|
|
ENQUIRE PROJECT DETAILS BY GENERAL PUBLIC |
| Project Details |
| Funding Scheme : | General Research Fund | ||||||||||||||||||||||||||||||||
| Project Number : | 12605520 | ||||||||||||||||||||||||||||||||
| Project Title(English) : | Mapping the Status of Political Incivility in Hong Kong's Digital Space: From the Umbrella Movement to the Anti-Extradition Protests | ||||||||||||||||||||||||||||||||
| Project Title(Chinese) : | 勾勒香港網絡不文明的立體面貌：從雨傘運動到反修例抗爭 | ||||||||||||||||||||||||||||||||
| Principal Investigator(English) : | Prof Song, Yunya | ||||||||||||||||||||||||||||||||
| Principal Investigator(Chinese) : | |||||||||||||||||||||||||||||||||
| Department : | Division of Emerging Interdisciplinary Areas | ||||||||||||||||||||||||||||||||
| Institution : | Hong Kong Baptist University | ||||||||||||||||||||||||||||||||
| E-mail Address : | yunyasong@hkbu.edu.hk | ||||||||||||||||||||||||||||||||
| Tel : | |||||||||||||||||||||||||||||||||
| Co - Investigator(s) : |
|
||||||||||||||||||||||||||||||||
| Panel : | Humanities, Social Sciences | ||||||||||||||||||||||||||||||||
| Subject Area : | Humanities and Arts | ||||||||||||||||||||||||||||||||
| Exercise Year : | 2020 / 21 | ||||||||||||||||||||||||||||||||
| Fund Approved : | 396,720 | ||||||||||||||||||||||||||||||||
| Project Status : | Completed | ||||||||||||||||||||||||||||||||
| Completion Date : | 30-6-2023 | ||||||||||||||||||||||||||||||||
| Project Objectives : |
|
||||||||||||||||||||||||||||||||
| Abstract as per original application (English/Chinese): |
From the 2014 Umbrella Movement to the ongoing anti-extradition bill movement, an increasing number of Hong Kong Internet users have adopted a style of behavior that is abusive and hostile, partially an expression of frustration with what is seen as the government's intransigent disregard for protesters' appeals. This online incivility—which broadly ranges from aggressive comments to hate speech—may possibly corrode reasoned discourse in cyberspace, fuel social climates of hate in public debate, and detract from the broader struggle for democracy and autonomy. Focused mostly on the U.S. or European context, however, prevailing research leaves unexamined uncivil behaviors in Hong Kong's online communities in general and specifically during the recent period. To help overcome these analytical and empirical gaps in our scholarly and practical knowledge, this project investigates in-depth the phenomenon of political incivility in Hong Kong's digital space from a longitudinal perspective (2014-2019). It devises an empirically clearer definition of online incivility, more precisely examines how and why a changing social atmosphere affects uncivil discourse, and more comprehensively delineates practical measures to tackle uncivil online behavior.
Procedurally, this project first develops a more empirically based and methodologically defensible definition of online incivility. We will conduct in-depth interviews with a broad sample of Hong Kong's Internet forum users to devise a community definition of incivility. In our view, research on incivility should consider the perceptual elements of incivility from a user-centered point of view, rather than by arbitrarily applying politeness standards that risk restricting the speech of certain social groups.
Next, computational approaches will be combined with qualitative discourse analysis to investigate the incivility in political discussion on Hong Kong's three most important Internet forums. The aim is to reveal the social and behavioral mechanisms through which the instances of uncivil behaviors at the individual level might possibly generate a self-perpetuating process of ever-growing uncivil and aggressive group behaviors. Using large-scale datasets of uncivil online speech collected from these Internet forums, this project will use a set of advanced data mining techniques to extract behavioral patterns on social media. The broader goal of this project is to understand the dynamics and societal effects of negative discourses and online hostile speech environments.
Finally, the analytical and empirical results of this project will be used to devise more effective mechanisms that help minimize the harm of online uncivil behavior while still allowing for the free flow information and facilitating diverse opinions and political tolerance. 從2014年的雨傘運動到現下的反修例抗爭，香港互聯網上的辱罵或敵意性行為甚囂塵上。造成這種現象的部分原因是人們認為政府沒有回應抗爭者的主張，因而通過這些方式來表達失望不滿。網絡不文明行為涵蓋了攻擊性評論和仇恨言論。這種行為會侵蝕網絡上的理性話語，給公共討論造成敵對氣氛，甚至可能會偏離民主之路。儘管美國、歐洲情境下的網絡不文明行為已經獲得了相當豐富的研究成果，但是有關於香港網絡不文明行為的研究卻明顯不足。本研究旨在填補相關的學術和實踐知識空白，採用縱向視角深入分析2014-2019年間香港網絡上的不文明行為。研究者基於經驗性證據對網絡不文明行為作出更清晰的界定，檢視動盪的社會氣氛如何影響不文明話語，並提出更全面的應對方法。 首先，本研究對網絡不文明行為給出的定義，著眼於實證及方法論的堅實基礎。研究者對香港的網絡論壇用戶進行深度訪談，藉以從用戶社群的角度提出對不文明行為的定義。網絡不文明行為研究應該從用戶感知角度出發，而不是機械地套用禮貌準則，生硬照搬可能會限制某些社會群體的發聲。 其次，本研究結合計算社會科學方法以及質性的話語分析法，對香港各大網絡論壇上的不文明行為進行調查分析，旨在揭示個體層面的不文明行為如何不斷衍生攻擊性群體行為的社會機制及行為機制。研究者應用一系列先進的數據挖掘方法，從網絡論壇上海量的不文明言論數據中挖掘出隱含的行為模式。本研究的目標是理解包括消極言論和網絡敵對言論環境的生態及其社會影響。 最後，本研究基於實證分析結果提出應對網絡不文明的有效機制，這些機制在保證資訊自由流動、意見多元及政治容忍的前提下，將網絡不文明的危害減至最低。 |
||||||||||||||||||||||||||||||||
| Realisation of objectives: | We have successfully accomplished all objectives outlined previously. To define online incivility from a Hong Kong user-centered point of view, we conducted interviews with online forum users in Hong Kong. The study delved into the intolerance and toxicity of specific uncivil behaviors. In an endeavor to categorize types of uncivil online speech, we used a 5-Likert Scale during the in-depth interviews to evaluate participants’ attitudes towards various uncivil behaviors, encompassing the degree of incivility perceived, the inclination to disengage from conversations upon encountering uncivil remarks, and the propensity to refrain from sharing content in response to such instances. Moreover, participants were asked to recount other instances of online incivility they had encountered, which were evaluated using the same criteria. Through this methodological approach, we systematically collected and synthesized prevalent forms of online incivility (Objective 1 & 2). To further map the status of political incivility in Hong Kong's online spaces, we undertook a computational analysis of a vast dataset of posts and comments on well-known online discussion forums. Specifically, we documented the use of explicit and implicit incivility during significant events like the Umbrella Movement and the Anti-ELAB Movement. Surveys were utilized to gauge the prevalence and types of incivility, along with the effectiveness of different intervention strategies (Objective 1 & 2). We documented interviewees' reactions to various forum threads through self-confrontation interviews, exploring the reasoning behind their unique perceptions of online incivility. A user-enhanced automatic coding method was employed for data analysis, offering a detailed insight into the online behaviors and perceptions of incivility. Interviewees' interpretations were found to be context-dependent and varied according to the discussion's nature and the involved participants. This variability underscored the subjective nature of online incivility, highlighting the importance of considering individual differences and the context in which interactions occur (Objective 2). We built a dictionary of uncivil expressions, which enabled us to develop automated coding to identify uncivil online speech. We began with extensive data collection from online forums to gather a robust sample of online conversations. These conversations were manually screened for potential uncivil language, which, upon validation by linguistic and cultural experts, formed the basis of a comprehensive dictionary. This dictionary then enabled the creation of an automated algorithm trained to detect uncivil expressions (Objective 2 & 3). Utilizing these analytic approaches, we leveraged computational and content analysis to track linguistic and behavioral changes over time. We carried out an extensive investigation into online incivility and its effects, specifically in the milieu of the Umbrella Movement and the Anti-ELAB Movement, within various online forums. Specifically, we examined how explicit and implicit uncivil expressions vary in discussions about the two movements, influencing deliberative discussion, uncivil reactions, and thread popularity. In terms of contexts, findings revealed that norms observed during different social protests moderate the correlation between deliberative discourse and incivility in digital communication. The emphasis on "rational and non-violent" behavior during the 2014 Umbrella protests fostered norms of civility and politeness, promoting the use of implicit incivility for constructive exchanges. Conversely, the Anti-ELAB protests saw the widespread adoption of aggressive tactics as effective norms. Implicit incivility declined as explicit incivility became more common. As such, we mapped the contexts and triggers of online uncivil behavior (Objective 3). To identify social and behavioral mechanisms underlying the individual-level manifestation of online incivility, the study investigated the contagion of offensive speech online within Hong Kong's forums, focusing on four social interactional mechanisms: generalized reciprocity, direct reciprocity, leader-mimicry, and peer-mimicry. Using the data from online forums covering original posts and user comments, a social network analytic approach was employed to examine these mechanisms' effects on political incivility. Interview results demonstrate that offensive speech influenced not only the targeted users but also bystanders, creating a collective discursive culture. Findings suggest that social norms in online communities emerge from the accumulation of micro-social interactions and adapt to shifting discussion dynamics. The contagion of offensive speech is thus driven by both direct and indirect social interactions within the forums (Objective 4). A set of computational approaches were utilized in this project. We proposed an online processing system to perform incivility detection and sentiment prediction for short texts. Its workflow involves collecting social media posts, preprocessing the data for analysis, using CNN-LSTM models to detect incivility and classify sentiment. When utilizing social media data to examine how disagreement can lead to incivility, we used Python to automate data collection and analysis, including sentiment analysis with VADER and network analysis with NetworkX. Time-series data were aggregated at three-day intervals to examine fluctuations in thread popularity, incivility, and identity expressions over time. In our analysis of sentiment within Cantonese political posts from online forums, we applied a spectrum of methods from dictionary-based to sophisticated machine learning models, including fine-tuned BERT and mBERT (Objective 5). We supplemented these computational methods with qualitative approaches to examine the spread of online incivility and to understand the factors influencing user intervention, or lack thereof, in uncivil online behaviors, utilizing the Bystander Intervention Model (BIM). We also explored how group identity and participant roles affect responses to incivility. Through interviews with Hong Kong forum users, we identified factors such as disinterest in topics, failure to recognize the severity of incivility, and a perceived absence of responsibility as reasons for non-intervention. We also found evidence of in-group favoritism and out-group discrimination, consistent with the Social Identity Model of Deindividuation Effects (SIDE), manifesting in more assertive responses to out-group incivility and a tolerance or endorsement of in-group incivility. This qualitative research underscores the importance of understanding bystander intervention dynamics and group identity to effectively address online incivility (Objective 5). To investigate the dynamics and social consequences of negative discourse, aggression, and hostility on Hong Kong's social media platforms, we leveraged panel survey data and social media analytics to study how disagreements escalate into incivility, including both exposure and personal expression of uncivil behavior. Our findings from a two-wave panel survey revealed that cross-cutting exposure initially increases polarization by amplifying exposure to uncivil messages and the expression of uncivil opinions, which in turn provoke negative emotions. Interestingly, cross-cutting exposure can also have an indirect depolarizing effect by promoting the expression of uncivil opinions and subsequently eliciting positive emotions, emphasizing the significance of active expression. Further, through computational content analysis and regression models, the study demonstrated that cross-cutting exposure intensifies both the encounter with and the expression of uncivil messages, which are predictive of political polarization. Additionally, we identified effective strategies for mitigating online incivility, including proactive and reactive moderation, pre-screening, and user contributions (Objective 6). We have produced high-quality papers for top-tier journal publications. One paper has been successfully published in Computer in Human Behavior (Impact Factor 9.9). Another paper is under review of Internet Research (Impact Factor 7.9) and has received a "minor revision" decision. Additionally, a paper submitted to Communication Research, one of the top two journals in the field of Communication, has also received the "minor revision" decision. We have also presented a series of papers at prestigious conferences in our field. In addition to the papers listed in Research Outputs, we have recently had another paper accepted by the esteemed AEJMC, which we will be presenting this coming August (Objective 7). | ||||||||||||||||||||||||||||||||
| Summary of objectives addressed: |
|
||||||||||||||||||||||||||||||||
| Research Outcome | |||||||||||||||||||||||||||||||||
| Major findings and research outcome: | This project charted the landscape of online incivility in Hong Kong's digital spaces. A longitudinal analysis of online forum data uncovered a rising tolerance for political incivility. Investigation into various social contagion mechanisms suggested that peer imitation is a predominant factor in the proliferation of uncivil behaviors, supplemented by generalized reciprocity and direct imitation. A further computational and quantitative inquiry into the dynamics of online incivility during social movements categorized it into implicit and explicit forms, assessing their effects on democratic discourse. Results indicate that while implicit incivility may foster deliberative discussions, explicit incivility tends to lead to a “leader mimicry” contagion effect. Conversely, implicit incivility appears to deter further uncivil responses, demonstrating its nuanced influence on online conversations. Furthermore, the study probed into how the social norms of movements shape uncivil behavior, uncovering varying effects depending on the characteristics of the movement. Methodologically, a comprehensive Cantonese profanity lexicon was developed to enhance the classification of different types of incivility and enable automated detection. In addition, panel survey data and computational analysis of online forum data were used to study the escalation of disagreements into incivility. Results showed that cross-cutting exposure initially increases polarization by exposing users to uncivil messages and opinions, which evoke negative emotions. However, this exposure can later reduce polarization by fostering the expression of uncivil opinions and subsequently eliciting positive emotions. In-depth interviews with Hong Kongers explored their perceptions of online incivility. The research revealed significant variations in how uncivil comments and their modes of presentation are perceived, with textual incivility in comments being particularly influential on forum participants. By applying the Bystander Intervention Model, the study identified several factors that influence users' reactions to online incivility, including their sense of responsibility, persuasive ability, and considerations of the costs and benefits of responding, as well as the influence of identity roles. Moreover, the research corroborates the Social Identity model of Deindividuation Effects, highlighting in-group bias and out-group prejudice in reactions to incivility. The interviews further explored the intolerance and toxicity associated with specific uncivil behaviors, pinpointing doxing, cyberbullying, discrimination, and hate speech as particularly egregious. It identified a range of motives behind such uncivil actions, including identity-driven, aggressive, counteractive, affective, and culturally proposed motives, with counteractive motives being most common. According to interviewees, strategies considered effective in mitigating online incivility encompass a range of proactive, reactive, and user-interactive moderation measures. | ||||||||||||||||||||||||||||||||
| Potential for further development of the research and the proposed course of action: |
Expanding the scope of this project beyond its initial Chinese context promises to unlock a broad spectrum for academic inquiry and practical implementation. A systematic study that spans diverse national domains and social media platforms will enrich our comprehension of online incivility and bolster the rigor of comparative research in digital communication. Additionally, by harnessing the synergy between GPT technology and Natural Language Processing, future research can refine the tools for detecting subtle shades of uncivil discourse. Specifically, grounded in a consensus on incivility categories identified by interviewees, future research can utilize the GPT model to expand the data corpus, followed by training machine learning models to discern levels of incivility within this dataset. These methodological strides hold the potential to significantly enhance strategies for moderating online behavior. In tandem with these technological advancements, our project advocates for the establishment of reputation systems on digital platforms. These systems will incentivize positive interactions while discouraging negative conduct, thereby encouraging users to uphold standards of respect and civility. By fostering a self-regulating community that values constructive dialogue, we aspire to create a more inclusive and harmonious online environment that celebrates diversity and fosters meaningful interactions. | ||||||||||||||||||||||||||||||||
| Layman's Summary of Completion Report: | This research project aims to understand and map online incivility in Hong Kong, especially during major social movements. Through analyzing online forum data, surveys, and user interviews, the study seeks to examine the nature, causes, and impacts of uncivil behavior. The research defines political incivility from the user's perspective, identifying behaviors deemed disruptive. This user-centered approach aids in creating automated methods to detect incivility, enhancing social media conversation metrics for more meaningful discussions. The study also explores the dynamics and societal impacts of hostile online discourse, aiming to create sustainable solutions to foster healthier online discourse. Key findings include the role of supportive expressions in reducing incivility and promoting positive engagement. Analysis of forum posts and social network data reveals how social dynamics, like peer mimicry, amplify uncivil discourse. Interviews and surveys highlight diverse motivations and perceptions of incivility influenced by context and individual differences. By integrating computational approaches with qualitative methods, the research provides a comprehensive understanding of online incivility's nature and effects. This research contributes valuable insights into mitigating online hostility, fostering productive social media conversations, and enhancing the integrity of participatory politics. | ||||||||||||||||||||||||||||||||
| Research Output | |||||||||||||||||||||||||||||||||
| Peer-reviewed journal publication(s) arising directly from this research project : (* denotes the corresponding author) |
|
||||||||||||||||||||||||||||||||
| Recognized international conference(s) in which paper(s) related to this research project was/were delivered : |
|
||||||||||||||||||||||||||||||||
| Other impact (e.g. award of patents or prizes, collaboration with other research institutions, technology transfer, etc.): |
|||||||||||||||||||||||||||||||||
| SCREEN ID: SCRRM00542 |