Graphicon studies have focused mainly on the meanings and functions of individual categories of graphicons (typically emoticons, emojis, or stickers) (see the review in Tang & Hew, 2019) rather than considering graphicons in relation to one another as part of a larger ecology, as recommended by Herring and Dainas (2017). Even fewer studies have explored the changes that take place over time within this ecology.

An exception is Konrad et al. (2020), who posited that graphicons typically pass through three phases: an early phase, a peak phase, and a phase of decline or conventionalisation. In Western contexts, Konrad et al. (2020) situated emoticons in the third phase, based on their reduced frequency and their conventionalised use as punctuation; emojis in the second, peak phase; and stickers in the first, earliest phase. They also predicted that the trajectory would be more advanced in Asian countries, with stickers in ascendance and emojis starting to decline. Their predictions were based on interviews and surveys conducted with English-speaking Facebook messenger users rather than longitudinal data, however.

Studies with actual longitudinal data so far have confirmed that emoji have overtaken emoticons in the West (Pavalanathan & Eisenstein, 2016). As emoticons have declined in frequency of use, their functions have become conventionalised as a type of punctuation (Markman & Oshima, 2007; Provine et al., 2007), with the conventionalisation further demotivating their usage (Albert, 2020; Konrad et al., 2020). A longitudinal study in a Chinese context that examined the frequencies of kaomojis (Japanese-style emoticons), emojis, and stickers found only partial support for the prediction of Konrad et al. (2020), however. Kaomojis exhibited a rise-and-fall pattern, but their decline was not accompanied by an increase in emojis or stickers (Zhang et al., 2022).

The discrepancy between model-based predictions and actual data highlights the need for further investigation into the factors influencing these trends. Konrad et al. (2020) propose that popularity (frequency of use) and shift in pragmatic function are the primary drivers of the transition between graphicon categories. However, these could be seen as effects rather than as causes of the rise and fall of graphicon categories. It remains to be determined what factors, especially technical ones, drive their evolution.

Additionally, although an increasing number of new graphicon designs within each graphicon category are available, little is known about how each unique graphicon is used in practice. Several quantitative studies have identified the top 20 most frequently used emojis in different corpora (e.g., Liu et al., 2020; Wang et al., 2024), but little is known about the use of less frequently occurring emojis. Zhang et al. (2022) found in their longitudinal corpus that the trends in graphicon types differ significantly from those for graphicon tokens; however, no reasons for this discrepancy were given.

Graphicons on Chinese social media constitute a unique ecosystem with distinctive designs and usage. The graphicons incorporate the logographic features of Chinese characters (Li & Zhu, 2019; Zhang et al., 2022) and convey rich cultural messages from Chinese idioms and traditions (de Seta, 2018; Zhang et al., 2021). The culturally enriched designs of graphicons on Chinese social media are preferred over Unicode emojis (e.g., none) by users of different platforms (de Seta, 2018). Chinese users tend to use emojis more frequently in contexts involving family and social relationships compared to their Western counterparts, who also use emojis extensively in professional and casual settings (Guntuku et al., 2019). Although several studies have pointed out that stickers are now very popular among Chinese social media users (e.g., de Seta, 2018; Ge, 2020; Ge-Stadnyk et al., 2025; Sandel et al., 2019), we have been unable to find comparable information on the frequency of sticker use.

Graphicons on Bilibili are distinctive in several ways that make the platform a valuable case study for graphicon usage. Bilibili is one of the most popular social media platforms in China, featuring high stickiness and interactivity (Gao, 2021; Sun et al., 2023; Tian, 2021). As we argue here, graphicons play a significant role in user interaction, cultural expression, and community building. Bilibili has released platform-specific sets of emojis and stickers that have been overwhelmingly popular – for instance, a TV set icon that is the symbol of Bilibili, and the Popular Words Series, which are graphic representations of selected popular verbal expressions from comments or ‘danmu’ (short messages overlaid on the video itself) (Zhang et al., 2022). These platform-specific graphicons, which are associated with Bilibili’s fandom cultures, enhance user engagement through expressive visual communication, allowing users to indicate their membership and build a sense of identity within the platform community (Wu et al., 2018). The platform also monetises graphicons by offering exclusive packs for purchase, encouraging users to invest more in their online presence (Yang, 2021).

The development and widespread adoption of graphicons, as with other social media affordances, are the result of ongoing interactions between the digital platform’s capabilities and the social needs and practices of its users. This interplay is what Trist (1978) called socio-technical systems theory, which emphasises the interaction of social and technical subsystems. The social subsystem focuses on human factors such as people, relationships, and norms, while the technical subsystem involves tools and technologies, such as hardware and software, required for tasks (Bostrom & Heinen, 1977; Kling, 2007). Both aspects are crucial for understanding digital communication.

A socio-technical perspective has been found useful in exploring the evolution of digital communication, which is marked by the mutual influence of social practices and emerging technologies. For instance, it has shown how digital maps transitioned from traditional cartographic methods to web-native media forms that reshape social practices and spatial awareness (Hanchard, 2024). Studies in this line demonstrate the potential of the socio-technical approach to shed light on the complexity in evolving digital communication.

To the best of our knowledge, this perspective has not been explicitly applied in graphicon studies, although the focus of most existing graphicon research could be categorised as either social or technical. Studies from the social perspective explore topics such as the meanings and functions of graphicons (Herring & Ge-Stadnyk, 2024; Logi & Zappavigna, 2021), usage patterns in relation to user demographics and profiles (Koch et al., 2022), and cultural influences on graphicon use and interpretation (e.g., Chik & Vásquez, 2017; Zhang et al., 2021). In contrast, the technical perspective has been considerably less studied. Yet technical factors such as the design and availability of graphicons clearly have an impact on their usage. For instance, Unicode, as a technical standard, distinguishes emojis from emoticons and other iconographic formats, enabling their use across devices, operating systems, and Internet platforms, thereby facilitating their widespread circulation (Berard, 2018). The layout of the emoji keyboard has an impact on the time required for emoji selection (Pohl et al., 2017), thereby affecting its usage. Stickers, unlike emojis, are not supported by Unicode or standardised formats, which could hinder the development of widespread conventions of sticker use (Tagg & Lyons, 2021). While social and technical perspectives both offer valuable insights into graphicon usage, studies investigating their interactions within the same longitudinal corpus are notably lacking.

This study extends the socio-technical perspective by integrating a historical analysis of both social and technical influences in a large corpus of graphicons. Specifically, we explore the following research questions: 1) What are the trends over time of the relative frequencies of Chinese graphicons on Bilibili? 2) What irregularities in the trends are observed when compared to the rise-and-fall model of graphicon evolution proposed by Konrad et al. (2020)? and 3) What socio-technical factors in the platform’s history explain the trends in the Bilibili corpus, particularly the irregular frequencies that deviate from the predictions of Konrad et al. (2020)?

Our corpus consists of 14 years of longitudinal comment data collected from the Bilibili platform. Bilibili is a video-sharing platform that, like YouTube, allows users to post comments below the videos and features short danmu messages that are overlaid on the video itself.

The Bilibili platform was chosen for several reasons. Firstly, it is one of the most popular Chinese social media platforms, with average monthly active users reaching 0.315 billion (around one-fifth of the Chinese population) in the first quarter of 2023 (https://www.Bilibili.com/read/cv24070769/, retrieved July 9, 2024). Secondly, the comments can include kaomojis, emojis, and stickers. Lastly and most importantly, Bilibili preserves a historical record of the comments posted below the videos, including the graphicons in the comments, and the comments can be captured automatically.

The data consist of comments and replies to comments from the channel of Bilibili’s annual Spring Festival Gala Show (hereafter, the Bilibili show) (https://space.Bilibili.com/1868902080, retrieved April 5, 2022). This channel was chosen because it is the only one that includes consecutive comments dating back to 2010. The show consists of a mash-up video of content provided by professional users to celebrate the Chinese New Year, and it is released on the eve of the Chinese New Year. Besides the show videos, the Bilibili show channel includes videos related to the gala show, such as trailers, teasers, and outtakes. Like the Bilibili show videos, these videos include older comments and replies and thus are included in our data.

Comments and replies from all 54 videos available in the channel, covering the years from 2010 to 2023, were captured and stored in March 2023 using Python and the Scrapy tool. A total of 992,856 messages (including both comments and replies) were collected. Danmu messages were excluded from the data due to their limited use of graphicons.

The analysis was conducted in two stages. To obtain the trends of kaomojis, emojis, and stickers in use, we first identified their occurrence in the messages using a semi-supervised process of deep learning (Qin et al., 2019) and manual identification and calculated their frequency. The frequencies were normalised based on the total number of messages each year. We also applied Voyant (https://voyant-tools.org/), a web-based text analysis tool for analysing word frequencies and trends in comments. We derived patterns for the corpus as a whole, as well as identifying deviations from the model predicted by Konrad et al. (2020).

In the second stage, to account for the patterns in the frequencies, we adopted a historically informed socio-technical approach. We conducted digital archaeology to reconstruct the history and usage of graphicons on Bilibili, based on which we identified factors that could plausibly be related to irregularities and classified them into social or technical categories. Digital archaeology refers to ‘the use of digital technology to investigate and communicate the past’ (Morgan, 2022, p. 214); in this case, the past being examined is also digital. Specifically, our search for historical information included the following major sources: 1) information available on Bilibili for the timeline of events, platform practices, and culture; 2) annual Bilibili financial reports for user numbers and profiles; 3) media reports about Bilibili for major events in its development and its social impact; 4) the internet archive Wayback Machine for old Bilibili webpages; and 5) questions and answers about Bilibili on platforms such as Zhihu (Chinese Reddit) for graphicon input settings and graphicon accessibility.

We recognise that our backgrounds as researchers influence our analysis of Bilibili’s user communities and platform dynamics. Our approach is informed by cross-disciplinary perspectives from computer-mediated communication, linguistics, and information science. Further, the third and fourth authors are native Chinese speakers who are users of Bilibili, and the first and second authors have previously researched Bilibili and other Chinese social media platforms. While we believe these perspectives enrich our analysis, we acknowledge that our interpretations may be shaped by our familiarity with Chinese cultural contexts. Concerning data collection and use, the study followed the platform’s guidelines for data access, scraping, and usage for academic research.

The normalised frequencies of all graphicon tokens for each of the 14 years are displayed in Figure 1.

These trends provide partial support for the evolutionary trajectory of graphicons proposed by Konrad et al. (2020), particularly as regards kaomoji. Deviations from Konrad et al.’s (2020) predicted rise and fall of each graphicon (emoji replacing emoticons (kaomojis) and stickers replacing emojis) include the following:

Although kaomoji rose and fell from 2010 to 2019, the decrease in kaomoji was not replaced by emojis or stickers, leading to a significant drop in graphicon use in 2017.

As part of the findings from our digital archaeology, a brief overview of the platform’s history is essential to understand the socio-technical factors contributing to the irregularities in graphicon evolution. Bilibili was founded in 2009 as a non-profit video-sharing community for Anime, Comics, and Games (ACG) fans (Sun et al., 2023). The platform derives its name from the popular anime A Certain Magical Index, where the protagonist Mikoto Misaka’s electric attack produces a sound humorously referred to as Bilibili, which later became the nickname given to her by the male protagonist (https://www.Bilibili.com/read/cv154805/, retrieved June 4, 2024). In 2013, Bilibili received its first investment from Tencent, was listed on the NASDAQ stock index in 2018, and was listed on the Hong Kong Stock Exchange in 2021. Along with its increasing commercialisation, Bilibili has seen changes in its policies, content, and the number and profiles of its users.

Bilibili released open registration in 2013 with a strict membership test to familiarise the newcomers with the platform’s culture and practices. Only individuals who passed the test were granted permission to interact with comments or ‘danmu’. The test was discontinued in 2015 to accommodate a broader group of users. Not long after, a paid membership was implemented in 2017, suggesting the platform’s full commitment to commercialisation. Together with these developments, the video content on the platform expanded from pure ACG to pan-ACG (including general games and animations, then movies and TV programs), then further to a broad spectrum of videos. Alongside the expansion of video content categories, the main user groups also shifted from ACG fans to pan-ACG fans, and then to pan-entertainment users. By 2023, the monthly active users had increased to over 300 million. The expansion of the platform from a closed community for ACG lovers to a massive pan-entertainment site is illustrated in Table 1, alongside trends in graphicon use.