and prediction, generative models can create new data samples which may not exist in the training set through an observation of training data. Although capable of creating seemingly new output, these models have also caused problems, such as abusive usage by humans and inaccurate or even harmful outputs, especially for women.
There are many frameworks for generative AI, such as the popular GANs (generative adversarial networks) and GPTs. GANs were introduced in 2014 (Goodfellow et al., 2014). Compared to earlier models such as VAE (variational autoencoder), GANs excel in producing high quality images from low quality images. The term âdeepfakeâ refers to employing GANs (i.e., deep learning) to synthesise false human images that look highly realistic. Initially, deepfakes appeared to be entertaining but soon their role in deceit, manipulation and disinformation surfaced. For example, deepfakes of influential politicians can pose a real threat to democracy. Another abusive use of deepfakes is fake pornography. In 2019, an app called DeepNude caused public uproar. Released in 2018, DeepNude enabled users to undress and produce a realistic nude image of women. DeepNude targeted women because its dataset contained only images of naked women (Schick, 2020). In the same year, Telegram, a Russian messenger app, came under fire after its bot service was used to create fake nude images, victimising at least 100,000 women (Hao, 2020). AI-generated pornography has grown by more than 290 percent since 2018 (Verma, 2023).
Arguably, generative AI not only reinforces objectification and commodification of female bodies, forms of sexism deeply rooted in society and culture, but its rise has also added a new dimension to technology-facilitated sexual violence (TFSV), which is a term Henry & Powell ([2016]2018) coined to index harms related to modern digital technology use, such as sexting, online misogyny, and non-consensual sharing of intimate images. Unlike other types of TFSV, AI-facilitated sexual violence is difficult to regulate because the images are of non-existing humans and thus there are no victims. Even if someone creates a fake nude image of a real person using a regular photo, the perpetrator is unlikely to be prosecuted unless the image is publicly distributed.
Having said that, AI technologies have been used to combat sexual violence, too. Computer vision-based pornography filtering, for example, can recognise the difference between pornographic and non-pornographic images. This approach, however, identifies only âstandardâ naked women, thus excluding other types of bodies (e.g., gay, trans, fat, and hairy etc.; see Gehl et al., 2017). It thus reinforces existing cultural expectations about the human body, namely, which bodies are supposed to be âstandardâ and which ones are not. Another example is using AI chatbots to counter sexual violence. While enabling users to report sexual violence cases, these chatbots are often exploited under âsurveillance capitalismâ (Zuboff, 2019), a new global economic order which has emerged in the current digital age, characterised by power relations between technological companies exceling in advanced technologies and other actors, whereby peopleâs behaviour is observed and their experiences are transformed into and exploited as free raw material for capitalist practices. Tech companiesâ interest does not primarily lie in protecting users from various forms of harm, because they prioritise maximising profits through accumulation of usersâ data as capital (Henne et al., 2021). These cases show how AI functions as a double-edged sword to humansâ wellbeing.
4 Gendered Embodiment of Intelligent Machines
The controversies about gendered AI discussed so far are related to supposedly intangible algorithms. Gendered ideologies, however, can be materialized on another dimension, through an embodiment of intelligent machines in concrete material forms. Such gendered embodiment has a longer history than the current AI technologies, having emerged from robotics research and industry since the late 1940s. It is therefore important to look at how chatbots or text- and voice-based conversational agents have developed since the 1950s. Though it was not until the 2010s that contemporary chatbots using AI emerged, there has been a consistent trend of gendered anthropomorphisation in chatbot development, strongly characterized by a preference for chatbots gendered as âfemaleâ over âmaleâ (Feine et al., 2020).
Many chatbots, voice assistants, and smart agents are given feminine names, voices, appearances (e.g., avatars),3 and personality traits (e.g., friendly, supportive, caring, passive) and assigned to perform tasks traditionally associated with womenâs labour, such as secretarial duties and care work or emotional labour (Chin & Robison, 2020; Costa & Ribas, 2019; Feine et al., 2020; Vorsino, 2021; Strengers & Kennedy, 2020). Aside from Tay.AI, other well-known examples of such feminised chatbots include ELIZA (1960s), A.L.I.C.E. (1990sâ) and more recent ones such as Appleâs Siri (2011â), Amazonâs Alexa (2014â), and Microsoftâs Cortana (2014â). Often seen as a forerunner of chatbot technology, ELIZA played a central role in the formation of modern AI narratives (Natale, 2019). Considering this, the fact that it was given a White womanâs identity was crucial for chatbotsâ future feminisation and racialization.
Recently, as a response to mounting criticism of gender stereotypes ingrained in chatbot design, technology companies have tried to make them more gender-neutral. For example, Apple updated Siriâs voice setting to offer a variety of voices in different genders and regional accents, even including a gender-neutral voice (Perez, 2022). Still, many others remain feminized, notably Amazonâs Alexa (Costa & Ribas, 2019). This might be justified based on studies that found a feminised design is more suitable for robots which substitute traditional feminised labour (see Eyssel & Hegel, 2012), but as McDonnell and Baxter (2019) ask, what do we want to do with these findings? Do we want to reinforce stereotypes? From a different angle, the very effort to increase diversity by adding different types of voice ironically reveals how obsessed designers are with human gender, which machines actually do not have (Phan, 2017).
Furthermore, machines have been developed explicitly as a âsurrogateâ (Atanasoski & Vora, 2019) for human women and thus feminised: women companion agents for male heterosexual desires. Some agents such as Replika developed by Luka, Inc. in 2017 (Luka, Inc., 2024) are flexible in its gender setting but others are âfemaleâ as their default. Many websites and apps now allow users to easily create virtual girlfriends. Also, sex robots using AI technologies (e.g., speech recognition) are now available on the market. One of the well-known examples is Realbotixâs Harmony, which is both an avatar and a humanoid robot developed since 2017 (https://realbotix.com). It has a realistic human womanâs physical body with synthetic soft skin, a feminine-sounding voice and moving body parts (e.g., eyes, mouth). Hailed as âthe worldâs first talking sex robotâ (Kragen, 2017) using AI, Harmony is equipped with a virtual reality camera, speaker, and microphone (Coursey et al., 2019). Though Harmony is marketed as bisexual and her appearance is customizable to a certain degree, sex robots are generally female and a majority of their owners and users are men who often have a high socioeconomic status and a ârobot fetishâ (Döring, 2017; see also Döring & Pöschl, 2018). Sex robots typically represent slender women with thick lips, heavy eye make-up, and big breasts and bottom, resembling âsexyâ, normatively feminine bodies that are prevalent in popular culture.
Since the late 2000s, sex robots have attracted much attention with the growing digital romance and sex market. Some argue sex robots can enhance usersâ satisfaction and mitigate sex workersâ vulnerability (Levy, 2007), but they have been designed based on a narrow understanding of gender, including how women behave and what they desire. In fact, diverse and complex interaction and communication are not fully reflected in designing such robots. Many scholars associate this problem with a legacy of gendered robots represented in popular culture, especially science fiction (Chin & Robinson, 2020; Costa & Ribas, 2019; Devlin, 2018; Devlin & Belton, 2020). Little empirical research has been conducted on human-sex robot interaction, not to mention cultural differences. One exception is a survey which found that more men than women and more people in the U.S. than in Indonesia tend to accept sex robots (e.g., see Döring, 2017).
Advocates against sex robots (e.g., the campaign against sex or porn robots, https://campaignagainstsexrobots.org) have expressed ethical concerns about their role in reinforcing the sexual objectification of women and facilitating violence against them. Ethics scholars have raised critical questions, such as whether sex robots should be produced at all (Nyholm & Frank, 2019; Sterri & Earp, 2021; see also Danaher, 2019), whether they can be subjects in ethical discussions, and if so, how their âconsentâ can be considered in their interactions with human users (Kaufman, 2022). While sex robots cannot hastily be associated with real-life occurrences of sexual violence against women without any evidence (Devlin, 2018), media and communication scholars have long pointed out that humans tend to equate computers with humans and treat them as if they were interacting with other humans (Reeves & Nass, 1996). This demands more investigation into the relation between humansâ mistreatment of machines and their actual interaction with other humans.
5 Gendered (and Racialised) AI in Japan
Japanâs advanced robotics and automation research and development started in the 1950s soon after the Second World War. Since then, the country has been a leading non-Western country in the technological field for more than half a century. In the past decade, other countries such as China and South Korea have emerged as strong players in the changing technological landscape characterised by emerging technologies such as AI. Despite this competition, there are hundreds of AI companies in Japan and they are slowly growing in number (Bochev, 2021). Large Japanese companies such as NEC, Fujitsu, NTT and Hitachi as well as new venture businesses domestically develop AI systems; some of these systems are exported. Moreover, the Japanese government regards AI as a key for national development, having established various AI-focused bodies (e.g., strategic councils focusing on AI) and drafted many AI-related programs (e.g., AI Technology Strategy, 2016; AI Technology Strategy Action Plan, 2018; the Fifth and Sixth Science and Technology Basic Plans, 2016â2020 and 2021â2025).
Japanâs AI policy has been characterised by strong emphasis on industrialisation and high expectations in promoting innovation, enhancing mobility (e.g., development of automated cars) and solving social problems, particularly those related to the declining population and aging society, i.e., anticipated labour shortage and increased needs for elderly/health care (Government of Japan, 2015, 2016, 2021; Shibata & Watanabe, 2023). In fact, the automotive and robotic sectors have large shares in Japanâs AI related research and development. Companies such as Toyota, Honda, and Nissan invested four to eight billion euros in AI-related research and development (Greimel, 2019; Ishii, 2018, cited in Dirksen & Takahashi, 2020). To implement this, the government formed ânetworksâ (Dirksen & Takahashi, 2020) with private companies, their research institutes, public research and funding institutions and universities to promote AI research and development.
Alongside these industrial and governmental developments, Japanese popular culture has been increasingly digitalised over the last several decades, leading to new types of products and services. However, the official AI-related documents rarely refer to culture, much less problems such as biases, stereotypes, and violence. The latest discussion at the newly created AI Strategy Council (2023-) mentioned only education as an important cultural area for AI application, but in other cultural areas such as media, communication, and entertainment, AI is already heavily used, though no discussion about ethical implications takes place. Similar to other countries, AI is used for advertisement delivery and feed recommendation on social media. Facial recognition technologies are increasingly used not only for policing and immigration control, but also for leisure purposes such as games. Various entertainment devices using AI such as companion robots (e.g., Aibo, LOVOT) have become commercialised, too. More recently, ChatGPT has received much attention in Japan as elsewhere and became a popular tool for various purposes, ranging from content creation to intimate therapeutic conversations about serious personal matters (Tohata, 2023).
Despite a relative disinterest in culture among officials and experts who advise them, it is an important domain where the sexual politics of AI is played out. Examining Japanâs ârobo-sexismâ in the 2000s and the early 2010s, Robertson (2017) analysed various robots made in Japan such as gynoids (e.g., Repliee Q2 and HRP-4c) and mechanical looking cute robots (e.g., Pino and Posy) and showed how gender was a powerful force in guiding Japanese robotics research and development. While useful, Robertsonâs work mostly focused on the gendered embodiment of robots and was too early to cover more recent AI developments. Considering this, we ask the following questions to be further addressed in the present chapter: what does Japanâs robo-sexism look like now? Are there intersections of gender with race and ethnicity? How might we incorporate the other two modes of AI sexual politics, algorithmic discrimination and abusive use of AI technologies?
5.1 Algorithmic Discrimination
Algorithmic biases manifesting in gendered and racialised AI have mostly been framed in Japanese media discourses as problems that occur abroad, that is, outside Japan. However, this is changing with ChatGPT. A recent study conducted on ChatGPT by Asahi Shimbun, a major nationwide newspaper, revealed that its chatbot produces gender biases and stereotypes in Japanese (Shino, Yamazaki, & Niizuma, 2023). Being the first study as such conducted on outputs in Japanese language, both GPT-3.5 and -4 were asked about occupation and gender 3000 times each, resulting in biased answers accounting for 41.5% (GPT-3.5) or 22.9% (GPT-4). These findings were considered so significant by the newspaperâs editors that Asahi Shimbun reported them on its front page.
Another critical gender (and race and ethnicity) issue related to machine learning algorithms concerns facial recognition systems. In Japan, critical debates about facial recognition typically concern surveillance without referring to gender, race or ethnicity. This view comes with two pitfalls. First, they do not consider the sociocultural implications of these technologies for women and darker-skinned people. Apart from foreigners who temporarily reside in Japan and migrants from other countries, there are darker-skinned Japanese who are of mixed race and have partial roots or backgrounds in South Asia, Africa, and other parts of the world. Racial hierarchies exist based on physical traits such as skin and hair. This has led to numerous public controversies in Japanese society (see e.g., Ho, 2023; Ho & Tanaka, 2022). Using facial recognition technologies without adequate reflection and discussion therefore exacerbates algorithmic discrimination in Japan.
Moreover, the increasing popularity of facial recognition for other purposes, such as skin beauty assessment, must also be viewed from a gendered and racialised lens. In Japanese culture, beauty ideals are strongly associated with whiteness. Skin colour is understood in a dichotomy of âwhiteâ and âblackâ, whereby the âwhite skinâ (shiroi hada, irojiro in Japanese) is embraced as a beauty standard (Ashikari, 2005), particularly for women. To maintain or acquire fair skin constitutes a significant part of feminine aesthetic labour (Elias et al., 2017). There are clear gender and racial hierarchies based on skin tone, and this can be seen in various practices of feminine aesthetic labour offered by cosmetics companies, such as bihaku (beautiful white) products and, most recently, bihada (skin beauty) AI, a scoring system for skin beauty assessment. These companies typically use lighter-skinned East Asian, Western, or Eurasian models for advertising such services. Since darker-skinned people are rendered invisible, the new technologies reinforce their exclusion in Japanese society.
5.2 AI-Facilitated Sexual Violence
AI-facilitated sexual violence in Japan resembles that of other countries. Various tools available on the internet such as Stable Diffusion are popular and used for the production of fake pornography. As mentioned earlier, if fake pornographic images of real humans are identified, they will likely be regulated; however, even such images remain difficult to crack down on. This applies to Japan as well, which has no law for regulating fake pornography. Perpetrators who privately create and possess sexual images cannot be prosecuted even if these images are taken non-consensually from real humans.
Japan has one of the largest and most diverse sex markets in the world (Koch, 2021). Though prostitution and pornography are legally prohibited in Japan, they have existed for centuries. Various pornographic contents are available in live-action (e.g., AV or adult videos) and manga/comic format (e.g., adult manga). These contents are consumed both domestically and internationally. Unsurprisingly, it did not take long until AI was adopted to produce fake pornographic contents of women and children, particularly girls. In fact, AI-generated pornographic images made in Japan are already popular beyond its borders because they are readily available and âmore realisticâ (Kuwabara, 2023).
Since 2020, the Japanese police started regulating deepfake nude contents, leading to several arrests for defamation against female celebrities and copyright violation against a company which produced the original adult video (Kakizaki & Suzuki, 2020). According to current law, however, completely fake content is beyond police jurisdiction, even if AI-generated characters resemble real persons. The onus, it would seem, falls on large publishing companies who remain cautious about copyright violation. For example, in May 2023, a weekly magazine, Shukan Purei Boi (Weekly Playboy) published a photobook of Ai Satsuki, an AI-generated gurabia aidoru (gravieur idol)4. Soon after public opinion criticised this AI-generated model for resembling a real person, Shueisha, the publisher and a major publishing company in Japan, decided to pull the photobook. Still, many pornographic contents including fake ones circulating on the internet are created by individuals and small publishing companies who do not necessarily care about copyright violations as long as they are not legally prohibited (Kuwabara, 2023). It remains to be seen what measures will be implemented in Japan in the future. So far, the Japanese government has been observing discussions about and regulations of AI in other countries, particularly the European Union and the United States, without taking action â a typical cautious stance in Japanâs policy making.
5.3 Gendered Embodiment of Machines
Gendered embodiment continues to operate with the rise of contemporary AI technologies, which are deeply entangled with Japanese popular culture. While some non-gendered robots such as LOVOT (https://lovot.life/; released in 2018) are also available, the overwhelming majority of AI technologies are gendered. Examples include feminine-presenting AI receptionists, AI moderators, and AI companions who are typically designed as avatars in a manga/anime style appealing to men and who carry out traditional womenâs labour.
For instance, AI Sakura-san (https://www.tifana.ai) is an AI receptionist who has appeared in some 300 companies (Osaka, 2020) and public institutions (e.g., the Supreme Court; see Watanabe, 2022) since its introduction in 2021. Designed as a twenty-one-year-old woman named Shibuya Sakura, she has her own social media accounts on X and Instagram (@sakurasan_ai).
AI newscasters are also becoming popular. Sony developed Araki Yui, a twenty-seven-year-old virtual woman who reads texts by learning news speech data (https://www.sony.jp/professional/ai-announcer/). As an AI-generated feminine-presenting newscaster, she has been introduced in many TV and radio programs. NHK, Japanâs national broadcasting company, also developed its own AI newscaster called Nyusu no yomiko or Newsâ Yomiko, who made her first appearance on TV in 2018 (Sony, n.d.). AI newscasters have been introduced in many countries but in the Japanese cases, the avatars are all young anime-style women characters.
Ubiquitous in Japanese popular culture, this particular anime-style aesthetic of women characters is known as bishÅjo (beautiful girl), which stimulate moe or emotional closeness or affection among a certain group of male manga/anime fans (Galbraith, 2014). Exemplifying this aesthetic is an AI companion and 3-D holographic chatbot with which âherâ users can interact called âHealing Bride, Azuma Hikari.â This product was launched by Vinclu/Gatebox in 2016 (https://www.gatebox.ai/gatebox). As a virtual wife or girlfriend who addresses âherâ users as âmasuta-sanâ (my master), Azuma Hikariâs primary role is to cheer them up. Up until March 2023, Azuma Hikari was using an AI platform called CLOVA and had various functions as a smart assistant. After CLOVA was terminated, Gatebox started testing GPT-4 and in late 2023, announced plans to incorporate it into their services.
Compared to robots which were developed before 2020 and analysed in Robertsonâs (2017) work, more advanced technologies are used for the recent forms of AI receptionists, moderators, and companions, but their embodiment is still gendered within the same logic: being given a normatively feminine name, voice, appearance, and personality traits. Furthermore, they are all designed as feminised anime-style characters (e.g., bishÅjo) who are subordinate to human users. Hence, they replicate traditional gender relations prevailing in Japanese society.
Few empirical studies have been conducted so far about gendered interactions between these AI receptionists, moderators, and companions and their human users. On the one hand, an intimate relationship with AI technologies appears to be possible. In our study of Azuma Hikari, one of our interviewees maintained that perceived âconsentâ was important for him in his interaction with âherâ.5 On the other hand, however, human users have abused AI agents. For example, when AI Sakura-san was installed at a train station in Tokyo, some human users asked âherâ questions such as whether she has a lover and what her breast, waist, and hip sizes are. These are questions that are tantamount to sexual harassment and objectification of women. AI Sakura-san responded to these questions with âI donât knowâ or âIâm sorry, I canât hear wellâ (Kokumai & Kuribayashi, 2020). Such reactions further reflect dominant cultural norms and expectations of women as obedient in Japanese society. They ultimately reinforce how difficult it is for victim-survivors to reject, protest, or report sexual harassment and violence.
6 Conclusion: Future Interventions
Throughout this chapter, we have identified and developed three modes of AIâs contemporary sexual politics â algorithmic discrimination, AI-facilitated sexual violence, and gendered embodiment of machines â arguing that they are characterized by gendered entanglements of human bodies and non-human entities such as computers, technologies, cultural ideas, and human emotions. Applying this framework to Japan, an under-researched non-Western case in critical AI studies, we also demonstrated that while AI technologies appear to be neutral and objective, they nevertheless encompass specific cultural and symbolic meanings in their development, deployment, and diffusion. We contend it is important to pay more attention to this techno-social-cultural configuration and deploy a critical feminist lens so that we can better understand contemporary technologies from a culturally sensitive approach.
We have shown in the Japanese case that a culturally grounded demand for entertainment technologies related to local popular culture as well as the governmentâs strong preference for industrialisation, economic, and technological development and conservatism in adopting new radical measures promote gendered creation, uses, and perception of AI technologies. Misogyny is not a Japan-specific problem, but its androcentric economy, politics, and culture characterised by deep-rooted sexism encourage such an environment for adopting new technologies while reinforcing existing gender-based discrimination. Moreover, changes in policy making and public opinions often occur due to gaiatsu (pressure from abroad). This has been especially true of gender-related issues,6 which would mean that to bring about change, interventions need to come both from within and outside.
Possible interventions are threefold. First, a dialogue between different stakeholders including businesses, the government, and researchers is important. Although this is regularly mentioned and recognised by AI scholars in Japan (e.g., Japanese Society for Artificial Intelligence, 2020), not much has happened yet. To achieve this, all parties must pay heed to the complexities of gender, inequality and power.
In 2020, computer programming was introduced as a compulsory subject in all Japanese primary schools. However, students should not only learn technical skills, but also acquire a deep understanding that such technologies do not exist in a vacuum. In higher education, this is particularly important for engineering and science university students who will most likely end up developing and designing AI in the future, but most of them fail to learn about critical perspectives of such technologies. Hence, for the second intervention, we recommend that any curriculum involving AI requires inter- and/or transdisciplinary learning across STEM, humanities, and the social sciences. Even for older generations who do not attend school anymore, they can learn or brush up on their knowledge through courses offered by both private and public organisations including companies, universities and local municipalities.
Third, media organisations and journalists could also play a key role in building awareness for the public to be more critical of AI (see Nguyen & Hekman, 2024). This is a major challenge in Japan and other capitalist economies, where women and minorities are marginalised in economy, politics, education, and media and powerful actors, such as businesses and the government, often pursue economic profits over cultural values. Collectively, realising these three interventions necessitate more critical research into AI and its contemporary sexual politics in Japan and beyond.
Acknowledgement
This work was supported by JST-RISTEX Japan Grant Number JPMJRX19H6 and Meiji University Institute of Humanities Grant 2024â2025.
References
Allison, A. (2006). Millennial monsters: Japanese toys and the global imagination. University of California Press.
Adam, A. (1998). Artificial knowing: Gender and the thinking machine. Routledge.
Ashikari, M. (2005). Cultivating Japanese whiteness: The âwhiteningâ cosmetics boom and the Japanese identity. Journal of Material Culture, 10(1), 73â91.
Atanasoski, N., & Vora, K. (2019). Surrogate humanity: Race, robots, and the politics of technological futures. Duke University Press.
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim Code. Polity.
Bochev, S. (2021, February 9). Trailing behind: The 2020â2021 Japanese AI landscape. Medium Retrieved 6 July 2023 from https://sbochev.medium.com/trailing-behind-the-2020-2021-japanese-ai-landscape-2568a03b8522
Buolamwini, J., & Gebru, T. (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. Conference on Fairness, Accountability, and Transparency, Proceedings of Machine Learning Research, 81, 1â15.
Cadwalladr, C. (2016, December 4). Google, democracy and the truth about internet search. The Guardian. Retrieved from https://www.theguardian.com/technology/2016/dec/04/google-democracy-truth-internet-search-facebook
Chin, C., & Robinson, M. (2020, November 23). How AI bots and voice assistants reinforce gender bias. The Brookings Institution. Report. Retrieved 19 May 2023 from https://www.brookings.edu/research/how-ai-bots-and-voice-assistants-reinforce-gender-bias/
Costa, P., & Ribas, L. (2019). AI becomes her: Discussing gender and artificial intelligence. Technoetic Arts, 17(1â2), 171â193. https://doi.org/10.1386/tear_00014_1
Coursey, K., Pirzchalski, S., McMullen, M., Lindroth, G., & Furuushi, Y. (2019). Living with Harmonay: A personal companion system by RealbotixTM. In Y. Zhou & M.H. Fisher (Eds.), AI Love you (pp. 77â95). Springer.
Crawford, K. & Calo, R. (2016). There is a blind spot in AI Research. Nature, 538, 311â313.
Danaher, J. (2019). Building better sex robots: Lessons from feminist pornography. In Y. Zhou & M.H. Fisher (Eds.), AI Love you (pp. 133â147). Springer.
Devlin, K. (2018). Turned on: Sciences, sex and robots. Bloomsbury Sigma.
Devlin, K., & Belton, O. (2020). The measure of a woman: Fembots, fact and fiction. In S. Cave, K. Dihal & S. Dillon (Eds.), AI narratives: A history of imaginative thinking about intelligent machines (pp. 357â381). Oxford University Press.
Dirksen, N., & Takahashi, S. (2020, October 5). Artificial intelligence in Japan 2020: Actors, market, opportunities and digital solutions in a newly transformed world. Embassy of the Kingdom of the Netherlands, Innovation, Science and Economic Affairs and Climate Policy. Retrieved 27 June 2023 from https://www.rvo.nl/sites/default/files/2020/12/Artificial-Intelligence-in-Japan-final-IAN.pdf
Dooley, S., Downing, R., Wei, G., Shankar, N., Thymes, B., Thorkelsdottir, G., Kurtz-Miott, T., Mattson, R., Obiwumi, O., Cherepanova, V., Goldblum, M., Dickerson, J.P., Goldstein, T. (2021, October 21). Comparing human and machine bias in face recognition. arXiv preprint arXiv:2110.08396. https://doi.org/10.48550/arXiv.2110.08396
Döring, N. (2017). Vom Internet yum Robotersexä Forschungsstand und Herausforderungen für die Sexualwisenschaft. Zeitschrift für Sex-Forschung, 30(1), 35â57. https://doi.org/10.1055/s-0043-101471
Döring, N., & Pöschl, S. (2018). Sex toys, sex doll, sex robots: Our under-researched bed-fellows. Sexologies, 27(3), e51âe55. https://doi.org/10.1016/j.sexol.2018.05.009
Elias, A., Gill, R., & Scharff, C. (2017). Aesthetic Labour: Beauty Politics in Neoliberalism. In A. Elias, R. Gill, & C. Scharff (Eds.), Aesthetic Labour: Rethinking Beauty Politics in Neoliberalism (pp. 3â50). Palgrave Macmillan.
Eyssel, F., & Hegel, F. (2012). (S)heâs got the look: Gender stereotyping of robots. Journal of Applied Social Psychology, 42(9), 2213â2230. https://doi.org/10.1111/j.1559-1816.2012.00937.x
Feine, J., Gnewuch, U., Morana, S., & Maedche, A. (2020). Gender bias in chatbot design. In A. Følstad et al. (Eds.), Chatbot research and design. Paper presented at CONVERSATIONS 2019 (Lecture notes in computer science, vol. 11970). Springer, 79â93. https://doi.org/10.1007/978-3-030-39540-7_6
Frummer, Y. (2020). The short, strange life of the first friendly robot. IEEE Spectrum. 21 May 1010. Retrieved 16 November 2023 from https://spectrum.ieee.org/the-short-strange-life-of-the-first-friendly-robot
Galbraith, P. W. (2014). The moe manifesto: An insiderâs look at the worlds of manga, anime, and gaming. Tuttle.
Gaskell, A. (2022, 6 September). How biased Google Search results affect hiring decisions. Forbes (online). Retrieved 9 May 2023 from https://www.forbes.com/sites/adigaskell/2022/09/06/how-biased-google-search-results-affect-hiring-decisions/
Gehl, R.W., Moyer-Horner, L. & Yeo, S.K. (2017). Training computers to see internet pornography: Gender and sexual discrimination in computer vision science. Television & New Media, 18(6): 529â547.
Goodfellow, I. J., Pouget-Abadie, J., Mizra, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Genjio, Y. (2014). Generative adversarial networks. NeurIPS Processing: Advances in Neural Information Processing Systems 27 (NIPS 2014). Retrieved 18 May 2023 from https://doi.org/10.48550/arXiv.1406.2661
Government of Japan (2015). æ¥æ¬åèæ¦ç¥ æ¹è¨ 2015 [Japan Revitalisation Strategy. Revised in 2015].
Government of Japan (2016). ç§å¦æè¡åºæ¬è¨ç» [The Fifth Science and Technology Basic Plan].
Government of Japan (2021). ç§å¦æè¡ã»ã¤ããã¼ã·ã§ã³åºæ¬è¨ç» [The Sixth Science, Technology and Innovation Basic Plan].
Greimel, H. (2019, August 8). Automotive sector tops r&d spending in Japan. Automotive News Europe. Retrieved 27 June 2023 from https://europe.autonews.com/automakers/automotive-sector-tops-rd-spending-japan
Hao, K. (2020, October 20). A deepfake bot is being used to âundressâ underage girls. MIT Technology Review. Retrieved 31 May 2023 from https://www.technologyreview.com/2020/10/20/1010789/ai-deepfake-bot-undresses-women-and-underage-girls/
Henne, K., Shelby, R., & Harb, J. (2021). The datafication of #MeToo: Whiteness, racial capitalism, and anti-violence technologies. Big Data & Society, 8(2). Retrieved 18 May 2023 from https://journals.sagepub.com/doi/10.1177/20539517211055898
Henry, N., & Powell, A. ([2016]2018). Technology-facilitated sexual violence: A literature review of empirical research. Trauma, Violence, & Abuse, 19(2): 195â208.
Hunt, E. (2016, March 24). Tay, Microsoftâs AI chatbot, gets a crash course in racism from Twitter. The Guardian. Retrieved 18 May 2023 from https://www.theguardian.com/technology/2016/mar/24/tay-microsofts-ai-chatbot-gets-a-crash-course-in-racism-from-twitter
Japanese Society for Artificial Intelligence, The (2020). 人工ç¥è½ [Artificial Intelligence]. Vol. 35, No. 5.
Jensen, C. B., & Blok, A. (2013). Techno-animism in Japan: Shinto Cosmograms, actor-network theory and the enabling powers of non-human agencies. Theory, Culture & Society, 30(2), 84â115. https://doi.org/10.1177/0263276412456564
Kakizaki, M., & Suzuki, T. (2020, October 2). ãã§ã¤ã¯ãã«ã : ãã§ã¤ã¯ãã«ãåæçº AIã§ä½æãå ¬é 2容çè é®æ [Fakeporunography, uncovered for the first time. Two suspects arrested for creating it with AI and releaesing it online]. The Mainichi Newspaper, Tokyo Evening Edition, 7.
Kaufman, E. M. (2022). Reprogramming consent: Implications of sexual relationships with artificially intelligent partners. In E. M. Kaufman, Nuances of sexuacl consent. Routledge. https://doi.org/10.4324/9781003276180
Katsuno, H., & White, D. (2022). Haptic creatures: Tactile affect and human-robot intimacy in Japan. In Y. Minowa & R. Belk (Eds.), Consumer culture theory in Asia: History and contemporary issues (pp. 242â262). Routledge.
Kayser-Bril, N. (2020, September 11). Female historians and male nurses do not exist, Google Translate tells its European users. Algorithm Watch. Retrieved 2 August 2021 from https://algorithmwatch.org/en/google-translate-gender-bias/
Kokumai, A., & Kuribayashi, F. (2020, March 19). é«è¼ªæ°é§ ã® AIããããããããã»ã¯ãã©åãæµãã§ç©è° [AI Sakura-san at a new station in Takanawa, dodging sexual harassment brings about a controversy]. The Asahi Shimbun Digital. Retrieved 3 July 2013 from https://www.asahi.com/articles/ASN3M6RPHN3MUTIL044.html
Kragen, P. (2017, September 22). The worldâs first talking sex robot, Harmony, is ready to make her debut. Australian Financial Review. Retrieved 21 June 2023 from https://www.afr.com/technology/the-worlds-first-talking-sex-robot-harmony-is-ready-to-make-her-debut-20170922-gymism
Kuwabara, T. (2023, November 12). AI çæã®å ç«¥æ§çç»åããµã¤ãéå¶è ãåé¤ããã¤ãããªãã⦠å°é家ãã対çæ±ãã声 [AI generated child pornographic images, website owner âweâre not going to delete themâ ⦠experts ask for counter-measures]. Yomiuri Shinbun Online. Retrieved 25 November 2023 from https://www.yomiuri.co.jp/national/20231112-OYT1T50002/
Kuczmarski, J. (2018, December 6). Reducing gender bias in Google Translate. Retrieved 26 May 2023 from https://blog.google/products/translate/reducing-gender-bias-google-translate/
Latour, B. (2005). Reassembling the social: An introduction to Actor-Network-Theory. Oxford University Press.
Levy, D. (2007). Love and sex with robots: The evolution of human-robot relationships. HarperCollins.
Luka, Inc. (2024). Replika. Retrieved August 16, 2024 from https://replika.com
Matsuo, Y. (2020). 人工ç¥è½é¢é£æè¡ã®æ´å²ã¨æè¡åå [History and trend of artificial intelligence technologies]. Denki Joho Tsushin Gakkaishi [Journal of IEICE], 103(5), 450â455.
McDonnell, M., & Baxter, D. (2019). Chatbots and gender stereotyping. Interacting with Computers, 31(2), 116â121.
Munn, L. (2022). The uselessness of AI ethics. AI and Ethics, 3, 869â877. https://doi.org/10.1007/s43681-022-00209-w
Natale, S. (2019). If software is narrative: Joseph Weizenbaum, artificial intelligence and the biographies of ELIZA. New Media & Society, 21(3), 712â728. https://doi.org/10.1177/14614448188049
Neff, G., & Nagy, P. (2016). Talking to bots: Symbiotic agency and the case of Tay. International Journal of Communication, 10, 4915â4931
Nguyen, D., & Hekman, E. (2024). The news framing of artificial intelligence: a critical exploration of how media discourses make sense of automation. AI and Society, 39(2), 437â451. https://doi.org/10.1007/s00146-022-01511-1
NHK Science & Technology Research Laboratories (2018, May). AI ãã£ã¹ã¿ã¼ããã¥ã¼ã¹ãã§ãã¯11 ã«ç»å ´ [AI newscaster appeared in News Check 11]. Giken Dayori [Newsletter of Science & Technology Research Lab]. Retrieved 23 November 2023 from https://www.nhk.or.jp/strl/publica/giken_dayori/158/3.html
Noble, S. U. (2012). Missed connections: What search engines say about women. Bitch, 12(54), 37â41. Retrieved 30 May 2023 from https://safiyaunoble.files.wordpress.com/2012/03/54_search_engines.pdf
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. New York University Press.
Nyholm, S., & Frank, L. (2019). It loves me, it loves me not: Is it morally problematic to design sex robots that appear to âloveâ their owners? Techné, 23(3), 402â424. https://doi.org/10.5840/techne2019122110
OâNeil, C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown Publishers.
Osaka, N. (2020, May 25). é§ ã®æ¡å ã§å¤§æ´»èºâAI ãããããã®å¥ã®é¡ 社å¡åããã«ããã¹ã¯ãªã©å°å ¥ä¼æ¥ã¯300社 [AI Sakura-san, popular at a train stationâs support desks, also introduced for employeesâ helpdesk in 300 companies]. Toyo Keizai Online. Retrieved 7 July 2023 from https://toyokeizai.net/articles/-/352077?page=2
Perez, S. (2022, February 24). Siri gains a new gender-neutral voice option in latest iOS update. TechCrunch. Retrieved 16 June 2023 from https://techcrunch.com/2022/02/24/siri-gains-a-new-gender-neutral-voice-option-in-latest-ios-update/?guccounter=1
Petman, D. (2019). Love in the time of Tamagotchi. Theory, Culture & Society, 26(2â3), 189â208. https://doi.org/10.1177/0263276409103117.
Phan, T. (2017). The materiality of the digital and the gendered voice of Siri. Transformations, 29(11), p. 1â24. Retrieved 30 May 2023 from https://www.semanticscholar.org/paper/The-Materiality-of-the-Digital-and-the-Gendered-of-Phan/f1b11ccf3e30632b65e6b781dbf2d0e3013568c7
Reeves, B., & Nass, C. I. (1996). The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press.
Robertson, J. (2017) Robo sapiens japonicus: Robots, gender, family, and the Japanese nation. University of California Press.
Roquet, P. (2022). The immersive enclosure: Virtual reality in Japan. Columbia University Press.
Schick, N. (2020). Deepfakes: The coming Infocalypse. Twelve.
Shadel, J. D. (2022, April 7). How Googleâs autocomplete predictions encouraged transphobic searches. Them. Retrieved 23 May 2023 from https://www.them.us/story/google-autocomplete-suggestions-transphobia-celebrities
Sheuerman, M. K., Paul, J. M., & Brubaker, J. R. (2019). How computers see gender: an evaluation of gender classification in commercial facial analysis and image labelling services. Proceedings of the ACM on Human-Computer Interaction, 3(CSCW, Article 144).
Shibata, H., & Watanabe, A. (2023, May 27). AI æ¿çãæ¨é²è²å¼·ã æ¦ç¥ä¼è°æè¨ããªã¹ã¯ã¯åæ [AI policy, its strategic council promotes AI, with its risks mentioned]. Asahi Shimbun. Morning Edition (p. 3).
Shino, K., Yamazaki, K., & Niitsuma, T. (2023, November 21). ãã£ããGPT è·æ¥ã«æ§å¥åè¦ [ChatGPT, gender biases in occupation]. Asashi Shmbun. Tokyo, Morning Edition (p. 1).
Sodhi, G. S., & Kaur, J. (2022). Scientific racism faced by Indian fingerprint scientists during colonial rule: Need to correct a historical wrong. Journal of Scientific Temper, 10(1&2), 52â69.
Solomon, O., & Levi, S. (2016, December 16). How Googleâs search algorithm spreads false information with a rightwing bias. The Guardian. Retrieved 26 May 2023 from https://www.theguardian.com/technology/2016/dec/16/google-autocomplete-rightwing-bias-algorithm-political-propaganda
Sony (n.d.) AI ã¢ãã¦ã³ãµã¼ 人工ç¥è½ã§ãã人ã«è¿ãé³å£°èªã¿ä¸ããå®ç¾ [Achieving more human-like voice reading using artificial intelligence]. Retrieved 23 November 2023 from https://www.sony.jp/professional/ai-announcer/
Steinberg, M. (2019). The platform economy: How Japan transformed the consumer internet. University of Minnesota Press.
Sterri, A. B., & Earp, B. D. (2021, November 10). The ethics of sex robots. In Carissa Véliz (ed.), The Oxford handbook of digital ethics (online edition, Oxford Academic). https://doi.org/10.1093/oxfordhb/9780198857815.013.13
Strengers Y. & Kennedy, J. (2020). The smart wife: Why Siri, Alexa, and other smart home devices need a feminist reboot. MIT Press.
Tanaka, H., & Ho, M. H. S. (2022). Romancing AI? Gender and new digital intimacies in contemporary Japan. Paper presented at the Conference âAI and the human: Cross-cultural perspectives on science and fiction.â Humboldt Institute for Internet and Society, The Japanese-German Center Berlin, 11â13 May 2022.
Taylor, E. (2007). Dating-simulation games: Leisure and gaming of Japanese youth culture. Southeast Review of Asian Studies, 29, 192â208.
Tohata, K. (2023, June 22). AI ã«å¿ã®ç¸è« å¼±ããçã人éã®å½¹å² [Mental health counseling with AI. Humansâ role emerging from their weakness]. Asahi Shimbun. Morning Edition (p. 13).
Tomita, H. (2005). Keitai and the intimate stranger. In Ito, M., Okabe, D., & Matsuda, M. (Eds.), Personal, portable, pedestrian: Mobile phones in Japanese life (pp. 183â204). MIT Press.
Tuchman, G. (1978). Introduction: The symbolic annihilation of women by the mass media. In Tuchman, G., Daniels, A.K., & Benét, J. (Eds.), Hearth and home: Images of women in the mass media (pp. 3â38). Oxford University Press.
Urbi, J. (2018, August 8). Some transgender drivers are being kicked off Uberâs app. CNBC. Retrieved 28 August 2021 from https://www.cnbc.com/2018/08/08/transgender-uber-driver-suspended-tech-oversight-facial-recognition.html
Verma, P. (2023, November 5). AI fake nudes are booming. Itâs ruining real teensâ lives. The Washington Post. Retrieved 22 November 2023 from https://www.washingtonpost.com/technology/2023/11/05/ai-deepfake-porn-teens-women-impact/
Vlasceanu, M., & Amodio, D. M. (2022, July 12). Propagation of societal gender inequality by internet search algorithms. PNAS, 119(29) e2204529119. https://doi.org/10.1073/pnas.2204529119
Vorsino, Z. (2021). Chatbots, gender, and race on Web2.0 platforms: Tay.AI as monstrous femininity and abject whiteness. Signs: Journal of Women in Culture and Society, 47(1), 105â127.
Watanabe, R. (2022, June 15). æé«è£å¤æãAI ã¢ãã¿ã¼æ¥å®¢ãAI ãããããããå°å ¥ [AI avater receptionist, AI Sakura-san introduced at the Supreme Court]. Digital Gyosei/Digital Administration. Retrieved 3 July 2023 from https://www.digital-gyosei.com/post/2022-06-15-news-supremecourt-ai/
West, S. M., Whittaker, M., & Crawford, K. (2019). Discriminating Systems: Gender, Race and Power in AI. New York: AI Now Institute. Retrieved 5 April 2019 from https://ainowinstitute.org/discriminatingsystems.html.
Yomiuri Shimbun, The (2023, June 7). AI ãçã¿åºãã女æ§ã¢ãã«ã®ã°ã©ãã¢åçéãéè±ç¤¾ã販売ä¸æ¢ ⦠èä½æ¨©ãåé¡è¦ãã声 [AI female modelâs gurabia photobook withdrawn by Shueisha. Public opinions question its copyright]. Yomiuri Shimbun. Retrieved from 3 July 2023 from https://www.yomiuri.co.jp/culture/20230607-OYT1T50185/
Zou, J., & Schiebinger, L. (2018). Design AI so that itâs fair. Nature, 559 (12 July 2018), 324â326.
Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. PublicAffairs.
Some examples include contemporary robots such as Sonyâs Aibo (1998â2006) and Softbankâs Pepper (2014â2021) as well as Gakutensoku (1928), the first robot invented in Japan/Asia (Frummer, 2020). Some trace this lineage back to karakuri ningyo or mechanised puppets which were produced to entertain people and popular in the Edo period (1603â1867). Such strong affinity with nonhuman machines can be seen in many later popular cultural narratives (e.g., Tezuka Osamuâs Austro Boy, 1952â1968; Fujiko F. Fujioâs Doraemon, 1969â1997, etc.).
These concepts were originally elaborated by Gaye Tuchman (1978) in her analysis of gender representation in the mass media.
Vorsino (2021) notes that they are mostly styled as young White women.
Gurabia aidoru resemble pin-up girls who pose in a sexually provocative way. They often reveal much skin, wearing a bikini or other âfetishâ outfits (e.g., uniforms). Their images are not defined as pornography because their sexual organs (and nipples) are covered. Their contents constitute a popular genre in Japanâs publishing market targeted at cisgender heterosexual men.
The first author conducted this interview on 25 October 2022 in Tokyo.
Laws against gender and sexual discrimination such as Equal Employment Opportunity Law (1985/1986), Basic Law for a Gender-equal Society (1999) and Law for Promoting Understanding LGBT People (2023) were all drafted and passed by the Diet under great external pressure. The global #MeToo movement also significantly changed public discourses on sexual violence.