Content uploaded by Angela Xiao Wu
Author content
All content in this area was uploaded by Angela Xiao Wu on May 06, 2021
Content may be subject to copyright.
This is a preprint publication. Cite as
Wu, A. X. (2022). The ambient politics of affective computing. Public Culture, 34(1).
The Ambient Politics of Affective Computing
Angela Xiao Wu
Media, Culture, and Communication
New York University
Abstract
Much attention to affective computing has focused on its alleged ability to “tap into human
affects,” a trope also foundational to broader theoretical frameworks of big data surveillance.
What remains understudied and undertheorized is affective computing’s social life, which
unfolds through ambient politics—the fraught processes whereby interested parties contest and
collude on its deployment. This essay calls to reorient critical analysis of affective computing
away from its design epistemics to its ambient politics and, in parallel, to shift the focus from
interiorized subjects to conditions of collective existence. My intervention begins with a
conceptual distinction between actual emotional experiences and emotionology—that is, the
symbolic-material infrastructures through which society recognizes and normalizes emotion
displays. Empirical inquiry into ambient politics, I argue, requires us to note this distinction and
approach affective computing at the level of the emotionological. To illustrate this research
strategy, I trace how such portable technologies as sentiment analysis and Like buttons wound
up redefining collective action in China, which partly explains the conservative turn observed in
Chinese online cultures since the mid-2010s. I demonstrate that in its social life, affective
computing is exceptionally malleable; social actors repackage, reinterpret, and remediate it to fit
their agendas. Its ambient politics is thus a politics of infusion and amalgamation, in contrast to
the simplification and standardization characteristic of its design epistemics.
Keywords
China, sentiment analysis, Like button, emotion, affect, emotionology, social media, surveillance,
positive energy, digital activism, affective governance
Acknowledgements
My thanks to Guobin Yang, Natasha Schüll, Elizabeth Lenaghan, three anonymous reviewers,
and the Public Culture editorial board for their invaluable comments at different stages. This
essay also benefits from conversations with He Bian, Yige Dong, Lily Chumley, danah boyd,
Sareeta Amrute, Melissa Gregg, as well as audiences at the Center on Digital Culture and Society
and the Center for the Study of Contemporary China at the University of Pennsylvania, the
Institute of Communications Research and the Department of Communication at the University
of Illinois Urbana-Champaign, Data & Society, and the Annual Meeting of Society for Social
Studies of Science in New Orleans, where I presented earlier versions. My fieldwork and writing
were supported by the Henry Luce Foundation/American Council of Learned Societies (ACLS)
Program in China Studies and the Chiang Ching-kuo Foundation.
2
In August 2009, an article entitled “Mining the Web for Feelings, Not Facts” introduced
readers of the New York Times to sentiment analysis, or algorithmic techniques for automatically
classifying open-ended written accounts by emotion.
1
While marveling at the power of
sentiment analysis to ride the “bull market” of social media, “opening a tantalizing window,”
and “translating the vagaries of human emotion into hard data,” the author also worried that this
systematic computation of emotion might sideline the use of factual processing to make decisions
(Wright 2009). Citing this coverage, critical scholarship designated sentiment analysis as a
primary instrument that marketing and political campaigns wielded to exploit people’s emotional
investments in participatory cultures (Andrejevic 2013, chap. 3). Today when we talk about
affective computing, at least in Euro-American societies, we think about recognizing emotions
from facial micro-expressions. Sentiment analysis, albeit an expansive subfield of affective
computing (e.g., Hussain and Cambria 2015), has fallen out of the limelight. This trajectory
alone should lead us to question the perennial hype about emergent technologies granting access
to psychological states. That said, affective computing invites more exploration, both empirically
and conceptually, that is not anchored by its capacity to capture mental activity, either treated as
truism or a scientific scandal.
By affective computing I mean calculative technologies that purportedly recognize
emotion based on data detailing a wide array of human behaviors. Such technologies include Like
metrics, whose nominal emotive claim heralded the arrival of the social web (Gerlitz and
Helmond 2013; Gehl 2014).
2
I keep the definition expansive not least because in technical
literature the boundary of affective computing is not conclusive (e.g., opinions differ as to
whether data are restricted to physiological traces), but more importantly for conceptual
interventions: Seemingly diverse technologies, so long as they claim to register emotion, have
consistently instigated theorization and critiques that share the same primary concern with
interiority.
Critical scholarship continues to situate a series of such technologies in what is described
as a longer history of the “becoming public of affect,” often entwined with the expansion of
extractive capitalism (Arvidsson 2011, 49). The same vocabulary—“opening up,” “getting in,”
1
Machine learning or dictionary-based, sentiment analysis breaks texts into smaller units and
accesses their combination and frequency. In practice, it may label “anger abounds” as “negative,”
“The government should step in” as “neutral” and, were the classification not binary, a fan’s
exaltation over their idol’s latest photoshoot, “I’m devastated. Help!” as “fear.”
2
Facebook Likes were also the building block of Cambridge Analytica’s “psychographics” (Hindman
2018).
3
“drilling deep,” “unexplored frontiers,” and so forth—has recurred in popular and academic
literature over a decade. The notions of “digital tapping into human affects” and “capitalizing on
emotions” have anchored early conceptualizations of sentiment analysis and Like buttons—to
which I will return—in strikingly similar manners. The same preoccupation with penetration is
also foundational to recent theorizations about big data surveillance where profit and
administration presuppose the capture and modulation of human interiority (e.g., Couldry and
Mejias 2019; Zuboff 2019). Amidst AI fears and uncertainties, nothing can better attest to this
prevalent imaginary than the popular portrayal of Chinese digital systems. As the liberal West’s
Other, China is deployed to fuel its darkest fantasies—as the ne plus ultra in their abilities to
track and discipline invasively. This essay departs from this preoccupation to center the social life
of affective computing (Appadurai 1988).
In the early days of sentiment analysis in the United States, social media content was
readily understood as material for sentiment analysis. Critics worried that the new digital
environment, equipped with new analytical tools, would sideline evidence-based rational
deliberation. Such an assumption reflects unspoken presumptions: ordinary users’ expressions
are feelings, whereas news from reputable media is facts. The truth is, however, that sentiment
analysis systems “see” emotion wherever they are deployed. Once fed to sentiment analysis,
news turns emotional; this was exactly what happened in the 2014 Facebook emotion study,
where news and personal posts indistinguishably constituted the “emotional makeup” of users’
News Feed (Kramer, Guillory, and Hancock 2014). And once put under the lens of sentiment
analysis, news further depreciates. It becomes another type of content automatable for “empathic
optimisation,” in the words of researchers of social media disinformation (Bakir and McStay
2018). Sentiment analysis, it seems, remains enmeshed in American concerns about political
polarization in ways that broadly accord with the valorization of the sensual versus the intellect
vis-a-vis the democratic process since Aristotle (see Gross 2007; Rancière 2013).
Meanwhile, sentiment analysis appears to strike a different chord in the Chinese context.
The COVID-19 pandemic prompted a national machine learning contest to train models on
curated datasets from Weibo, China’s much-larger-than-Twitter platform. The aim was “to
quickly discover people’s mood swings, so that policies and directives are made with
pertinence—a mission of great social value” (SMP2020-EWECT 2020). Thanks to its language-
specificity, sentiment analysis has separate routes of annual patenting activity (Figure 1).
Whereas the number of sentiment analysis-related patents has dropped sharply of late in the US,
Chinese patenting shot up in the mid-2010s. But rather than a lag in cutting-edge research, as
4
this essay will show, the contour of Chinese sentiment analysis patenting suggests aspiration for
entrepreneurial implementation irrespective of the technique’s (in)validation within scientific
communities.
Figure 1. “Sentiment analysis” Patenting by China and the US, based on Google Patents.
The divergence between American and Chinese trajectories stems from ambient politics.
3
I use “ambient politics” to refer to the processes whereby social actors contest and negotiate over
affective computing’s interpretation and deployment. Compared to the epistemological premises
built into the technology, the ambient politics of affective computing have received relatively
little attention.
4
Yet the social consequences of affective computing can be political, economic,
and epistemic, and they cannot be deduced from the epistemics embedded in its design features.
It is instead contingent on whether and how the technology—which makes autonomous and
universalist assertions about emotion—is made part of local projects, whether and how it is
lodged into the symbolic-material infrastructures through which a society recognizes, valorizes,
and regulates emotional expressions. To illustrate this, I interrogate how, through ambient
politics emergent from diverse agendas of Internet governance, such portable technologies as
sentiment analysis and Like buttons wound up reconfiguring what collective action is about in
3
Such divergent trajectories echo what Jasanoff (2015) highlights in her research program on
sociotechnical imaginaries of desirable futures—that is, the “reception” of science and technology by
non-scientific actors and institutions often varies across contexts.
4
For a related overview of the science and technology studies literature, see Jasanoff’s (2015)
discussion on the field’s relative inattention to technological systems’ interplay with local political
authority, moral order, and cultural resources, in contrast to its concentration on the social
construction of their formulation and materialization.
5
China.
5
At its heart, this essay seeks to reorient theorization of affective computing away from
individualized and interiorized subjects to shared conditions for collective agency and sociality.
From Design Epistemics to Ambient Politics
Emotionology: the attitudes or standards that a society, or a definable group within a
society, maintains toward basic emotions and their appropriate expression; ways that
institutions reflect and encourage these attitudes in human conduct (Stearns and Stearns
1985, 813).
In their interrogation of historical changes of emotion, social historians have advanced
the term emotionology to avoid conflating actual experience of emotion with norms about
emotional expressions. Loving behavior by parents is not the same as recommendations about
how to behave lovingly in child-rearing manuals; actual patterns of anger cannot be assumed
from a community’s repression of anger through socialization. Emotionology appraises emotion
and designates the location of emotional expression. It can be thought of as symbolic-material
infrastructures buttressing the physique of our emotional world. These infrastructures consist of
cumulative, layered technologies of perception (and occlusion) that institutions of oppression
and exploitation set in place (Gross 2007; also see Ahmed 2004). They may afford the enslaved
and the poor narrower emotional ranges, seek out signs of elusiveness in women (Alder 2002),
and interfuse negative energy and social criticism. To conceptually differentiate the actual
emotional experience of those under power is thus also politically meaningful (Reddy 1997).
Investigating emotionology proper (and steering clear of interior workings) is important in its
own right. Emotionology may serve as a powerful force in shaping behavior, independent of
actual change in emotional levels (Stearns and Stearns 1985). It is part of what Jacques Rancière
(2013, 8) calls the “regime of the sensible:” a shared “system of a priori forms determining what
presents itself to sense experience,” the condition under which collective agency and communal
practices are imagined.
We see a certain degree of emotion-emotionology conflation in both the critical literature
on affective computing and in extant discussions of China’s societal campaigns to spread
“positive energy” which serve as the backdrop for my telling of local histories of sentiment
5
My empirical analysis draws on preserved web archives, scholarly papers, media reportage, and
industry and government documents in the course of the 2010s, as well as fieldwork in the online
opinion analytics sector conducted between 2016 and 2019.
6
analysis and Like buttons. Let me begin with the latter.
The Chinese phrase “positive energy” (zhengnengliang) can be traced to July 2012,
during the London Olympics torch relay, when Weibo users celebrated Chinese torchbearers for
“igniting positive energy.” Soon picked up by state media and official documents, it turned into a
versatile catchphrase associated with a cheerful mentality, deeds conforming to social norms, and
national economic and cultural superiority (P. Yang and Tang 2018). In the eyes of its critics, the
discourse of positive energy epitomizes Chinese “therapeutic governance” which fosters affective
self-fashioning in service of neoliberalism (P. Yang and Tang 2018; Hird 2018; Chen and Wang
2020; Zhang 2015).
As ubiquitous as this rhetoric may be, people’s internalization of state-led imperatives
cannot be assumed. Observing migrants’ vernacular practices to “be happy” and “get by” in
Shanghai’s urban ruins, ethnographers remind us: “To read subjectivity directly from public
discourse is not to investigate subjectivity at all” (Gregory Simon, quoted in Richaud and Amin
2020, 83; also see Reddy 1997). This point resonates with social historians’ earlier note that
“there has been […] too much temptation to assert novel emotional experience, on the basis of
admittedly novel emotionology, than the facts warrant” (Stearns and Stearns 1985, 825).
The same conceptual and methodological tensions underlie interrogations of affective
computing and its implications. Much critical work centers on design epistemics—that is, the
various presumptions and values that are built into seemingly neutral calculations. Thereupon it
takes two routes. The first lands on the creation of new subjects and sociality per design
epistemics, in ways not dissimilar to discursive production. The interface of mood tracking apps,
for example, sometimes bespeaks the remaking of the self. Similar conflation between emotion
and emotionology also extends to broader theorizations on big data surveillance (Couldry and
Mejias 2019; Zuboff 2019). This muddiness stems from the fact that the vocabulary associated
with affective computing, like educational manuals from the past and the positive energy
discourse from contemporary China, revolves around constructs of emotion.
6
In contrast, I advocate for approaching affective computing strictly at the level of the
emotionological, which is also a prerequisite for any empirical inquiry into its ambient politics.
Affective computing designates particular human behavior—whose digital traces undergo
algorithmic calculation—as a venue to “display” discrete emotions, not unlike a wide spectrum of
6
Exceptions include Davies (2017) and Williamson (2017), whose analyses remain focused on how
technologies impose “politically and commercially desired forms of feelings” as integrative to the
valuations of the emotions in implementation contexts.
7
modes and systems, past and present, for emotion-gauging. From embodied-experiential
sympathizing to self-reports of subjects, from “superficial” gestures to visceral-physiological
recordings, all of these technologies essentially proffer standards for denoting emotions. (see
Dror 2011) Seen in this light, practices in critical humanities are included that, in advocating
perceptual and interpretive changes, foreground feelings as being triggered, addressed, and
channeled through images, architectures, connections, and rhythms.
7
To investigate affective
computing with an insistence on the emotion-emotionology distinction means to refrain from
recognizing emotions in one’s own analysis and instead focus on tracing the contours and politics
of the social recognition of emotions, i.e., emotionology. This in turn avoids arguments about
people’s emotional life and leaves the exploration of emergent subjectivities and senses of self to
socio-anthropological investigations of users and usage (e.g., Schüll 2016).
The second route taken by critical work on affective computing is to discredit its design
epistemics by foregrounding scientific disputes and developers’ arbitrary choices. Techniques to
recognize affect in written accounts, facial micro-expressions, tones of voice, and gaits alike tend
to posit universally applicable emotion taxonomies and a fixed relationship between people’s
inner states and their behavioral displays (see Crawford et al. 2019). Historians and
anthropologists have amply challenged the existence of psychological constants; even
psychologists and computer scientists disagree about measuring emotions by behavioral traces,
however granular (Stark and Hoey 2021; also Leys 2017). Yet as urgent as this line of
interrogation is, we should note that affective computing’s traction and hence social
consequences occur outside the academy, where it is taken up as expertise, service, and product.
Arguably the early twentieth-century antecedent of affective computing, the polygraph (once
named the emotograph) remained popular in police investigation and employee screening for
decades in the United States (but nowhere else), despite cumulative scientific evidence disputing
its accuracy (Alder 2002). Likewise, Chinese sentiment analysis reigns not due to its scientific
validity, let alone accuracy, but its legitimation from political and commercial establishments
(also see Wu 2020a).
7
See Leys (2017) for a political critique of affect theory as a mode of analytics whose anti-
intentionalism converges with post-WWII empirical approaches to the emotions. Critical “affect-
recognition” scholarship also includes a diverse body of works on the “affective politics” of digital
media that considers affect and emotion’s promises and liabilities in civic digital culture (e.g., Dean
2010; Papacharissi 2015). What these studies share is a recognition of human emotions and how
they operate (or are operated on) in digital environments. As should become clear, my approach
takes a different tack.
8
While affective computing may be designed to enact, and while it may even succeed in
enforcing existing dominant emotionology (e.g., Williamson 2017), that its social life unfolds
beyond design intentions points to its transformative potential. But to investigate this process,
our analysis should no longer treat affective computing as a computational endeavor to capture
emotions, but instead as quantum media with polysemic emotional signification. Quantum
media are enumerative and algorithmic outputs. Being consumed by broader society, they are
prone to constant remediation (Wernimont 2019). Wrist wearable devices, for example,
produce certain “media accounting” of selfhood for sharing and pondering (Crawford, Lingel,
and Karppi 2015; also see Humphreys 2018). Polling graphs, which hail public opinion into
being (Herbst 1998), acquire variant connotations as they circulate. Yet as quantum media,
affective computing creates more spacious room for social actors to maneuver because the
meanings of emotion terms are extraordinarily malleable.
The construction of affective computing faces, more generally, competing established
theories of emotion for model building, and more specifically, “confusing emotional signals”
from humans whose contexts remain inscrutable.
8
Emotional polysemy may start as a
conundrum within design epistemics. But I wish to analytically invert it into an issue of strategic
ambiguity that ambient politics exploits. Consider as an analogy emoji, such as the “upside-down
face,” whose ambiguity makes them exceptionally appealing (Figure 2) (Auerbach 2019, 254).
As media with emotional polysemy, affective computing provides streams of fodder for
interested parties to repackage and remediate. This may include conveniently shifting between
motivational, evaluative, and experiential connotations of emotion as a composite phenomenon,
and attaching new meanings to the discrete emotions that affective computing posits to measure,
such as negativity, anger, and happiness.
Figure 2. The “upside-down face,” or Unicode character U+1F643.
Online Activism as Sentimental Waves
The Chinese internet has been an eventful place. While the Chinese government seeks to
8
These theories variably conceptualize emotion as motivating behavior, as evaluative signals
indicating judgment, as conscious feeling states, and as their hybrids (Stark and Hoey 2021).
9
prevent street demonstration and autonomous organization (Wang and Minzner 2015), the
digital domain has absorbed immense protest energy and containment efforts by authorities.
Thanks to its non-institutional and extra-institutional nature, Chinese online activism, compared
to its American counterpart, can appear more spontaneous and radical (G. Yang 2016). These
collective actions are known as “online events,” echoing the Chinese reference to offline protests
as “mass events.” Having witnessed the blossoming of online activism in the 2000s, many
observers noted that toward the mid-2010s, contentious events were giving way to “consensus”
events conforming to the ideological mainstream (e.g., G. Yang 2017). Wagner-Pacifici (2017,
85) writes: “forms … are the matter of events.” An event is hailed out of occurrences in ongoing,
everyday life by historically contingent forms such as portraits, gestures, naming conventions,
and rhetoric devices. Retracing this observed historical change, I inquire into the forms that “act
representationally, demonstratively, and performatively” to let events appear (and disappear) on
the Chinese internet.
9
The story of Chinese sentiment analysis charts its incorporation into one
such ascending form through ambient politics.
Chinese scholarship on the eventful Internet serves as our vantage point. This scholarship
has two separate origins: “new media event” research and yuqing research. Dating to the 2009
“New Media Event” workshop organized by Hong Kong scholars, the former is well-known
transnationally. The workshop hosted participants from mainland China and Taiwan, as well as
Daniel Dayan, who had just edited a book on the Beijing Olympics as a media event. This
alchemy of energy birthed a research agenda culminating in a field-defining collection published
in China (Qiu and Chen 2011). Extending Katz and Dayan’s (1992) typology of media events
(e.g., contest, conquest, coronation), Chinese new media event research sought to categorize
online collective action in terms of how its “narrative forms” relate to larger structures of power.
But unlike Katz and Dayan, who focused on how state and media elites forge broadcast media
events, Chinese scholars were interested in grassroots voices seizing microphones in the Internet
age. Their main methods were discourse analysis, online ethnography, and interviews.
Though much obscured to international observers, the other origin of studying online
events—yuqing research—soon triumphed through an outpouring of government and industry
support (Figure 3). Although frequently translated as “public opinion,” yuqing more precisely
9
I treat digital technologies as supporting architecture that renders assembly legible—what Judith
Butler (2015) calls “conditions of appearance.” This angle differs from the main literature on how
digital technologies affect politics, examining either their cost-reducing logic in collective action (e.g.
Bennett and Segerberg 2012) or, conversely, their appropriation by the establishment to gain
support or dampen resistance (e.g. Schradie 2019).
10
refers to intelligence on shifting expressions from the masses. While leaked yuqing reports from
government agencies always offer sensational peeks into China’s political system (Pan and Chen
2018; Batke and Ohlberg 2020), yuqing analytics as a genre actually pervade academic
publications. Notably, however, the high frequency of keywords such as “guidance,”
“management,” and “governance” in these publications alludes their applied orientation in
stability maintanance (see also Wang and Minzner 2015).
Figure 3. Volumes of journal publications on “yuqing event” (blue line), “new media event” (green line),
and “online event” (orange line), 2007-2019, based on China Science Periodical Database via Wangfang
Data.
The Yuqing Monitoring Office (renamed Yuqing Data Center in 2017) of People’s Daily,
the mouthpiece of Chinese Communist Party, started releasing annual yuqing reports as early as
2008. It did so first in collaboration with the Chinese Academy of Social Sciences and then as a
“data-based consultation think-tank” that sells customized yuqing insights and training (which
went public in 2012). In its first report, published in a high-ranking media and communication
journal, yuqing events were defined as “events most attended by netizens,” based on the total
numbers of posts containing specific keywords from three Chinese Bulletin Board Systems (BBS)
(Zhu, Hu, and Sun 2008). Over time, while more data sources continue to be incorporated into
these analyses (e.g., search engine queries, blogs, Weibo, WeChat, news media, mobile news
apps), the convention of event-picking by keyword frequency persists in yuqing research. Over
the years yuqing research has experimented with a variety of methods, including oft-crude
diagnostic description and content analysis to summarize “attitudes,” expert panel ratings along
linear schemes to indicate “social pressure,” and utilizing platforms’ own metrics such as Weibo
11
Trending, despite its mysterious algorithm.
Increasingly, these experiments are giving way to ready-made commercial yuqing software
that scrapes, computes, and visualizes publicly accessible data from the “Total Web,” including
but not limited to major platforms. China’s billions-worth yuqing industry has the so-called
“digital listening” sector as its equivalent in liberal democracies (Kotras 2020; Karpf 2016;
Kreiss and Mcgregor 2018). Some crucial differences include that while both serve a diverse
clientele including digital marketers, corporations, universities, and political campaigns, the
yuqing business’s major patrons are government agencies. Their transactions are visible in
procurement documents as early as January 2007 (Batke and Ohlberg 2020). The unverifiability
of sentiment analysis can be particularly appealing to government clients. Studying 653 leaked
yuqing reports that a prefecture government used for internal upward communication,
researchers remark that: “The overall assessment of sentiment is almost always positive” (Pan
and Chen 2018, 609). Another, and directly related, difference is that yuqing analytics is attuned
to crisis management, designed to spot sporadic, unexpected, transient motions online, not to
infer and symbolize public opinion for long-term chronological comparison.
As the coupling between academic research and the yuqing industry tightens (Wu
2020a), the latter’s vision techniques have achieved a wide subscription for probing online
activism. This extends far beyond scholarly knowledge. Yuqing analytics’ formulaic yields,
including graphics and narration according with industrial protocols, are curiosities in wide
circulation. Because of this high visibility, ordinary people have come to recognize that
collectively boosting online activity (i.e., creating “yuqing crises”) may draw authorities’ attention
to local injustice and cover-ups.
The epistemic shift renders visible some aspects of collective action while obscuring
others. Revealed through thick descriptions and critical semiotic interpretations, “new media
events” are ethnographically identified and they unfold in interaction with existing power
structures. Yuqing events, in contrast, are quantitatively emergent content clusters.
10
Furthermore, yuqing analytics are wave analytics. They render a subset of the Total Web data
demarcated by select keywords into a set of time-series plots. These plots show contours of
different variables including discrete emotions from sentiment analysis and word frequencies
derived from word-cloud analysis. The curving portions represent the events. As the intensity of
10
Unlike the theory-driven typology of “new media events,” yuqing events are loosely classified by
objects of attention, such as corrupted officials, business scandals, celebrity controversies, and
“social conflicts” (between people).
12
content production rises and falls, the collective action comes and goes. Sentiment analysis
usually occupies the most prominent panel of digital displays and the defining section of reports.
As sentiment analysis converts waves of aggregate content into waves of emotion, the nature of
collective action is rendered primarily emotional (Figure 4).
Figure 4. Sentiment “waves” and pie charts of overall emotion proportions from (a) a popular yuqing
software using keyword “garbage sorting” (a new policy) during June 27-August 7, 2019, based on
725,188 “Total Web” content items, and (b) a leaked report that a different yuqing company gifted to
central authorities on the death of Covid whistleblower Dr. Li Wenliang, based on “Total Web” data
collected within its 24-hour range; also see Wu (2020a). The former categorizes emotion into Positive
(blue), Negative (red), and Neutral (orange), and the latter, titled “Computing Emotion, Measuring
People’s Heart,” has five categories: Optimism (green), Anger (red), Sadness (yellow), Disgust (blue), and
Fear (green).
“The epistemic effect” of wave inscriptions in social scientific analysis, writes Stefan
Helmreich (2020, 292–94), is to “authorize wave lines as well as claims about waves as material-
processual things in the world.” This entails disambiguating now formally “wavy phenomenon”
from material causation that elides the inscription techniques. Even though yuqing analysts can
investigate each emotion category in the software by clicking all the way down to individual
posts, these posts appear as long lists of fragments lifted out of their contexts of utterance.
Sentiment-centric yuqing analytics thus disintegrate local discursive interactions to re-compose a
global facade of fluctuating affective curves. In addition, yuqing reports always have negative-
sounding emotions, such as negativity, anger, and sadness (Figure 4), as their focal “scientific
objects.” In the same timescape, the reports note actions taken by the implicated party (e.g.,
announcement, distraction, flooding, or—less common—censorship). If coincidental with the
13
inflection points of sentiment contours, especially the negative ones, these actions are considered
to have altered the event’s unfolding—a typical case of conjuring causality with formulaic wave
inscriptions. But ultimately, wave analytics posit the passage of time as the explanatory variable.
A yuqing wave reifies a purported law of nature: a yuqing event is bound to vanish, just like the
next will eventually arise. As such, yuqing events are isolated, periodic repetitions of emotional
outbursts that cannot build upon one another and bear no relevance to any structural
transformations.
11
Finally, yuqing analytics’ interpretive framework grows out of crowd psychology. In the
inaugurating report, People’s Daily’s analysts concluded by suggesting that deviant (piancha)
yuqing was driven by a “web mob” (baomin) (Zhu, Hu, and Sun 2008, 39): “It is a bizarre
phenomenon of ‘collective intolerance.’ …Mass agitation replaced rational self-restraints… In the
heat of the moment, [these people] cannot look beneath the surface at all the complex social and
psychological causes. They habitually rush to persecute what they think are the ‘bad guys.’”
Training workshops, manuals, and actual yuqing reports draw on a fusion of concepts and
statements from century-old crowd psychology and modern-day communication research. This
knowledge universe is evident in a photo from a senior yuqing practitioner showcasing the books
on his desk. Posted on his WeChat account for connecting with clients and colleagues, the photo
was meant to signal professional status. Next to his mug and keyboard, the vertical stack consists
of ten workplace wellness and team-building books and a tasteful reprint of archaic sex columns
from a century ago. On top of this tight stack conveniently lie four translated volumes, all
recently published: Malcolm Gladwell’s (2002) The Tipping Point: How Little Things Can Make
a Big Difference, Gustave Le Bon’s (1895) Crowd: A Study of the Popular Mind, Eric Hoffer’s
(1951) The True Believer: Thoughts on the Nature of Mass Movements, and George Orwell’s
(1945) Animal Farm. Striking as this assortment is, all items speak to the law of crowd
dynamics, despite the substance of contention. They may gather for anything. The crowd’s
unrestrained nature makes it inherently dangerous.
Tapping Likes to Spread Positive Energy
Like metrics are arguably the most widespread quantum media that are polysemic in
11
Compared to yuqing events’ confinement by the analytics’ imposition of keyword filters and
timescale, hashtag activism’s boundary is never set as it is shaped participants’ own deployment of
hashtags, which include recycling them in new contexts to constitute long-term struggles (see
Bonilla and Rosa 2015).
14
terms of emotion. As discussed, early conceptual work around Likes has informed theorization
about more sophisticated technologies of affective computing and surveillance more broadly.
Scholars designated the button as Facebook’s 2009 design triumph for monetizing emotion, a
manifestation of Eva Illouz’s (2007) “emotional capitalism” that is about harnessing the inner,
subjective life of the private self (Gehl 2014, 88). The rise of the “the social web,” or Web 2.0,
was attributed to Like data, which was put in contrast with the hits and links of the past
“informational” Web 1.0. “[T]he Like button’s capacity to instantly metrify and intensify user
affects” enabled the “Like economy” (Gerlitz and Helmond 2013, 1349).
This notion about unmediated tapping was complicated by later empirical studies that
showcase the multifarious strategizing users bring into their decision to click Like (Eslami et al.
2016). The semantics also matter. Experiments show that thumbs-up buttons marked with
“Respect” or “Recommend” instead of “Like” fare systematically differently (Stroud,
Muddiman, and Scacco 2017). Whereas these findings problematize generalizing the Like
button’s societal implications from its design premises, my history of the Chinese Like button
below focuses on what ambient politics can achieve by engaging with its polysemic affective
signification.
This history is most illuminating when told in parallel, and then entwinement, with the
ascension of positive energy as a hegemonic discourse since late 2012. In China the Like button
is known as the zan button. Zan is conventionally used as a verb meaning not “like,” but
“praise” or “commend.” At the end 2012, Yaowenjiaozi (2012), an old-guard authoritative
Chinese language magazine included zan in Ten Annual Catchphrases. The entry marveled that
zan began to function as an adjective meaning “good.” “It became popular first online and then
broke into traditional media… A possible origin is Taiwan, as a Taiwanese newspaper voted zan
as Taiwan’s annual character for 2011.” The entry failed to mention that Facebook’s Chinese
version, which conquered Taiwan, translated Like into zan. Also missing was that, in China’s
online universe that blocked Facebook, WeChat introduced a Like button, a heart icon next to
the character zan, to 100 million users in April 2012. It was rolled out with Moments
(pengyouquan, or “friend-circle”) which features updates from contacts, a function crucial to the
messaging app’s expansion into a social networking platform. WeChat’s user base doubled to
200 million in the next five months.
In January 2013, Weibo, China’s (then-)bona fide monopolistic social platform,
introduced Likes using a thumbs-up icon to its 500 million users. Based on various reports on
“Hot Weibo posts” and numerous screenshots from the day, the Like function was not a game
15
changer in user interaction; its counts were typically much smaller than counts of both
Comments and Shares (mostly with comments). A possible explanation is China’s entrenched
blogosphere and BBS traditions, which normalized textual forms of public participation. In fact,
following its own Blog platform, Sina designed Weibo to be text-heavy, allowing commentary to
branch out in complex hierarchical manners, in stark contrast to Twitter. Evidently, the
company set hopes high on the Like button as the external plugin to draw traffic from outside of
Weibo (which did not work as hoped) (Jingyu 2013). Zan found its way into the 2013 Ten
Catchphrases as well, albeit as dian-zan, meaning tapping Likes. This time Yaowenjiaozi (2013)
was unequivocal about its origin: dian-zan “came from the Like function of major social
networks. …Now frequently used in print media, its meaning has changed to indicate assessment
of sorts (dianping). But different from assessment, dian-zan is saying good things only.” (Serious
or tongue-in-cheek, this concluding clarification would become jarring in the discursive climate
of 2014 and onward, after official discourse hijacked the lighthearted phrase for its positive
energy campaign.)
Following this Yaowenjiaozi release, People’s Daily published an editorial “Tap Likes for
China Power,” which attempted to review China’s achievements in 2013 using all the
catchphrases. The editorial regarded many cadres’ existing hostility toward online parlance as
unwarranted. Instead, it called on official language to embrace “net talk.” What was implied is
that the linguistic fusion is instrumental to consensus building, via which “‘China power’ may
consolidate, and the positive energy of reform may spurt.” Dian-zan was the recurring phrase
until the final passage:
In this “wacky” (qipa) age when new things keep emerging, faced with unyielding
collective endeavors to pursue dreams, anecdotes of good Samaritans, and wonders
ranging from online shopping to Big Data, people “tap Likes” (dian-zan) with flying
fingers, joining each other [online] to keep everyone warm. […] While “tapping Likes,”
let us march toward the starting line of the New Year…” (Li 2013)
“Tapping Likes for” (wei…dian-zan) has since become a fixed collocation juxtaposing a
digital neologism with objects associated with state-sanctioned values. Its metaphoric usage
began infesting media and scholarly commentary. Xi Jinping’s (2014) New Year’s Address gave
it a decisive push: “Without people’s support, [our cadres’] work could not have been done so
properly. I tap Likes for our great people.” The last line was taken as the title to his speech in
16
popular media. People’s Daily later named dian-zan one of the buzzwords that Xi brought into
vogue (dai huo, literally “set on fire,” Figure 5): “Xi’s Address received massive Likes-tapping,
which shows his close proximity with the people and the genuine resonance between them”
(Sheng and Wang 2015).
Figure 5. PLA (People’s Liberation Army) Daily’s Weibo account published twelve “Likes-tapping Xi”
cartoons, one for each buzzword he purportedly “set on fire,” annotated with People’s Daily’s original
content (weibo.com/2280198017/C39EGjQyf).
The Xi administration’s expropriation of Like buttons spearheaded its program to “sync”
official talk with online vernaculars (see Wu 2020b). But taking an epistemological-
infrastructural perspective, we may recognize this as a move that alters the legibility of collective
action online. In China, with measures such as real-name registration and binding mobile phone
numbers registered with national IDs, online activity is readily traceable. Stories of being
contacted, warned, and penalized by security agents, often extra-legally, due to web usage
circulate widely. In other words, Chinese online activism entails real bodily risks.
In this context, online participation comes with a spectrum of vulnerability. As police
action against “rumors” and “illegal speech” makes clear, contributing original content is the
most incriminating activity, but sharing is also risky because it indicates rumor mongering intent
(Huang 2017). Relatively speaking, tapping Likes—notably we are now concerned with content
deemed destablizing—is the safest. This spectrum unsurprisingly aligns with platforms’ own
17
governance strategies. During the Covid outbreak, for example, the major social media network
Douban disabled the Share and Comment functions of many posts from Wuhan, but left their
Like buttons intact. Also unsurprisingly, when evidently sentitive content surfaces (e.g.,
oppression of minority groups, high-level corruption, government malpractice, unjust policy), the
numbers on its Share counters may keep growing and those on its Like counters may grow even
faster, but its Comment section remains empty, even though it is still functional. In these
precarious scenes, the Like button provides a shielding architecture akin to barricades and
sunken shelters for street assemblies. It allows people to show up and gesture towards others to
join, while refraining from verbalizing (also see Butler 2015).
As quantum media, Comment and Share counters indicate the amount of attention
content receives; however, the Like metric connotates the volume of kindred sentiments
expressed toward content. The sheer counts of Likes, easily in the hundreds of thousands and
sometimes millions in China’s populated online space, imply the congregation of bodies amassing
under a message. In a country where no images of spontaneous physical assembly are in
circulation, these are stunning scenes to behold—a fascinating case of “statistical panic,” the
harrowing ruptures we feel at the sight of striking numbers presented by new vision techniques
(Woodward 2009). As with historical precedents wherein power sought to explain away the
threatening presence of novel social connections that asserted themselves, the positive energy
discourse churned out the Likes-tapping rhetoric to offset the stressful opacity behind Like
counts.
12
Affective Computing in Chinese Emotionology
On May 19, 2018, more than forty academics and corporate data scientists spoke at the
“New Media Transmission of Positive Energy” symposium in Beijing, which was jointly
sponsored by the Chinese Academy of Social Sciences (CASS) and Weibo. In a major
presentation, CASS researcher Liu Ruisheng provided three scales:
(1) On the macro-level “newness”: The role of new media in the development of Chinese
society amounts to positive energy;
(2) On the meso-level “changes”: In the past two decades, despite constant changes in media
dynamics, what remains unchanged is that as long as we work with the laws of the
12
My analysis draws inspiration from Jones-Imhotep’s (2017) work on the rise of sentimentalism to
account for large gatherings when old social structures were crumbled in the late-18th-century
France.
18
communicative ecology, online mainstream opinion remains a terrain of positive energy;
(3) On the micro-level “details”: Rather than fixate on a few extreme expressions and
jumping to conclusions, we must conduct more scientific analysis of online opinion and
upgrade our strategies of guidance and management, so that new media spread more
positive energy and fulfill people’s aspiration for a good life online. (Sun 2018)
Liu’s speech revealed the hidden truth about positive energy from a governance
perspective. That is, its ontology is contingent on evolving digital infrastructures—ever-new
media, changing communicative ecologies, and persistent “scientific” wrangling over online
happenings. Over the course of the 2010s, the congested ambient politics engendered by
sentiment analysis and the Like button eventually lodged these technologies into the dominant
emotionology that keeps positive energy vital and viral.
On the one hand, the ways in which sentiment analysis gets repackaged and deployed in
yuqing analytics congeal with the broader Chinese imaginary about the crowd. From the 1970s
onward, US academia largely rejected crowd psychology. Sociologists and historians instead were
keen to investigate people’s deliberation in planning, organizing, and exercising self-control in
temporary gatherings (McPhail 2017). But the Chinese trajectory is distinct.
13
Maoist politics
imbued mass gatherings with supreme agency and designated emotional excess as an expression
of moral commitments (Dutton 2016). Drastic negation followed. The postsocialist era imposed
a “scientific” censure of crowd mentality, buttressed by a flurry of foreign monographs (Xiao
2017), like those exhibited on the yuqing practitioner’s desk. It is under this emotionology that
Chinese sentiment analysis has come to scaffold the narrativization of collective action online.
Importantly, sentiment analysis performs this role only when integrated into the peculiar
format of yuqing analytics, which grew out of the interstitial space straddling political, economic,
and academic fields. Uniform data generated by monopolistic platforms enable macroscale
aggregation that dwarfs forceful local activism. Yuqing events are also cauterized once the curves
plateau out, regardless of persistent struggles. The shifting sentimental waves serve as a theatre of
fate. At the resolution of the Total Web, human actors are dissolved; what they consider and
aspire to as they plough through webs of meanings are displaced by crowd psychology narratives.
An institution of management for diverse clients, yuqing analytics strategically interpret
13
The imports of European crowd psychology in the first half of the twentieth century led to the
discovery of the Chinese crowd. Placed at the center of China’s revolutionary struggles, this figure
once channeled conflicting political ideals before the Communists came to power (Xiao 2017).
19
nominally negative sentiment waves. They are sometimes signals of offline unrest, sometimes
indications of opinionated disapproval, and other times a lack of reason. But overall online
activism is reduced to a temporary emotional discharge bound to misfire. The epistemic shift
from “new media event” research to yuqing analytics, I suggest, partly explains what appeared to
be a flourishing of consensus events and a decline of contentious events on the Chinese Internet.
On the other hand, the Chinese government has historically relied on instituted practices
to “weave together multiple and parallel [affective] strands and flows into one big concept” to
produce political outcomes (Dutton 2016, 723; Perry 2002). Defining Like-tapping as emotional
expressions tied to positive energy, in this sense, extends this ethos of governance into technical
infrastructures. What Like metrics imply, accordingly, became as versatile as the content of
positive energy, which seamlessly glides between the feeling of happiness that needs
maximization, giving approval and support, and a proclamation to serve (state-promoted) moral
agendas. This changing signification of Like metrics further led to their integration into various
algorithmic regimes, which in concert facilitates the ascension of China’s new affective
hegemony.
Further, being framed as an act to spread positive energy, “like-tapping” can be analyzed
as a conformist gesture rather than a manifestation of psychology—the focal point of much
existing scholarship on positive energy. As Sara Ahmed poignantly observes, under the veneer of
emotional contagion—in our case the “spread of positive energy”—“it is the objects of emotion
that circulate, rather than emotion as such. [...] Such objects become sticky, or saturated with
affect, as sites of personal and social tension” (Ahmed 2004, 11, also see 2010, chap. 1).
“Tapping Likes for” something thus is a gesture to openly acknowledge that thing as a positive
energy object, be it chicken soup for the soul, health tips, selfless deeds, congratulatory
commentary, GDP growth, military parades, or marvels from ancient dynasties. Accompanying
this has been Weibo’s incorporation of Likes metrics into its curatorial algorithms, which expose
to people what their social networks Like and rank comments by the number of Likes instead of
Replies received—because, notably, “inappropriate or controversial content tend to attract more
Replies” (Weibo 2016). Furthermore, platform channels of state media and government
agencies scramble to promulgate a hodgepodge of “positive energy” clickbait to cumulate more
Likes for performance evaluation (Lu and Pan 2021).
14
Through affectively subsuming Like
14
The main index for their performance is Communicative Power (chuanboli), a phrase borrowed
from Xi’s 2013 speech to urge government agencies expand their propaganda work on social media.
It assigns heavy weights to the number of Likes.
20
buttons, the positive energy imperative fosters a massive choreography on social media where
people seek out and tap particular content, in concert with algorithmic augmentation of its
exposure. This distributed deployment of affective computing, I argue, operates alongside the
disciplinary measures over content creation and distribution in Chinese Internet governance (see
G. Yang 2017).
Chinese sentiment analysis and Like buttons have also jointly ushered emotionological
change that predisposes people to engagement with preexisting semantic objects, rather than
creating their own. Online content faces a divide-and-rule strategy with regard to its associations
with dominant sociocultural and political norms. The so-called “clean and bright (qinglang)
cyberspace” that official documents and speech promote consists of feel-good, positive energy
content, which verifies those who “tap Likes for” it and simultaneously gets heightened and
disseminated by Like-tapping. Folded underneath the “clean and bright cyberspace” is the
“negative energy” content, which attracts “web mobs” subject to wave dynamics; its
multiplication, however staggering, is momentary and bound to wear out. Both Like-tapping
conformism and yuqing-sentiment analytics work to hamper ordinary users’ linguistic
expression. Attuned to display the intensity of people’s interaction with a priori delimited
content, neither renders visible the fraught process of semantic exchange, articulation,
complication, and transgression. In particular, yuqing analytics invite people to hover above the
context from “on high,” effectively installing an “ironic point of view” that denies any generative
potential in linguistic representation (Colebrook 2004).
15
These latent laws of optics accord with
the interest in stability and profits shared by Chinese government, platforms, and institutional
content providers. They together summon, both epistemically and behaviorally, an online
universe diffused with positive energy. This provides a substantial explanation for the widely
observed conservative turn of Chinese online cultures since the mid-2010s (G. Yang 2017; Wu
2020b).
Conclusion
The Chinese case I present illustrates a general research strategy to investigate affective
computing—at the level of the emotionological, with a focus on its ambient politics.
Emotionology refers to the symbolic-material infrastructures via which society recognizes and
normalizes emotion displays. The concept aids researchers to heed the professed norms and
15
This also sets yuqing waves apart from hashtag activism, whose constant supply for intertextual
linkages is a source for transformative political energies (Bonilla and Rosa 2015).
21
“broader affective logic” that shape any inquiry—technical, critical, and socio-anthropological—
into the realm of human emotion (Dror et al. 2016, 11). Behind both computing’s prevalent
adoption of “inside-out” behavioralist models of mental activity (also see Binder 2020), on one
hand, and its critics’ cherishment of individual sovereignty over the interior as a sheltered space
(see Amrute 2020 for a review of Zuboff [2019]), on the other, is one peculiar emotionology
that upholds a liberal subject with innate emotion. In addition, that these technologies can access
interior workings, on which much of their critiques rest, is also aggressively promoted by
corporate and political powers wielding them in self-interest. Attending to emotionology, in this
sense, may productively reorient our analysis away from a preoccupation with individual subjects
or disciplinary bodies to the construction of conditions of collective existence (also see Brown
2015, 73–78).
A parallel analytical shift is from affective computing’s design epistemics to its ambient
politics. While it is indispensable to delineate the values that are built into technologies, the
social consequences of these technologies hinge on whether and how they become embedded
into local emotionological infrastructures. In this fraught process, affective computing is seized
upon as quantum media with emotional polysemy. From a vast range of behavioral traces to a
standard set of discrete emotions, it renders the most bewildering and ambivalent into the most
comprehensible. But it does not end here. Affective computing produces output that is also the
most accommodating for a variety of social actors to recalibrate, reinterpret, valorize, and
narrativize as part of their own projects. The ambient politics of affective computing, therefore, is
one of infusion and amalgamation, distinct from what has been noted as the growing “pressure to
simplify emotion’s measure or monitoring so it can become machine-readable and machine-
expressible” (Pasquale 2020, 214). It is from these maneuvers and contestations that
emotionological shifts may arise, which in turn have an impact on our imaginaries and social
behaviors beyond the realm of emotion. By repackaging and remediating calculative
infrastructures that denote emotions, the ambient politics of affective computing serves a central
nexus linking, on one side, institutions of political and economic governance, and on the other, a
social world increasingly exposing itself to datafication.
Angela Xiao Wu is an assistant professor in media, culture, and communication at New York
University. She investigates the connections between media technologies, knowledge
production, and politics. Her current book project examines how public culture takes shape
when systems thinking informs its conception and governance.
22
References
Ahmed, Sara. 2004. The Cultural Politics of Emotion. New York: Routledge.
Ahmed, Sara. 2010. The Promise of Happiness. Durham, NC: Duke University Press.
Alder, Ken. 2002. “A Social History of Untruth: Lie Detection and Trust in Twentieth-Century
America.” Representations 80, no. 1: 1–33.
Amrute, Sareeta. 2020. “Sounding the Flat Alarm.” Boundary 2. January 27.
www.boundary2.org/2020/01/sareeta-amrute-sounding-the-flat-alarm-review-of-
shoshana-zuboff-the-age-of-surveillance-capitalism/.
Andrejevic, Mark. 2013. Infoglut: How Too Much Information Is Changing the Way We Think
and Know. New York: Routledge.
Appadurai, Arjun, ed. 1988. The Social Life of Things: Commodities in Cultural Perspective.
Cambridge: Cambridge University Press.
Arvidsson, Adam. 2011. “General Sentiment: How Value and Affect Converge in the
Information Economy.” The Sociological Review 59, no. s2: 39–59.
Auerbach, David. 2019. Bitwise: A Life in Code. Vintage.
Bakir, Vian, and Andrew McStay. 2018. “Fake News and The Economy of Emotions.” Digital
Journalism 6, no. 2: 154–75.
Batke, Jessica, and Mareike Ohlberg. 2020. “Message Control: How a New For-Profit Industry
Helps China’s Leaders ‘Manage Public Opinion.’” ChinaFile. December 20.
www.chinafile.com/reporting-opinion/features/message-control-china.
Bennett, W. Lance, and Alexandra Segerberg. 2012. “The Logic of Connective Action.”
Information, Communication and Society 15, no. 5: 739–68.
Binder, Jeffrey M. 2020. “The 18th-Century Elocution Movement and Facebook: Reading
Emotion before and after the Subject.” Media Culture & Society 42, no. 5: 762–76.
Bonilla, Yarimar, and Jonathan Rosa. 2015. “#Ferguson: Digital Protest, Hashtag Ethnography,
and the Racial Politics of Social Media in the United States: #Ferguson.” American
Ethnologist 42, no. 1: 4–17.
Brown, Wendy. 2015. Undoing the Demos: Neoliberalism’s Stealth Revolution. Cambridge: MIT
Press.
Butler, Judith. 2015. Notes Toward a Performative Theory of Assembly. Harvard University
Press.
Chen, Zifeng, and Clyde Yicheng Wang. 2020. “The Discipline of Happiness: The Foucauldian
Use of the ‘Positive Energy’ Discourse in China’s Ideological Works.” Journal of Current
Chinese Affairs 48, no. 2: 201–25.
Colebrook, Claire. 2004. Irony. Hove, UK: Psychology Press.
Couldry, Nick, and Ulises A. Mejias. 2019. The Costs of Connection: How Data Is Colonizing
Human Life and Appropriating It for Capitalism. Palo Alto: Stanford University Press.
Crawford, Kate, Roel Dobbe, Theodora Dryer, Genevieve Fried, Ben Green, Elizabeth
Kaziunas, Amba Kak, et al. 2019. “AI Now 2019 Report.” AI Now Institute.
ainowinstitute.org/AI_Now_2019_Report.html.
Crawford, Kate, Jessa Lingel, and Tero Karppi. 2015. “Our Metrics, Ourselves: A Hundred
Years of Self-Tracking from the Weight Scale to the Wrist Wearable Device.” European
Journal of Cultural Studies 18, no. 4–5: 479–96.
Dayan, Daniel, and Elihu Katz. 1992. Media Events: The Live Broadcasting of History.
Cambridge: Harvard University Press.
Davies, William. 2017. “How Are We Now? Real-Time Mood-Monitoring as Valuation.”
Journal of Cultural Economy 10, no. 1: 34–48.
Dean, Jodi. 2010. Blog Theory: Feedback and Capture in the Circuits of Drive. Cambridge: Polity.
Dror, Otniel E. 2011. “Seeing the Blush: Feeling Emotions.” In Histories of Scientific
23
Observation, edited by Lorraine Daston and Elizabeth Lunbeck, 326–348. Chicago, IL:
University of Chicago Press.
Dror, Otniel E., Bettina Hitzer, Anja Laukötter, and Pilar León-Sanz. 2016. “An Introduction to
History of Science and the Emotions.” Osiris 31, no. 1: 1–18.
Dutton, Michael. 2016. “Cultural Revolution as Method.” The China Quarterly 227, September:
718–33.
Eslami, Motahhare, Karrie G. Karahalios, Christian Sandvig, Kristen Vaccaro, Aimee Rickman,
Kevin Hamilton, and Alex Kirlik. 2016. “First I like It, Then I Hide It: Folk Theories
of Social Feeds.” In Proceedings of the 2016 CHI Conference on Human Factors in
Computing Systems, 2371–2382.
Gehl, Robert W. 2014. Reverse Engineering Social Media: Software, Culture, and Political
Economy in New Media Capitalism. Philadelphia: Temple University Press.
Gerlitz, Carolin, and Anne Helmond. 2013. “The like Economy: Social Buttons and the Data-
Intensive Web.” New Media & Society 15, no. 8: 1348–65.
Gross, Daniel M. 2007. The Secret History of Emotion: From Aristotle’s Rhetoric to Modern Brain
Science. Chicago: University of Chicago Press.
Helmreich, Stefan. 2020. “Wave Theory ~ Social Theory.” Public Culture 32, no. 2: 287–326.
Herbst, Susan. 1998. Reading Public Opinion: How Political Actors View the Democratic Process.
Chicago: University of Chicago Press.
Hindman, Matthew. 2018. “How Cambridge Analytica’s Facebook Targeting Model Really
Worked – According to the Person Who Built It.” The Conversation, March 30.
theconversation.com/how-cambridge-analyticas-facebook-targeting-model-really-worked-
according-to-the-person-who-built-it-94078.
Hird, Derek. 2018. “Smile Yourself Happy.” In Chinese Discourses on Happiness, edited by
Gerda Wielander and Derek Hird, 106–28. Hong Kong: Hong Kong University Press.
Huang, Haifeng. 2017. “A War of (Mis)Information: The Political Effects of Rumors and
Rumor Rebuttals in an Authoritarian Country.” British Journal of Political Science 47, no.
2: 283–311.
Humphreys, Lee. 2018. The Qualified Self: Social Media and the Accounting of Everyday Life.
Cambridge: MIT Press.
Hussain, Amir, and Erik Cambria. 2015. “‘Socio-Affective Computing’ Series.” Springer.
www.springer.com/series/13199.
Illouz, Eva. 2007. Cold Intimacies: The Making of Emotional Capitalism. Polity.
Jasanoff, Sheila. 2015. “Future Imperfect: Science, Technology, and the Imaginations of
Modernity.” In Dreamscapes of Modernity: Sociotechnical Imaginaries and the Fabrication
of Power, edited by Sheila Jasanoff and Sang-Hyun Kim, 1–33. Chicago: University of
Chicago Press.
Jingyu. 2013. “Xinban weibo kehuduan: Shi shihou dian ‘zan’ le (Weibo upgraded: it’s time to
tap Likes).” 36Kr. June 28, 2013. www.36kr.com/p/1641746841601.
Jones-Imhotep, Edward. 2017. “The Unfailing Machine: Mechanical Arts, Sentimental Publics
and the Guillotine in Revolutionary France.” History of the Human Sciences 30, no. 4:
11–31.
Karpf, David. 2016. Analytic Activism: Digital Listening and the New Political Strategy. Oxford:
Oxford University Press.
Kotras, Baptiste. 2020. “Opinions That Matter: The Hybridization of Opinion and Reputation
Measurement in Social Media Listening Software.” Media Culture & Society 42, no. 7–8:
1495–1511.
Kramer, Adam D. I., Jamie E. Guillory, and Jeffrey T. Hancock. 2014. “Experimental Evidence
of Massive-Scale Emotional Contagion through Social Networks.” Proceedings of the
24
National Academy of Sciences of the United States of America 111, no. 24: 8788–90.
Kreiss, Daniel, and Shannon C. Mcgregor. 2018. “Technology Firms Shape Political
Communication: The Work of Microsoft, Facebook, Twitter, and Google With
Campaigns During the 2016 U.S. Presidential Cycle.” Political Communication 35, no. 2:
155–77.
Leys, Ruth. 2017. The Ascent of Affect: Genealogy and Critique. Chicago: University of Chicago
Press.
Li, Hongbing. 2013. “2013 reci: wei zhongguo liliang ‘dianzan’ (Buzzwords 2013: Tap-Likes for
China power).” People’s Daily, December 20.
Lu, Yingdan, and Jennifer Pan. 2021. “Capturing Clicks: How the Chinese Government Uses
Clickbait to Compete for Visibility.” Political Communication 38, no. 1-2: 23–54.
McPhail, Clark. 2017. The Myth of the Madding Crowd. New York: Routledge.
Pan, Jennifer, and Kaiping Chen. 2018. “Concealing Corruption: How Chinese Officials Distort
Upward Reporting of Online Grievances.” American Political Science Review 112, no. 3:
602–20.
Papacharissi, Zizi. 2015. Affective Publics: Sentiment, Technology, and Politics. Oxford: Oxford
University Press.
Pasquale, Frank. 2020. New Laws of Robotics: Defending Human Expertise in the Age of AI.
Cambridge, MA: Harvard University Press.
Perry, Elizabeth J. 2002. “Moving the Masses: Emotion Work in the Chinese Revolution.”
Mobilization 7, no. 2: 111–28.
Qiu, Linchuan, and Taowen Chen, eds. 2011. Xinmeiti shijian yanjiu (New Media Event
Research). Beijing: Renmin University Press.
Rancière, Jacques. 2013. The Politics of Aesthetics. London: Bloomsbury.
Reddy, William M. 1997. “Against Constructionism: The Historical Ethnography of Emotions.”
Current Anthropology 38, no. 3: 327–51.
Richaud, Lisa, and Ash Amin. 2020. “Life Amidst Rubble: Migrant Mental Health and the
Management of Subjectivity in Urban China.” Public Culture 32, no. 1: 77–106.
Schradie, Jen. 2019. The Revolution That Wasn’t: How Digital Activism Favors Conservatives.
Cambridge, MA: Harvard University Press.
Schüll, Natasha Dow. 2016. “Abiding Chance: Online Poker and the Software of Self-
Discipline.” Public Culture 28, no. 3: 563–92.
Sheng, Hui, and Simin Wang. 2015. “Zheliangnian, Xi Jinping daihuo de 12 ge reci (Positive
energy, the 12 phrases Xi Jinping set on fire).” People.Cn. February 5, 2015.
politics.people.com.cn/n/2015/0206/c99014-26519924-10.html.
SMP2020-EWECT. 2020. “The Evaluation of Weibo Emotion Classification Technology.” The
9th China National Conference on Social Media Processing, September.
smp2020ewect.github.io/.
Stark, Luke, and Jesse Hoey. 2021. “The Ethics of Emotion in AI Systems.” In Proceedings of
ACM Conference on Fairness, Accountability, and Transparency (FAccT’21), Virtual
Event, March.
Stearns, Peter N., and Carol Z. Stearns. 1985. “Emotionology: Clarifying the History of
Emotions and Emotional Standards.” American Historical Review 90, no. 4: 813–36.
Stroud, Natalie Jomini, Ashley Muddiman, and Joshua M. Scacco. 2017. “Like, Recommend, or
Respect? Altering Political Behavior in News Comment Sections.” New Media & Society
19, no. 11: 1727–43.
Sun, Long. 2018. “‘Xinmeiti zhengnengliang chuanbo’ yantaohui zaijing juxing (‘New Media
Transmission of Positive Energy’ Symposium Held in Beijing).” CASS. May 19, 2018.
news.cssn.cn/zx/bwyc/201805/t20180519_4265147.shtml.
25
Wagner-Pacifici, Robin. 2017. What Is an Event? Chicago, IL: University of Chicago Press.
Wang, Yuhua, and Carl Minzner. 2015. “The Rise of the Chinese Security State.” The China
Quarterly 222, June: 339–59.
Weibo. 2016. “User Guide on the New Commenting Function 2.0.” October 20, 2016.
https://weibo.com/ttarticle/p/show?id=2309404032707440077749.
Wernimont, Jacqueline. 2019. Numbered Lives: Life and Death in Quantum Media. Cambridge,
MA: MIT Press.
Williamson, Ben. 2017. “Moulding Student Emotions through Computational Psychology:
Affective Learning Technologies and Algorithmic Governance.” Educational Media
International 54, no. 4: 267–88.
Woodward, Kathleen. 2009. Statistical Panic: Cultural Politics and Poetics of the Emotions.
Durham: Duke University Press.
Wright, Alex. 2009. “Mining the Web for Feelings, Not Facts.” New York Times, August 23.
Wu, Angela Xiao. 2020a. “Chinese Computing and Computing China as Global Knowledge
Production.” Catalyst: Feminism, Theory, and Technoscience 6, no. 2.
Wu, Angela Xiao. 2020b. “The Evolution of Regime Imaginaries on the Chinese Internet.”
Journal of Political Ideologies 25, no. 2: 139–61.
Xi, Jinping. 2014. “Guojia zhuxi Xi Jinping fabiao 2015 nian xinnian heci (New Year’s Eve
Address by Chairman Xi).” XinhuaNet. December 31.
www.xinhuanet.com//politics/2014-12/31/c_1113846581.htm.
Xiao, Tie. 2017. Revolutionary Waves: The Crowd in Modern China. Cambridge, MA: Harvard
University Asia Center.
Yang, Guobin. 2016. “Activism.” In Digital Keywords, edited by Benjamin Peters, 1–17.
Princeton, NJ: Princeton University Press.
Yang, Guobin. 2017. “Killing Emotions Softly: The Civilizing Process of Online Emotional
Mobilization.” Communication & Society 40: 75–104.
Yang, Peidong, and Lijun Tang. 2018. “‘Positive Energy’: Hegemonic Intervention and Online
Media Discourse in China’s Xi Jinping Era.” China: An International Journal 16, no. 1:
1–22.
Yaowenjiaozi Editorial Staff. 2012. “2012 nian Shida Liuxingyu (Top ten catchphrases of
2012).” Yaowenjiaozi, no. 2: 4–6.
Yaowenjiaozi Editorial Staff. 2013. “2013 nian Shida Liuxingyu (Top ten catchphrases of
2013).” Yaowenjiaozi, no. 2: 4–6.
Zhang, Li. 2015. “Psychotherapy, Spirituality, and Well-Being in a Transforming Urban China.”
In Handbook of Religion and the Asian City, edited by Peter van der Veer, 315–32.
Berkeley, CA: University of California Press.
Zhu, Huaxin, Jiangchun Hu, and Wentao Sun. 2008. “2007 nian zhongguo hulianwang yuqing
fenxi baogao (Chinese Internet yuqing report, 2007)” Jinchuanmei, no. 2: 31–40.
Zuboff, Shoshana. 2019. The Age of Surveillance Capitalism: The Fight for a Human Future at
the New Frontier of Power. London: Profile Books.
