Gender Representations on the Chinese Internet

(1)

LEIDEN UNIVERSITY FACULTY OF HUMANITIES

MASTER THESIS

Gender Representations

on the Chinese Internet

a search engine analysis

Pieter Velghe s1578472

Chinese Studies 2015-2016

A lot of gratitude goes out to my supervisor prof. Florian Schneider for all his help and enthusiasm during the writing of this thesis. And of course to Cosima Nimphy for all the support.

(2)

List of illustrations

Homepage of ChinaSo...21

First results for Baidu's 'showcased celebrity' and ChinaSo's 'celebrity'...23

Baidu’s first result for the ‘small fresh and clean’-category...25

First results of ChinaSo’s ‘jewelry’-category...27

Baidu's first page of results for ‘nanren’, ‘nüren’, ‘shuaige’ and ‘meinü’...29

Recommendations and auto-complete using the different 'gendered words'...31

ChinaSo's top results for ‘nanren’, ‘nüren’, ‘shuaige’ and ‘meinü’...32

Compilation of ChinaSo's recommendations for the 'gendered terms' in the search bar...33

Google HK's top results for ‘nanren’, ‘nüren’, ‘shuaige’ and ‘meinü’...34

(4)

List of tables

Variables for content analysis with different values……….………...38 Relative frequency of results content analysis……….40

(5)

1 Introduction

The internet market in China is booming. With an explosion of online shopping, upcoming internet giants like Baidu, Alibaba and Tencent, internet celebrities and online gaming, the internet has had a huge effect on Chinese society and economy. Amid this celebration of next-level consumerism, the Chinese people are trying to measure the impact digital technologies have on their lives, and constantly find new ways of improving the technology or putting it to better use while information, gossip and advertisements seem to be forced into everyday life on an unprecedented scale. This creates opportunities for companies, media and governments to broadcast their message even further into the worlds of individual users.

Since the economic reforms, there has been a clear increase of sexualized images in media and advertisement (Andrews and Shen 2002). This tendency is also clearly noticeable online, with for example the easy accessibility of pornographic material (Jacobs 2012). This thesis is an approach to study gender representations on the internet, as these representations are increasingly shaping our image of gender. More specifically search engines will be used for analysis, as these in many instances serve as our portal to the Web. As the study is on representations on gender, the visual aspects related to gender will be analyzed, together with some specific mechanisms of the search engines, such as the auto-complete suggestions.

The aim of this research is to come up with a systematic and valid approach to studying (gender) representations using search engines in China, as studying gender representations using search engines and the internet is a fairly new approach to study representations. I acknowledge that this is just one possible approach and the outcome of this research will decide if further research should continue down this path or try a different approach.

The thesis will consist of four parts. The first part consists of two chapters in which the first chapter deals with the position of gender in China and theory about how gender is being constructed in Chinese society, and the second chapter discusses theory about the digital, search engines and ways of studying digital processes. The second part, the methodology-section, discusses the particular

(6)

search engines on the basis of their standard appearance and available functions (without searching) and on the basis of a set of searches to compare the different representations on gender offered by the search engines.

(7)

2 Women and gender in contemporary China

In this chapter the position of women and gender in contemporary China will be discussed, together with the different forms of representation of gender in contemporary China. The first subchapter deals with the effects of the socio-economic changes after the economic reforms on gender relations in China. In the second subchapter, I discuss the way gender is being constructed in China and I argue for the importance of expanding the study of gender representations to digital media.

2.1 Changes after reforms

Since the ‘80s there has been a continuous development of research on women and women’s

conditions in China (Hershatter and Honig 1998). An increase in publicity and awareness of abuse and sexual violence against women, and an increase in the commercialization of women in pornography and prostitution (Evans 1995; 2000) has led many researchers and feminists to commit themselves to improve the position of women in China (Hershatter et al. 1998).

Yan (2000) argues that consumerism has replaced Communist doctrine to become the paramount cultural ideology of contemporary Chinese society. This replacement with consumption, according to Croll (1995: 151): ‘brings with it life-styles and desire, phenomena that highlight individual differences and sexual difference, in contrast to uniformity and gender sameness’. The Chinese became consumers and at the same time they, and especially the Chinese women, became the sexualized objects of consumption (Hooper 1998).

The position of the women seems precarious in post-reform China (Liu 2007). In popular media images, the Chinese woman is being represented as a ‘glorified housewife whose health and happiness depended on the initiatives of others (their husbands or a business enterprise)’ (Zurndorfer 2016: 7). Rarely does she appear as an independent or autonomous person (Evans 2000). A famous excerpt from a 2010 popular television dating program seems to confirm this role with one candidate bluntly stating that: “she would rather cry in a BMW than laugh on the back of a bike owned by a jobless man” (Song and Hird 2014: 160). A 2010 survey confirmed this view with nearly 50% of

(8)

people, and 7.3% more women than men, agreeing with the statement: ‘to marry well is better than to work well’ (Ding and Li 2014: 11). Compared with the 2000 survey, the female and male approval rates to this statement have increased by 10.7 and 10.5% respectively in 2011 (Research Group 2011).

Although Xu (1990) states that increasing numbers of young women are no longer willing to play the role of passive instrument of their husband’s desire, these are still a minority. The majority of unmarried young women view their bodies as valuable commodities to be exchanged for income and security (Zhang 2010), while in pursuing a kind of ‘global hypermasculinity’ (Ling 1999: 280), the male perfection in today’s image is the confident businessman whose financial success entitles him to attractive young women (Hinsch 2013).

It is difficult to assess what the current status is of gender equality in China. Liberating as the economic reforms were, by diversifying employment opportunities and choice (Tan 1998), due to discrimination and inequality in education and on the labor market, the reforms ‘opened possibilities but also shut them down’ (Shen 2015: 5-11). Empowered by consumerism, the reassertion of a natural femininity naturalized assumptions about gender and sexuality (Evans 1997). Be it the influence of consumerism, the absence of gender expression in the earlier period, or what Ding et al. (2014) describe as the people’s attitudes showing a tendency to regress to traditional (Confucian) points of view , or a mixture of these explanations; Tan (1998) concludes that the reforms have ‘exacerbated gender differences in power, income and status.’ (Evans 2000: 224).

2.2 Gender and representations

According to important gender theorists (Butler 1990; Moore 1994; De Lauretis 1984), gender has to be perceived not as a static identity or (bodily) difference that is simply ‘being there’ (Geller and Stockett 2006: xvi), but as forming ‘through multiple processes, both conscious and unconscious, within the context of dominant discourses and categorizations and the differences within and between them’ (Moore 1994: 49-50). Evans (1997: 18-19) adds that these ‘discourses and their texts, …, participate in the production of their readers as subjects through offering sets of positions to identify with and challenge’, and ‘whether or not individual persons consciously acknowledge the dominant

(9)

gender categories of these discourses, they also participate in reproducing them by making representations and self-representations – both consciously and unconsciously – with reference to them’.

Following authors like Butler (1990) and Moore (1994): ‘sex, sexuality, and gender are socioculturally constructed categories rather than natural ones; categories created through discourse, representation, and repetitive performance within the nexus of the body’ (Geller et al. 2006: 8). The goal of feminist anthropology is to counter assumptions that gender and sexuality (together with class, race and nationality) are essential and universal categories (Gaetano 2016), to document gender construction over space and time and challenge their ‘naturalness’ (Geller et al. 2006: 125), and to deconstruct what being a woman (or man) means in different context (Evans 1997).1

Evans (1995; 1997) has shown how in post-reform China a discourse on gender and sexuality is being produced, based on the discourse created at the beginning of the People’s Republic. This discourse uses science ‘to legitimate fundamentally hierarchical gender relationships’ and holds ‘the assumption that gender characteristics of women are inseparable from women’s reproductive function’ (Evans 1995: 360). By linking sexual activity to its reproductive purpose, concerns with healthy reproduction determine the parameters defining acceptable sexual activity and provide the rationale for regulating sexual behavior (ibid.: 366) This way excluding the ‘abnormal’, like female celibacy and homosexuality, and defining female sexuality principally in relation to masculinity and not existent in itself outside the context of the heterosexual relationship where female desire is constructed in response to male desire (ibid.: 370-3).

In contrast with China in the 1950s, official agencies now no longer act alone in generating and controlling this dominant discourse but the availability of popular materials expands the status as well as the readership of this discourse (ibid.: 389). As a result, ‘the representational terrain is much greater than before, generated through an unprecedented range of forms from the commercialized

1 The concept of discourse I use here is very much influenced by Foucault and his notion of discourse in its connection to the structures of knowledge and power. Cf. Foucault (1984) and his explanation of the ‘technology of sex’ that places the body under control of marriage, family and social welfare.

(10)

images of the eroticized female body to those of the responsive, supportive, and self-sacrificing wife’ (ibid.). These representations may seem contradictory but Ballaster et al. (1991) suggest that the media recognize and reflect that ‘women’s lives in patriarchal societies are contradictory’ (Gill 2007: 193).

In a similar way as Evans (1995; 1997; 2000) has deconstructed discourses and

representations to explore how the media in China (official and/or commercial) are actively involved in creating gender (Gill 2007), many scholars have turned to what de Lauretis (1989) calls the

‘technologies of gender’: media like cinema, television, magazines, that construct the representation of gender.2_{Gill (2007: 2) stated that ‘the media, gender relations and feminist ideas are constantly}

changing and in flux’, and with the rise of for example the internet and social media the way of studying gender and doing ethnographic research has also changed (Rogers 2013; Hallett and Barber 2014).

As far as the internet is concerned, there have been no attempts in the English-speaking academic literature to study gender representations on this medium due to some theoretical and methodological concerns.3_{Because of the increasing importance of the internet and digital}

technologies in our daily lives, this topic deserves more scholarly attention. Combined with the fact that there is a large increase in visual culture (due to this explosion of digital technologies since the

2 For studies on gender representations in China, cf. Zurndorfer (2016) for television, Wu and Chung (2011) for TV commercials, see Evans (2000), Andrews et al. (2002) and Ying and Lowry (2012) for studies on magazines; Liang et al. (2016) discuss sex education picture books.

3 There have recently been more studies on how to research some kind of inherent bias in the internet and its search engines (cf. Baker and Potts 2013; Datta and Tschantz 2015) but no real attempt has been made to provide for a systematic approach to study representations online, and more

specifically using search engines. For studies on gender representations in online video games, cf. Brehm 2013 and Fox & Tang 2014.

(11)

late twentieth century) (Lister and Wells 2001),4_{it is important to expand the study of representations}

to these ‘new’ media and platforms.

Visual culture does not simply ‘reproduces structures of “reality” ’ (Kress and Van Leeuwen 2006: 45), but is created (or ‘constructed’) in a particular cultural context and with a particular purpose. This raises questions concerning the power the media, as the producer of this visual culture, has (ibid.). To quote Kress et al. (ibid.: 45): ‘they (the pictorial structures) produce images of reality which are bound up with interests of the social institutions within which the pictures are produced, circulated, read. They are ideological.’ Thus to study the visual in the context of the internet, it is important to explore questions about the power this medium has in creating representations and cultural meaning and its implications for using the internet as a source.

3 The Chinese internet and online search

This chapter discusses the properties of the Chinese internet and the role of search engines in indexing and regulating online content. The first subchapter deals with ways of perceiving and studying the internet and the digital. The second subchapter discusses the power of search engines as they, for an important part, decide on which content users access. The subchapter concludes by discussing the implications of this for studying gender representations on the internet.

3.1 The Chinese Web

Since China connected to the World Wide Web in 1994, the rise of information and communication technologies (ICT) has been exponential (Zhang and Zheng 2009). Reaching almost 300 million users in February 2008, the number of Chinese internet users has become the largest in the world (Xinhua 2008). By an estimate of the Internet Live Stats, users have increased with another 420 million since 2008, and with still only 52.2% penetration (ILS 2016), there is still much room for development.

4 Kress and Van Leeuwen (2006) argue that this increase in the visual has maybe come about because of the verbal not being adequate anymore to deal with information that has become so vast and complex.

(12)

The topic of the internet in China attracts much scholarly interest. It has been widely argued that the internet, with its inherent promise of democratization, would change the authoritarian nature of the Chinese state (Zheng 2008). However, through the Chinese Communist Party’s (CCP) managing of the internet, that promise has stayed a promise (Zheng 2009). The way China manages the internet at times seems contradictory, comparing for example, the official discourse of banning online

pornography and severely punishing its users and distributors, with its seemingly easy accessibility (Jacobs 2012); still the censoring of politically sensitive or possibly state-harming material is very real and has large effects for the Chinese internet users (King et al. 2013).

If scholars want to gain a deeper understanding of for example internet censorship in China or other aspects of the Chinese Web, digital methods can be used to study the natively digital or digitized (Rogers 2013). To ‘follow the medium’, as Rogers puts it (ibid.: 59), one can research which sites are still available and which are not, to what other sites certain sites link, etc.5_{Because the world is ‘made}

up of specific configurations of global, national, and local networks in a multidimensional space of social interaction’ (Castells 2009: 19), Schneider (2015a: 59) points out that ‘interactions are increasingly defined by digital communications.’. As technology more and more plays an important part in our lives, the technology, and by extension, ‘the medium matters’ (ibid.).

Following Zuckerman (2013), we must not limit how we think of the digital processes around us by seeing them as an elsewhere. For the case of the internet in China, the national manifests itself throughout its digital networks in what Schneider (2015a: 73) calls a ‘kind of sovereign, national intranet’, and which is thus not a borderless ‘elsewhere’. Going against the more optimistic views that ICT connects people behind borders (Shirky 2008), Schneider has shown how Chinese hyperlinks mainly link to other hyperlinks within the borders of the People’s Republic and rarely refer to sources outside its national borders (Schneider 2015a; 2015b). It seems that by censoring and tightly

5 See for instance the work of King and his colleagues on tracing online censorship by the Chinese government (cf. King et al. 2013) or Ng (2013) who did a similar thing for the microblogging platform Sina Weibo.

(13)

controlling and keeping the flow of information inside the national borders, authorities hope to ‘guide public opinion’ (China Media Project 2013) and maintain social stability.6

Benney (2014) states that digital technologies are not value-neutral but are often created by elites to serve certain purposes. In a Marxist sense, where the people who own and control the material means of production are the same as those who also control the production and distribution of ideas in society (Gill 2007: 54), it thus seems that the internet functions as a ‘traditional medium’ where the state and big companies increasingly set up the rules, despite the internet’s promise of democratization (Schneider 2015b: 68-72). Marwick (2013: 62) adds that ‘Leaders in this culture are the same people that technology discourse has celebrated for fifty years: young, white, rich men.’; this could have deep implications for gender dynamics and representations on this medium.7

If the digital world is defined by national borders and ‘traditional’ power structures, the implications for research are that the expertise of area studies and (critical) ethnographic approaches is needed to try to make sense of the digital (Schneider 2015a). According to Rogers (2014), there are three methodological ways of studying the digital: traditional methods, digitized methods, and natively digital methods. As the traditional methods (such as interviews, surveys or observations) are more used to study digital media usage (Schneider 2015a), these will not be used for this research. The focus of this research will be on representations in images returned by search engines, thus combining the study of the digitized (the online images, which function as any image and thus can be analyzed using traditional methods like visual analysis) and the natively digital (the search engine). Rogers (2014: 21) writes that: ‘The issue no longer is how much of society and culture is online, but rather how to diagnose cultural change and societal conditions by means of the Internet. The conceptual point of departure is the recognition that the Internet is not only an object of study but also a source.’ So to

6 Since the Arab Spring and the Snowden revelations, the CCP has re-nationalized search engines and increased censorship to increase its grip (Jiang 2014c).

7 This could again be compared with traditional media like news or journalism where a small, white-male elite make up the agenda to the effect that women issues like violence against women, or big events in non-Western countries tend to be underreported or even ignored (Gill 2007).

(14)

study representations using digital methods is like studying representations using any medium, taking into account the peculiarities of that medium.

As the focus will be more on the representations, the first (the digitized images) will be more important in this research. However, since search engines are my medium of study, it is important to first discuss this medium, its functioning and the difficulties or challenges of using search engines for digital methods research.8

3.2 Online search and the visual

Since 1997, search engines have become our main portal into the Web and have increasingly consolidated their position on the Web (Van Couvering 2009). By way of virtual integration, they provide a whole array of different services that ‘lock in’ the user (Schneider 2016: 71), binding surfers to their product. Google, the leading search engine in market share in the Western world, has become synonymous for surfing the Web and has become an important part of our (digital) lives (König and Rasch 2014). This dominance raises many concerns, for example considering the fact that

internet-users not always seem to be able to distinguish a search engine from a regular homepage (Hargittai 2007), and the built-in search plug-ins of more recent times will probably not help spread awareness (Schneider 2016: 66).

Search engines also are not public archives but serve a mere economical function where the engines profit more if they can attract more users and so collect more data that they can sell to

advertisers (König and Rasch 2014).9_{To attract the most users, they have to be efficient (meaning they}

8 Schneider (2015a: 63) discusses three channels into the study of the (natively) digital: homepages, the hyperlink and search engines. For the purpose of this research, only search engines will be used. But it would be useful to follow up this research by adding an analysis of the other two to gain two new dimensions on the topic. Schneider omits an important fourth channel: the mobile application.

9 This also raises questions concerning privacy as personal data is being sold to companies. Cf. König and Rasch 2014

(15)

return the ‘right’ results for a certain query) and accessible (meaning the technical skill required of the users is not too high) (ibid.).10_{Search engines filter results for its users, thus filtering away the}

irrelevant to present the valuable information (Schneider 2016), and thus functioning as the gatekeeper who decides what info we get to see. In this sense, Hogan and Luka (2014: 301) argue that search engines function as a ‘hidden’ infrastructure that is online and accessible to everybody, but that does not reveal its true workings of how its algorithmic functions index and filter online material.

Rogers (2013: 86) adds that the search engine also functions as a ‘status-authoring device’ that ‘delivers those “deserving” to be listed as the top sources.’ As the search engine in a Foucauldian sense, constructs ‘the natural order of things’ (Jiang 2014b: 227), this holds very important implications for earlier questions raised about who holds the power on the internet; as the search engine is our main portal into the Web and studies show that users most of the time only read the first results page (Silverstein et al. 1999)11_{, users therefore validate ‘without question the hierarchy}

arbitrarily established by the algorithm.’ (Miconi 2014: 39). Combined with the fact that the search results are filtered (and also censored, in the case of China) before they reach the user, Halavais (2009: 1290) argues that these services increase ‘the attention paid to those people, institutions, and ideas that have traditionally held sway’, and thus ‘represent a largely conservative force’.

Despite the clear bias that is built into the engine, the search engine is still perceived as being neutral, partly because of the obscurity of its workings (König and Rasch 2014). According to Schneider (2016: 65-6), this bias can be negative, that is: ‘a warped perspective or false truth that someone has intentionally manipulated for their personal gain’; but is also often positive: ‘in the sense of a warped perspective that confirms what I (the user) expected’. Either way it shows ‘a distorted

10 Search engines thus qualify as mass media (or ‘mass art’) as their goal is to reach the largest possible audience (cf. Carroll 1998).

11 Through what Jiang (2014a: 1089-90) calls ‘search concentration’, or Shirky (2003) the ‘power law’, the hyperlinks that search engines always rank the highest, receive most attention and will increase their chances of always being ranked high-up on the search result page, thus always reinforcing their strong position online.

(16)

perspective’ when showing the world (ibid.: 66), as the technology ‘ends up sitting between you and reality, like a camera lens’ (Pariser 2011: 237). In this sense, the search engine also qualifies as a ‘traditional medium’ as it shows the user its own vision on reality.12

Additionally, search engines have become increasingly personalized as they define a user according to geolocation, search history and (online) social network, as to better serve users and (perhaps more importantly) advertisers (Hogan and Luka 2014). As a fourth definer, I add the element of language as studies have shown that not only the language one puts into the engine is of importance for the results it returns (cf. Schneider 2015a), but also the language setting can make a difference (cf. Jobin and Glassey 2014).13_{This personalization creates what Pariser (2011) calls a ‘filter bubble’ that}

not only distorts the view one gets on reality in a highly biased way, but this view is also different than the view anybody else gets as it is completely personalized, thus possibly locking the user in its own comfortable (and very biased) ‘bubble’.

Hogan and Luka (2014) argue that it is thus important to see what is revealed and what is concealed by the engine, what the reality is that is being presented. Through these personalization effects, the classic is filtered away to favor the recent, the local is favored over the global and the personal over the universal (ibid.); technological and commercial incentives thus draws user’s attention closer to home instead of directing them outward (Pariser 2011). For search engines on the Chinese intranet, besides the technological and commercial, it is maybe more importantly the national

12 Compared to traditional media, Pariser (2011: 186-194) points out a ‘shift in agency’ inherent to digital media. As you may choose search engine, you may not choose the algorithm as the working of the algorithm is obscured.

13 For example in Schneider’s (2015a) study, where differences in putting in simplified or traditional characters in Google, Baidu and a number of other Chinese search engines are already examined, the element of language setting could be added as Google can be set in Chinese (simplified or traditional). Baidu’s language setting for example, cannot be changed. It only has the option to that the results will be only in simplified or traditional characters.

(17)

that search engines use to keep users inside the borders, consulting state-approved content and building an ‘imagined community’ (Schneider 2015b; cf. also Anderson 1991).

For the purpose of this research, the internet will be used as a source, where search engines offer a ‘unique empirical window into the study of culture’ (Sanz and Stancik 2014: 4). Ørmen (2014) rightfully states the importance of archiving and studying online content. Although this is difficult, as websites are constantly changing and slowly dissolving and search engines’ algorithms frequently get updated (Rogers 2013). These issues and the consequences of personalization for using search engines for research will be discussed in the methodology section.

As computer media function as ‘meta-media’ that embed the production of other media in their design, ‘traditional’ methods can be used to study the digitized (ibid.: 25-6; Schneider 2015a). But as computers and digital interactions are increasingly shaping our world, it is also important to take into account the characteristics of computer-based signs and see how these are different from text, images or film (Andersen 1997; 2001). As Floyd (1987: 191-8) recommends, we should move from a ‘product-oriented perspective [that] regards software as a product standing on its own’ to a ‘process-oriented view [that] views software in connection with human learning and communication.’14

With the increase of the digital, there has also been a clear increase in the visual (Kress and Van Leeuwen 2006). Together with the ubiquitous social media, search engines have an important role and power position on the internet in general, and thus also in the ‘semiotic landscape’ of everyday life (Kress and Van Leeuwen 2006: 33), making search engines crucial in understanding this aspect of culture. Therefore, the images returned by the search engines will be analyzed using visual analysis to study gender representations on the Chinese internet, drawing on theory from content analysis to social

14 Andersen (1997) argues that we should dedicate a separate field of semiotics, computer semiotics, to the study of computer signs as computers differentiate themselves as ‘sign vehicles’ because their existence depends on human use and interaction (ibid.: 23). Manovich (2013) argues that software, in its ubiquity in all computer applications, should be studied in a separate software studies to deal with the causes (and not only the effects) of the software, as it has become so vital to our epistemology.

(18)

semiotics and computer semiotics. Studying (gender) representations using the internet and search engines is an essential endeavor in a world increasingly dominated by digital communications, and the increase in the visual that came along with it and that only promises to expand.

4 Methodology

The goal of this research is to find a systematic and valid approach to studying (gender)

representations using search engines in China, as this digital platform is a place where ‘different meanings of the category “woman” (or “man”) are being contested’ (Evans 2000: 233) or created. Using search engines for research results in some unique methodological challenges.

One challenge is that the query results from a search engine are not ‘just there’ to be found by the searcher, but are ‘coauthored’ by the search engine and the searcher (Rogers 2013: 93). Combined with the far-reaching effects of search personalization as discussed earlier, search engine research becomes unreplicable (Feuz and Stalder 2011). But instead of trying to achieve complete objectivity and reliability of doing search engine research, Ørmen (2014) argues that researchers should actively work with these limitations and treat search results ‘as historical documents archived at a specific time and place by researchers with more or less clear biases in their approach’ (ibid.: 234).

To limit some of the bias through personalization effects, I will change my IP-address using the Firefox Foxy Proxy-application to locate my computer in China when searching Baidu and ChinaSo, and in Hong Kong to search Google HK (cf. Schneider 2015b). For Google HK, I will also change my language setting to simplified Chinese as this may also change results. Before searching, I will clear my search history to further reduce personalization effects (Rogers 2013).

Another challenge is that images pulled by a search engine lack their original context (in what kind of website, advertisement, article, they were used). As the context of an image is very important

(19)

for analysis (Collier 2001), it is possible to ‘visit the page’ and place the image back in its original context. But as the focus of this research is representations in search engines, I argue that the surroundings of the search engine itself are sufficient as context for the purpose of this analysis.15

The search engines that will be used for analysis are Baidu 百百, ChinaSo 百百百百(Zhongguo

sousuo) and Google Hong Kong, as these give a good overview of the internet content in and around

China.16_{The data will be pulled from the search engines on one single day after a period of}

monitoring.17_{The images and other content used for analysis are those that are returned by the search}

engines on the first page of the results, as this is in line with research saying users rarely go beyond the first results (Silverstein et al. 1999).

The searches that will be performed will use two sets words: on the one hand the popular addresses ‘pretty lady’ 百百 (meinü) and ‘handsome guy’ 百百 (shuaige), and the neutral ‘man’ 百百 (nanren) and ‘woman’ 百百 (nüren). These popular addresses changed from being an address to a

15 Given the ‘meta-nature’ of the internet, search engines and multimedia, this study could be expanded towards a multimodal analysis (cf. Kress and Van Leeuwen 2001), including for example a study of homepages, or even use more digital methods and map out a network of online connections of these webpages (cf. Schneider 2015a; 2015b), to this way be able to put the content into more context.

16 Baidu is the biggest search engine in terms of market share and users in China (CNZZ 2014). ChinaSo (formerly Jike) only manages ca. one percent of China’s search traffic but as a state-run engine it should give an interesting ‘official’ look into how the CCP manages the internet through this search engine (cf. Jiang and Dîngă 2014). Google would be expected to give its users an ‘outsider-look’ compared to the Chinese search engines, however Schneider (2016) has shown that it returns much more official content than for example Baidu, making it also interesting for this study. Google left Mainland China in 2010 but is still accessible in Honk Kong.

17 All data will be collected on Saturday the 28th_{of May, after started monitoring Friday}

the13th. By using this ‘short burst model’ (in comparison to a longitudinal model, cf. Ørmen 2014: 230), there is a chance that the search engine realizes it is being studied as it will return more variability in results the faster you poll it (Zuckerman 2011).

(20)

genuine handsome guy or pretty lady to a general term to address anybody in their youth (Wang and Lu 2010). As these addresses have changed in meaning while still holding the connotation to a

person’s beautiful appearances, I argue that these terms are ideal to research representations of gender, especially to study representations online using a search engine, as these terms can be used as search terms and thus openly show how these terms or their neutral counterparts are constructed by the online community in China.

The research will consist of three parts: the first part deals with search engines and their visual and interactive properties. The focus will be on the different interfaces and more important on the different interfaces of the ‘images’-section of the different search engines (if available). The images of these different categories will be qualitatively analyzed. The second part deals with the composition and content of search results and the different suggestions the search engines provide. In the third part, different sets of images of different searches will be analyzed using a content analysis.

Apart from the social semiotic theory I will use for the analysis, I will use computer semiotic theory by Andersen (1997) to account for digital and computer processes encountered during the analysis. The goal of this research is not to achieve completeness in analyzing the data, as this is impossible, but rather to look for and analyze ‘deviance (irregular or odd query-building) and anomaly (results that amuse or offend)’ (Rogers 2013: 110).

(21)

5 Search engine’s gender representations

The analysis will consist of three parts. The first part deals with search engines and their visual and interactive properties. The second part deals with the composition and content of search results and the different suggestions the search engines provide. In the third part, I will analyze different sets of images of different searches using a content analysis.

5.1 Homepages

In this part, the homepages and standard properties of the different search engines will be analyzed. The focus will be on the composition and the different interfaces as these interfaces serve as the parts ‘of system processes that are seen or heard, used, and interpreted by a community of users’ (Andersen 1997: 201), which makes them crucial to the analysis. As this part deals with the default set-up of the different search engines, no search will be used but only the interfaces already there, provided by the search engine, will be discussed.18

18 Benney (2014) argues that design choices matter a lot as they decide how a service will be used, what choices are there for the user and which are omitted. These choices in turn serve the

(22)

I will analyze the different interfaces of the ‘images’-section of the different search engines in their relation to gender. In these interfaces, different categories of images are presented. The images of categories that represent gender in a certain way will be qualitatively analyzed on the basis of criteria of representational, interactive, and compositional meaning (cf. Jewitt and Oyama 2001).19_{I will}

analyze two categories of Baidu and two of ChinaSo. As Google does not offer such categories, it will not be incorporated in this analysis.

5.1.1 Composition

All three of the homepages of the different search engines, Baidu, ChinaSo and Google HK, work according to the same compositional principle of centrality. The search bar, as search is the main function of the search engine, is the most important element of the homepage that holds the marginal elements, such as affiliated sites and functions, log-in options and settings, together. Typically, the search bar, together with the logo of the engine, is the most salient.

business interests of the providers of these services and are thus not neutral.

19 From the different categories, I choose those whose pictures represent gender in a certain way. I made this selection after a cursory survey of going through each category and dismissing them when they didn’t have any pictures where men and/or women are depicted. I did not choose categories consisting of cartoon figures. Although these also represent gender in a certain way, the focus here is on realistic representations of gender.

(23)

The different engines of course offer different services and structure their homepage in different ways. User-friendliness and a simple, inviting design seems to be a key aspect of successful engines like Google and Baidu.20_{Exception to this rule is the homepage of ChinaSo, which offers a}

much wider range of services and information.

Figure 1 Homepage of ChinaSo. (Left is above the fold, right is below the fold.)

It almost looks like a mix between a search engine and a news site, as users can scroll down to find various news content ‘below the fold’ (figure 1). Beneath the search bar there also are two separate, constantly revolving interfaces where the top one shows various headlines and the bottom, larger one, shows news headlines with a pictures and popular sites. The bottom interface has as a caption saying ‘what netizens are searching’ 百百百百百 (wangyou zai resou), creating a sense of ‘imagined community’ among its users. In the search bar, already a search is recommended to a certain news item, and clicking the search bar shows ten of ‘today’s hot (news) items’ 百百百百 (jinri redian), ranked from one to ten.

The headlines and the pictures quickly reveal that most content is government-related, but there also seems to be room for more popular items. But apart from the logo also the search button

20 It could of course be argued that search engines like Baidu and ChinaSo copied Google’s successful model.

(24)

gives away that this is a CCP mouthpiece search engine, where you can expect that your search results will be highly ‘national’ as you not simply search but ‘search the nation’ 百百百百 (guosou yixia). The engine in general gives a very screamy and attention-seeking impression; with the overload of content and the fast rhythm of the revolving interfaces (they change ca. every three seconds). Also note the striking resemblance with its logo and Google’s logo.

Interfaces that are present both in Baidu’s and ChinaSo’s homepage are QR-codes that you can ‘swipe’ 百百百 (souyisou) to connect with your smartphone in a true multimedia way. Baidu also offers the possibility to use an image as search input and to get results based on that image. Google does not offer any of these services and its layout is in general much more sober compared to Baidu and ChinaSo. These two Chinese search engines seem to much more create a discourse of advanced, cutting-edge technology, where the search engine can always find new ways to incorporate new technologies in its design to always stay relevant and remain its strong position on the internet.

5.1.2 Categories of images

The ‘images’-section 百百 (tupian) of ChinaSo and Google can be accessed directly from the homepage, whereas for Baidu it has to be found in the ‘other products’-interface 百百百百 (gengduo

chanpin). Baidu and ChinaSo offer different categories of suggested images that mirror popular

searches or topics.

Baidu offers ten categories, ranging from artistic photographs and landscapes, to pets, cartoons, wallpapers and avatars. For the purpose of this research, I choose the two categories: ‘showcased celebrity’ 百百百百 (mingxing ruzhu) and ‘small fresh and clean’ 百百百 (xiao qingxin). ChinaSo offers eight ‘main’ categories that are visually showcased in the center on the page, with another ten in the recommendations bar above, and another eight in the ‘more’ 百百 (gengduo) option in the bar. Also in the center of the page, next to the eight main categories, is a fast-changing interface showing three different photo reportages of travel destinations in- and outside of China. The main categories include giant panda’s, penguins, different kinds of flowers, an online game and popular travel destinations. The categories in the recommendations bar include wedding, cars, famous

(25)

mountains and animals, in the ‘more’-section there is among others fashion, gardens, furniture and civil aviation. Of all ChinaSo’s categories I choose ‘celebrity’ 百百 (mingxing) and ‘jewelry’ 百百 (zhubao) for analysis.

Baidu’s ‘showcased celebrity’-category literally showcases one particular celebrity for five days. There then is a video or interactive postcard with QR-code underneath the search bar before the results where the celebrity promotes Baidu and him- or herself. The showcased celebrity on the day of analysis is Huang Xuan 百百, a Chinese actor.21_{ChinaSo doesn’t showcase one particular celebrity in its}

‘celebrity’-category, but standard shows the ‘popular recommendations’ 百百百百 (remen tuijian). The images shown however clearly favor one particular celebrity, a Chinese singer and actor Lu Han 百百, which is the first name in the right recommendations bar with the names of popular celebrities. Besides this recommendations bar, there is another one on the left where you can choose between male and female celebrities, and celebrities from what place you would like to see, from Mainland, Hong Kong and Taiwan, Japan and Korea, or Europe and America. Note that for each place, there first has to be distinguished between genders.

Comparing and analyzing these two categories from Baidu and ChinaSo, it is important to first note that the images returned by the search engine, (as the user chooses a certain category), contain a

priori a conceptual structure where the person on the images is not defined on the basis of he or she

‘doing something’ but for ‘being something’, viz. being a celebrity (Jewitt and Oyama 2001: 143-4).22

21 During the 18 days of monitoring, the three celebrities that were showcased before Huang Xuan were all female.

22 This choosing of category functions as if the user him- or herself would search for a certain celebrity or other terms like ‘cars’ or ‘landscapes’. As the uses chooses category, the definer of the specific category gets entered into the search bar as a query for the ‘images’-section of the search engine.

(26)

These images thus rarely have a narrative structure as they miss a vector (an action that helps in narrating a story in the image) (ibid.). We can see in figure 2 that in most of the images the vector is subservient to the celebrity itself that is portrayed. For the ChinaSo images, there are a few that show more than one person and there is a vector connecting these two persons (in leaning into each other or holding each other), but these also serves a conceptual purpose to classify all participants as

‘celebrities’.

The shape and layout of the images is different for Baidu and ChinaSo, with the first using horizontal rows of connecting rectangular images and the latter vertical rows of similar shaped images (albeit a bit longer in length) with a caption for which celebrity is shown. Baidu’s structure thus resembles Google Images’ and one could argue that Baidu was eager to copy the successful model of Google while ChinaSo went its own, more Chinese way where the vertical rows reflect the traditional Chinese writing that goes from top to bottom and right to left.23_{The rectangular shape of the images is}

historically grown as this makes it easier to stack images and objects (Kress and Van Leeuwen 2006:51).

To compare the images of Huang Xuan and Lu Han (I leave the other celebrities on ChinaSo away for now), it is obvious that they both have a very different style and image, where Huang Xuan seems more of a macho (especially if you look at the first row of pictures) and Lu Han more of a nice and not so aloof popstar. They are both young, attractive, popular and male. The pictures of both celebrities mostly use medium shots or close-ups, which creates a social or intimate and personal relationship with the viewer (Jewitt and Oyama 2001). Among the pictures of Huang Xuan there are also a few using long or full body shots that create an impersonal relationship and help create this macho or ‘bad boy’-look (ibid.).

Analyzing the gaze of the two celebrities in the picture, we see Huang Xuan alternating between looking into the camera (or direct at the viewer) and away from the camera in the pictures, whereas when he looks at the camera he or smiles cordially, or gives a cool ‘bad boy’-look. Lu Han is

23 Kress and Van Leeuwen (2006: 3-4) argue that ‘visual grammar’ followed lexical grammar to express meaning and thus is cultural specific. So the dominance of the left-right structure in Western culture followed the writing direction of the Roman script.

(27)

shown here more looking away from the camera creating an ‘offer’-look that implies passivity, something that could create a less aloof image and thus a closer bond with fans (compare with the cool and distant look of Huang Xuan), or smiling into the camera.

(28)

Combining the results of this small analysis with some of the accessories, clothes and other elements in the pictures, I assume that the setup of these images matches the image these celebrities want to achieve for themselves: the cool Huang Xuan with the sunglasses and the open shirt and the nice popstar, that maybe also leans a bit to the ‘cool-side’ but still remains less aloof. The modality shows naturalistic colors (with one exception with Huang Xuan) which makes these pictures of these celebrities appear more ‘real’, especially as these search engines want to bring their fans as close as possible to their idols.

(29)

Figure 3 Baidu’s first result for the ‘small fresh and clean’-category.

Baidu’s ‘small fresh and clean’-category, of which the first page of results is shown in figure 3, does not give away that it is linked to gender from the name itself. However, the images clearly show a certain representation of gender: along with the flowers, dreamy landscapes and wallpapers, we see a couple of young, pretty looking women, two small girls, two young couples, a wedding advertisement from Baidu Hunqing 百百百百 (hunqing), Baidu’s own ‘wedding celebration’ planner, and a tattooed female torso with the hands covering the breasts.

What are we to make of this? It could be argued that the ‘small fresh and clean’ in the form of flowers or dreamy landscapes, connotes to innocence, purity and maybe even virginity, therefore the pictures of the young girls and women. This is reflected in the recommended ‘related searches’ 百百百百 (xiangguan sousuo) above, which range from ‘small fresh and clean flower’, ‘cartoon’, ‘landscape’, to ‘pretty girl’, ‘schoolgirl’, ‘schoolgirl avatar’ and ‘tattoo’. But if tattoo’s then also then this category, the framed image that shows only a tattooed torso with tattooed hands covering the breasts and the image of the tattoo of ‘me’ and ‘your mom’ on an unidentified body part, are definitely interesting search results.

The images all have a naturalistic modality, with some pictures high color saturation maybe even leaning towards a sensory modality, which creates a fantasy like, ‘more than real’-feel to the pictures (Jewitt and Oyama 2001). It could be argued that most of the pictures again display a conceptual structure, except for the two couple pictures (and the wedding advertisement) where the action-reaction narrates the nature of their relationship (cute or playful), as they don’t interact with the camera or viewer. The pictures of three young women all use medium shots, creating a social

relationship, while the images of the couples use an impersonal long shot. The two little girls stick out with one close-up and the other being a long shot, and looks like the image could be found in a catalogue or advertisement for children’s clothes. The two young women in the first row look away from the camera creating an ‘offer’ or passive feel to the pictures which fits the innocence or purity that this category wants to display. Overall, the techniques used in the images and analyzed here, create a representation of the innocent, young and attractive women or innocent child.

(30)

ChinaSo’s ‘jewelry’-category is not as much about the jewelry but more about the women wearing the jewelry (figure 4). It shows all female models or celebrities on the catwalk wearing fancy dress and showing off their necklaces and earrings. The captions below the images describe the celebrity or model and the design of the jewelry worn to specify the images as in: ‘Chen Huilin’s jewelry atop black dress; long pearl necklace’ for the second image in the third column (Van Leeuwen 2005: 229-30).24_{The images show a clear connection between wealth or success and femininity. But}

they can be read either as the women having obtained success and its financial rewards, or as the (high class, beautiful) women, together with the jewelry, are the objects to be ‘obtained’, by perhaps a

Figure 4 First results of ChinaSo’s ‘jewelry’-category.

wealthy businessman. In that last reading, the woman becomes as the jewelry: expensive, high class, beautiful and a way to show of success and wealth.

This last reading seems to match the first four photographs, the woman on the cover of the magazine and the woman in the right corner best where the participant of these images do not appear to be a celebrities, but ‘mere’ models or beauty pageant participants, not somebody who has achieved top financial success. The gaze of the women in the first, third and fourth picture seems to demand attention while the gaze of the second picture, the woman in the magazine and the woman in the right corner display an offering gaze, which makes them appear more passive. Their distance to the viewer

(31)

ranges from impersonal (long shot) to social (medium shot, the woman on the magazine cover and the woman in the last picture of the first row), and one intimate or personal relationship (close-up of the bottom right picture.

In contrast to these are the two women on the second horizontal row, the one in white, and the one in black. Their pictures use of a long shot, the trained camera-smile and the background revealing some sort of red carpet or important event, imply that these are celebrities. It could be argued that the woman in the last picture of the first row also fits this first reading more as she appears quite

ambiguous and is therefore harder to classify. From the caption we can read that she actually is a celebrity.

5.1.3 Conclusion

We here have already seen one of the main workings of the search engine, as the search engine’s interface works according to the property of the ‘recursive process’, in the way that it uses its own output as input, as was the case for the categories of images discussed here (Andersen 2001: 421). This recursive process is also true for the ubiquitous recommendations that all the search engines offer at various places, which together with the general ranking and working of the search engine, I will discuss in the following part.

This part of the analysis has shown how different representations of gender seemingly innocently creep into aspects of daily life like in something that has become so commonplace as a search engine. This is a place where people presumably pay little attention to things like how gender is represented, as the search engine in anyway is seen as neutral. The discourse that is created by Baidu and ChinaSo is one that naturalizes a world of always young, attractive and glamorous celebrities, a consumer model where everything (and everyone?) has a price, and where virtues like innocence and purity are very important. The set-up of the images and the marketing of ‘showcasing’ a celebrity create a greater feeling of intimacy between its users and their celebrity idols.

(32)

5.2 Interfaces

In this part I will focus on the general ranking and working of the search engine and how its interfaces, like the auto-complete in the search bar and the recommendations, interact with the two pairs of ‘gendered terms’: ‘handsome guy’ or ‘shuaige’, and ‘pretty girl’ or ‘meinü’, and the neutral ‘man’ or ‘nanren’, and ‘woman’ or ‘nüren’. I will look for any bias or interesting results while I discover more about the workings of the three different search engines and how they react similarly or differently to these search queries.

The top results together with any advertisement, auto-complete or suggestion and recommendation which the search engine gives, are of much interest to the analysis. As Shirky’s (2003) above discussed ‘power law’ has shown, these are the results that receive most of the attention and

therefore continue to be listed as a top result or top recommendation by the search engine. As all three search engine function in their own way, they will be discussed separately.

5.2.1 Baidu

When searching for one of our ‘gendered terms’ and comparing the set-up and results of the different searches, we already see much difference (figure 5). The lay-out for the search results of the four searches is almost different each time. Ranking for similar results is also different, with for example the ‘images’ appear as the first result for the two female terms ‘nüren’ and ‘meinü’, while for the search ‘nanren’ they appear lower and for ‘shuaige’ even lower still (they are not seen any more in figure 5).

(33)

To the right of the results, we also get a different lay-out: for ‘nanren’ we get one

advertisement and for ‘nüren’ it stays blank,25_{while for ‘shuaige’ and ‘meinü’ we get two elaborate}

columns with ‘related words and phrases’ 百百百百 (xiangguan cihui) and ‘related persona’ 百百百百 (xiangguan renwu). For ‘shuaige’ they appear as first the ‘related words and phrases’, with the ‘related persona’ below it while for ‘meinü’ it is the other way around. This shows that most of the different elements of the search engine are ‘transient features’, as they change according to search query (Andersen 1997: 193). The result pages for ‘shuaige’ and ‘meinü’ are much more filled with content. This could perhaps be due to popularity of the terms and they thus are more popular search terms than ‘nanren’ and ‘nüren’.

Almost all of the results for all searches (not counting the advertisement) are Baidu’s own sites like its ‘images’-page, Baidu’s own social platform, Baidu Tieba 百百百百, its own encyclopedia, Baidu Baike 百百百百, and its own library database, Baidu Wenku 百百百百. This confirms the ‘own-content bias’ also shown in Jiang (2014a: 1091-2) and Schneider (2016). Exception is ‘nanren’’s ‘recent relevant

25 I forget to capture a screenshot of Baidu’s result for ‘nüren’ on the day of pulling the result and later I did not manage anymore to connect to Baidu through a Chinese server using FoxyProxy, therefore the screenshot of the results of ‘nüren’ used is one taken while accessing Baidu from my home server. So where there should have been one advertisement to the right, like on ‘nanren’’s screenshot, is now blank.

(34)

news to men’ 百百百百百百百百百 (nanren de zuixin xiangguan xinxi) and ‘meinü’’s ranking list, but this together with the ‘related words and phrases’ and ‘related persona’, all refer back to Baidu’s own search content.

The content shown on these first results pages moves from the seemingly neutral ‘nanren’, with the news articles and encyclopedia entry, to very sexualized images of both ‘nüren’ and ‘meinü’ (which will be analyzed in the next part), attractive looking celebrities and a range of questionable ‘related words and phrases’ such as ‘shuaige’’s ‘away with homosexuality’百百百百 (qu tongxinglian) and ‘meinü’’s ‘young (under aged) girl’ 百百 (shaonü), ‘beautiful legs-lovers’百百百 (mei tuikong) and ‘f-cup’ f百百 (zhaobei). Clicking through to these two last pages respectively shows pictures of long legged beautiful women and women with very large breasts. The recommendations are endless and allow the user to browse away ad infinitum while the content seems to become increasingly explicit or sexual. The algorithm will also remember your choices and will continue to feed you content in a certain direction.

Interestingly, the ‘away with homosexuality’ of ‘shuaige’’s ‘related words and phrases’ contrasts with the below-placed ‘male comrades’ 百百百 (nan tongzhi) in the ‘related persona’ column. As ‘comrade’ in Chinese already serves as slang for ‘homosexual’, the ‘male’ before it seems quite redundant. If you navigate to the search results, you say many pictures of homosexual couples kissing. On the larger picture of the recommended pictures, one of the man’s pants is hanging on his knees so his erect penis is sticking out. The results of ‘away with homosexuality’ contrast these with top results asking questions like ‘How will homosexuals ever obtain sexual satisfaction?’ and ‘Would you go to a country where homosexuality is legal?’26

The advertisement of ‘nanren’ links to the webpage of a Japanese company selling a substance ‘for external use’ to prevent premature ejaculation. The webpage is full of super muscular men, sometimes topless or sometimes naked but framed so the private parts are always covered. One topless guy is holding a topless woman who is covering her breasts with her arms. The website makes an appeal to their customers by asking ‘Successful man, can you give happiness to your woman (or to

26 Original text: ‘百百百百百百(百百)百百百百百百百百百百百?’ from 百百百百 and ‘百百“百百百”百百百百百百百百百百百?’ from 百百百百.

(35)

‘women’ plural)?’ and gives a lot of statistics claiming that many women think that their happiness is dependent on the ability of their man to give them an orgasm.27

Figure 6 Recommendations and auto-complete using the different 'gendered words'.

The recommendations and the auto-complete mechanism in the search bar (figure 6) show mostly references to popular culture in the form of film, TV-shows and music groups. But there are also results that display a certain representation of gender (that is, not taking into account how gender is represented in the different instances of popular culture recommended here). For instance ‘nanren’’s recommendation of ‘man stands up’ 百百百百 (nanren xiongqi) must be read as ‘man-up’ or ‘man gets erect’, judging by the pictures of men in tight speedo’s and of men performing oral sex on a

woman and the links with tips against erection problems the search engine returns. For ‘meinü’, the ‘small games’ 百百百 (xiao youxi) auto-complete leads to animated games which have a clear sexual motive and seem to be going into the direction of the Japanese animated pornography, hentai. Interesting is for the auto-complete of ‘shuaige’ that different results appear when you click in the search bar than if you type ‘‘shuaige’’ in the search bar. So there appears the auto-complete of ‘the big bird of ‘shuaige’’ 百百百百百 (shuaige de daniao), which is not present at the ‘normal’ auto-complete suggestions. This suggestion shows pictures of muscular men, only wearing very tight underwear or nothing at all.

27 Original text: ‘百百百百百百百百百百百百百百’ and ‘78%百百百百百百百百百百百百百百百’.

(36)

5.2.2 ChinaSo

The search results of ChinaSo for the ‘gendered terms’ show a much different view (figure 7). We see similar columns to the right of the results as was the case with Baidu’s ‘shuaige’ and ‘meinü’, but the search results themselves appear quite different than Baidu’s. No result of an ‘images’-page is among the top results and the results seem to consist primarily of news content. The ‘own content-bias’ is also clearly visible here as most content comes from ‘national search’s 百百 (guosou) own encyclopedia, news application and ‘special interest’-interface 百百 (guanzhu), but there are also other news sources and a reference to a movie for both ‘nanren’ and ‘nüren’. Apart from the columns to the right, you

could argue that the search results are not always very relevant to the search query, with the overload of news content. For ‘meinü’, there are no columns but a repetition of the ranking list of news bulletins we have seen in part one as suggestions in the search bar. There seem to be no advertisements.

The content of the columns range from ‘other people also searching’ 百百百百百 (qitaren hai sou) (compare with the ‘what internet friends are searching’ on the homepage) to ‘persona’ 百百 (renwu), ‘musical works’ 百百百百 (yinyue zuopin) and ‘literature’ 百百 (shuji). In these categories, I presume users find a lot of celebrities and references to popular culture, but also for example a link to ‘Thai ladyboys’Figure 8 Compilation of ChinaSo's recommendations for the 'gendered terms' in the search bar.

(37)

百百百百 (Taiguo renyao) at ‘nanren’’s page. One book title out of ‘nüren’’s ‘literature’-column reads: ‘: ‘I want a pair of celebrity legs.'28

The suggestions under the search bar show more references to popular culture for ‘nanren’ and ‘nüren’ but for ‘shuaige’ and ‘meinü’ they show a highly stereotypical and sexist picture: ‘shuaige’ has suggestions of ‘shuaige’ carrying their daughter (as a show of good or correct male behavior?) and ‘riot police ‘shuaige’’ 百百百百 (shuaige tejing), while ‘meinü’ has suggestions for women in traditional Chinese dress, the qipao 百百, women as slaves and ‘lift the skirt of a ‘meinü’ games’ 百百百百百百百百 (xian

meinü qunzi xiao youxi).

The auto-complete suggestions (figure 8) again show a lot of references to popular culture, mainly for ‘nanren’, ‘nüren’ and ‘shuaige’, although ‘nanren’ also has two recommendations to two different men’s magazines with pinup cover girls. ‘meinü’’s recommendations however offer a wide

range of sexist content, with pictures of women in lingerie or bikini, pictures of dancing women and what look like strippers; and even pornographic content. This was also to be found for ‘shuaige’’s ‘‘shuaige’ sleeping with ‘meinü’’-result 百百百百百百百 (shuaige yu meinü shuijiao) but it is not clear if the links go to actual pornsites or are mere clickbait to attract people to virus-infected sites, or both. Also there are recommendations to video chat with ‘meinü’ and ‘meinü’ undergoing a ‘big body change’ 百百 百 (da bianshen) linking to all kind of content related to plastic surgery and cosmetics.

(38)

5.2.3 Google Hong Kong

Google HK’s search results for the ‘gendered terms’ (figure 9) seem more simple than Baidu’s or ChinaSo’s as they only have the listed search results and no recommendations under the search bar or extra interfaces to the right of the screen. The ‘images’-section is within the top results for all searches and the very first result for both ‘shuaige’ and ‘meinü’. Above ‘nanren’ and ‘nüren’’s ‘images’ there are a couple of sponsored links which are divided from the ‘images’ by a very soft frame line. Google HK does not seem to favor its own content except for the ‘images’-section.

The content of the top links seems fairly innocent with some lifestyle websites for both ‘nanren’ and ‘nüren’. For ‘shuaige’ and ‘meinü’ the sites are more visual with compilations of

pictures of ‘shuaige’ and ‘meinü’ but of the top results, ‘meinü’ had four of these (that are all the top results except for the ‘images’-section) against one of ‘shuaige’.

For the suggestions in the search bar (figure 10), most seems not out of the ordinary, with again plenty of references to popular culture. An interesting suggestion is ‘nüren’’s ‘woman flavor’ 百百 百 (nüren wei) (or ‘feminine’ according to the site itself) which links to a Baidu Baike entry claiming:

(39)

‘Woman flavor is an only by women possessed charm and attractiveness, it is the glamourous reflection of women’s tenderness, grace, kindheartedness, intelligence and

independence.’29_{Further in the entry, the different ‘traits’ 百百 (tezheng) and ‘standard’ 百百 (biaozhun)}

of ‘woman flavor’ are given. This entry is a prime example of the naturalizing of a distinct gender difference, combined with the authoritative voice of an encyclopedia.

Furthermore, the recommendations for ‘shuaige’ show a whole range of pictures and avatars, with many pictures showing a very sexualized ‘shuaige’, showing of his muscular body. However, there appear to be no suggestions for ‘meinü’.

5.2.4 Conclusion

We have seen that the search results returned differ a lot for search engine used, but also for different search terms for the same search engine. Every search engine seems to have a certain way to rank its permanent features and alternate between the various transient features. These differences also show the different reality being constructed by the search engine and thus the different discourses or representations of and on gender.

(40)

By just using these simple search terms, there was already a whole bunch of eroticized, sexist and even homophobe and homoerotic content (of Baidu). In general, the women terms and especially ‘meinü’ returned most explicit and sexist material, with the very sexual ‘little games’. ‘shuaige’ also appeared in many instances in a very sexualized way but still less than ‘meinü’, where the ideal ‘meinü’ according to Baidu has long beautiful legs and an f-cup. For the men a discourse is created of the strong man that has to perform (also sexually) and be successful, and take good care of his woman.

5.3 Images

In this part of the analysis, different sets of images of the different searches for the ‘gendered terms’ will be analyzed using a content analysis based on variables Bell (2001) used in his study. Jewitt and Oyama’s (2001) social semiotic approach will also be used as example. To better fit the data, I will complement with more variables, specific to the images that will be analyzed here. The images used in the analysis can be found in the appendix.

5.3.1 Variables

Following Bell (2001), the images will be analyzed based on the variables social distance, visual modality, behavior or contact, and judged age ranges. These I complement with ‘race’ and ‘level of undress’. These variables will be put into a number of different values as to be able to quantify them for analysis.

For social distance, following Jewitt and Oyama (2001: 146), I use the shots frame sizes: close-up, medium shot, and long shot. They respectively stand for an intimate/personal relationship, a social relationship and an impersonal one. I add one value of the ‘selfie’, which counts as a close-up but is still different as it is taken by the participant of the picture him- or herself and thus appears even closer than the close-up. This can also be a measure for professionality of the picture as a ‘selfie’ points to a regular, amateur internet or social media user.30

Gender Representations on the Chinese Internet