Cody Buntain, Yelena Mejova, Diyi Yang
The pre-conference day included a wide array of workshops and tutorials, spanning a range of topics. The tutorials covered the latest techniques in machine learning (including deep learning and BERT), information extraction, causal inference, word embeddings, and the use of Twitter API v2, and addressed use cases including mis/disinformation and business decision making. The workshops included those on Cyber Social Threats (CySoc), Social Sensing (SocialSens): Special Edition on Belief Dynamics, Images in Online Political Communication (PhoMemes), Novel Evaluation Approaches for Text Classification Systems on Social Media (NEATCLasS), Social Media for Emergency Response (SoMER), Data for the Wellbeing of Most Vulnerable, and News Media and Computational Journalism (MEDIATE). A Data Challenge was also held on this day, with a special focus on Health-Related Discourse on the Web. The proceedings of the workshops are availableat http://workshop-proceedings.icwsm.org/index.php?year=2022.
For the main conference, 454 reviewers and 86 senior PC members evaluated 455 papers submitted to the conference, with 122 being accepted for publication. Out of these, 20 described a publicly available dataset, supporting the future growth and diversity of the community. Additionally, 15 posters were presented during a live poster session and 2 online ones. The papers were presented at 24 sessions, the topics of which spanned bias & fairness, conspiracy, polarization, and misinformation, applications in vaccination, music, and social good, research on recommender systems, moderation, influence, norms, and human behavior modeling, and the detection of toxic behavior, bots, and spam.
After consideration by a select committee, two papers were chosen for awards: “Pathways through Conspiracy: The Evolution of Conspiracy Radicalization through Engagement in Online Conspiracy Discussions” by Shruti Phadke, Mattia Samory, Tanushree Mitra for the Best Paper Award, and “Measuring the Monetary Value of Online Volunteer Work” by Hanlin Li, Brent Hecht, Stevie Chancellor for the Best Paper Runner Up. Further, a Test of Time Award was given to the paper titled “The Livehoods Project: Utilizing Social Media toUnderstand the Dynamics of a City” by Justin Cranshaw, Raz Schwartz, Jason I. Hong, Norman Sadeh, which was published in 2012. Finally, the Adamic-Glance Distinguished Young Researcher Award went to Robert (Bob) West (École polytechnique fédérale de Lausanne, Switzerland) for his contributions to the understanding of online polarization and radicalization on digital platforms in the aims of strengthening the Wikipedia ecosystem.
The program also included two invited speakers: Chris Bail, a Professor of Sociology and Public Policy at Duke University with a talk on “Perceived Gender and Political Persuasion: A Social Media Field Experiment during the 2020 Democratic National Primary” and Alice Oh, a Professor in the School of Computing at KAIST, who spoke on “The Importance of Multiple Languages and Multiple Cultures in NLP Research”. Additionally, three panels were organized around the topics of Pandemic & Social Media, Diversity at ICWSM, and Women in Computational Social Science. Other activities includeda student panel where students had a chance to interact with more senior mentors, and a town hall meeting with a discussion of conference organization and future direction.The organizing committee included the General co-chairs: Diyi Yang (Georgia Institute of Technology, USA) and Yelena Mejova (ISI Foundation, Italy), and the PC Chairs: Jisun An (Singapore Management University), Luca Maria Aiello (ITU Copenhagen), Tanu Mitra (University of Washington), along with important contributions by virtual chairs, workshop, tutorial, dataset paper track, data challenge, and other organizers.Next year, in 2023, the 17th International AAAI Conference on Web and Social Media will take place in Cyprus, chaired by Jahna Otterbacher. For more information, visit https://icwsm.org/
Cyber Social Threats (W1)
The role of online platforms as a prime, daily communication tool is coincident with a sharp rise in its misuse, threatening our society in large. These platforms have been implicated for promoting hate speech, radicalization, harassment, cyberbullying, fake news, human trafficking, drug dealing, gender-based stereotyping, and violence among other ills, with a significant impact on individual and community well-being. Especially problematic in recent years, and of particular interest for CySoc 2022, has been the proliferation of vaccine misinformation. Such content and behaviors are inherently multi-faceted, making the recognition of their narratives challenging for researchers as well as social media companies. The implications to individuals and communities require reliable models and algorithms for detecting, understanding, and countering the malevolent behavior in such communications. These challenges have led to a rising prominence of analysis of online communications in academia, politics, homeland security, and industry using computational techniques from natural language processing, statistics, network science, data mining, machine learning, computational linguistics, human-computer interaction, and cognitive science. To meet these challenges, this workshop aims to stimulate research on social, cultural, emotional, communicative, and linguistic aspects of harmful conversations on online platforms and developing novel approaches to analyze, interpret, and understand them. No formal report was filed by the organizes for this workshop.
Social Sensing: Special Edition on Belief Dynamics (W2)
The social sensing workshop (started in 2015) is a multidisciplinary meeting place that brings together social scientists, computer scientists, cognitive scientists, and other disciplines interested in social media analysis, around research that interprets social media as measurement instruments. Social media democratized information production offering an unprecedented view into human habits, customs, culture, stances, and indeed descriptions of physical events that transpire in the world. They also give unprecedented opportunities to spread misinformation, influence opinion, change beliefs, distract from truth, or advance specific agendas, hidden or overt. The potential of social media to influence populations has brought about interest in understanding human belief dynamics and their relation to social media influence, whether by means of benign information sharing or coordination to alter people’s opinions, emotions, behaviors, or understanding of events. What are the scientific foundations for modeling social media as a new communication, measurement, and influence channel? How to utilize information media signals to better understand social systems, communities, and each other? How to identify and mitigate misuse of this medium? What specifically can one measure or influence, what underlying theoretical framework allows one to do so, and what applications are enabled by the endeavor? Since measurement and influence operations are well‐studied in many physical domains, what can one learn from the physical domain (e.g., from the signal processing literature, management systems, or process control) to enable novel social media analysis methods? This scope brings about new interdisciplinary research challenges and opportunities at the intersection of communication, sensing, social network analysis, information theory, data mining, natural language processing, artificial intelligence, cognitive models, and social sciences. No formal report was filed by the organizes for this workshop.
Images in Online Political Communication (W3)
This year’s ICWSM saw the first run of the PhoMemes workshop on Images in Online Political Communication. Visual media has long been a key element of political discourse, and as new online media spaces increasingly focus on imagery (e.g., Instagram, Snapchat, and TikTok), new opportunities arise to study how politicians, political elites, and regular users use such imagery. Despite these advances, our understanding of how images are used for online political discussion lag behind our understanding of text. This workshop is meant to assess these current limits. Topics include roles in misinformation/disinformation, political advertising, and mobilization. PhoMemes accepted seven papers on these topics, primarily on the use of images by disinformation agents, misinformation, and meme characterization. We have also designed a data challenge to push the limits of these topics, around disinformation-agent classification, hate, and screenshot propagation. This challenge attracted five submissions, with submissions from the National University of Singapore and CMU topping the leaderboards across the first two challenges.Beyond these submissions, PhoMemes was also fortunate to have two excellent keynotes: Sefa Ozalp, PhD, Principal Data Scientist at the Anti-Defamation League, and Prof. Kevin Munger, Assistant Professor, Pennsylvania State University. Sefa brought a much-needed perspective on the realities of applying computational tools to combat hate groups and the necessity of including visual media in studies of online discourse–lest we miss a substantial portion of the interaction. Likewise, Kevin described the significant difficulties in understanding the message present in memetic imagery, the interplay with generational politics in America, and how much of that message is socially constructed and absent from the visual itself.
Erik Bucy, Cody Buntain, Andreu Casas, Jungseock Joo, Dhavan Shah, and Zacharcy Steinert-Threlkeld graciously gave their time as co-chairs for this workshop. This report was written by Cody Buntain.
Novel Evaluation Approaches for Text Classification Systems on Social Media (W4)
The automatic or semiautomatic analysis of textual data is a key approach to analyse the massive amounts of user-generated content online, from the identification of sentiment in text and topic classification to the detection of abusive language, misinformation or propaganda. However, the development of such systems faces a crucial challenge. Static benchmarking datasets and performance metrics are the primary method for measuring progress in the field, and the publication of research on new systems typically requires demonstrating an improvement over state-of-the-art approaches in this way. Yet, these performance metrics can obscure critical failings in current models. Improvements in metrics often do not reflect improvements in the real-world performance of models. There is clearly a need to rethink performance evaluation for text classification and analysis systems to be usable and trustable.
If unreliable systems achieve astonishing scores with traditional metrics, how do we recognise progress when we see it? The goal of the workshop on Novel Evaluation Approaches for Text Classification Systems on Social Media (NEATCLasS) is to promote the development and use of novel metrics for abuse detection, sentiment analysis and similar tasks within the community, to better be able to measure whether models really improve upon the state of the art, and to encourage a wide range of models to be tested on these new metrics. No formal report was filed by the organizes for this workshop.
Social Media for Emergency Response (W5)
Social media platforms have evolved to play multiple roles in contemporary society – such as primary sources for communication, acting as a proxy for information needs, and also becoming a medium for seeking assistance. One of the interesting roles social media platforms have started being used nowadays for is – emergency response. During emergency events, delivering the right information in a timely manner is crucial, not only for emergency response organizations but also for other social media users in the vicinity. Social media platforms allow people who are actually at ground zero to express their opinions and provide information about the event, thus becoming in most cases, primary sources of information. However, mining social media data to provide insights that can be used by emergency response comes with significant technical challenges, such as real-time information extraction, information ranking, credibility, multimodality, information visualization, data sampling, etc. Tackling these problems require researchers to take a truly inter-disciplinary approach to enable the use-case of using social media data for emergency response holistically. Through this workshop, we want to provide a cross-disciplinary forum for researchers from different fields to present and discuss research on this topic from different perspectives.
When an emergency event first occurs, getting the right information as quickly as possible is critical in saving lives. When an event is ongoing, information on what is happening can be critical in making decisions to keep people safe and take control of the particular situation unfolding. In both cases, first responders, rescuers, peacekeepers, and others have to quickly make decisions that include what resources to deploy and where. Fortunately, in most emergencies, people use social media to publicly share information. Social media platforms have changed emergency response, in the sense that information publicly posted on such platforms is now critical, not just in understanding what’s happening on the ground, but also in making critical decisions on where resources are needed and how to deploy them. In many cases, social media platforms have also been used to communicate with communities impacted by emergencies- such communication happens between members of the community, emergency response planners, non-governmental aid agencies, and government agencies.
Since the popularization of social media platforms, thousands of research papers have been published on highly diverse topics of interest to different research fields- from computer science to sociology. Scientists have explored networks, recommendations, ranking, propagation of information, influence, and many other technical topics. In spite of its importance, the research community, however, in our view, has not done as much work in relation to how social media data can be used for emergency response. There are significant efforts at various universities working on these topics, but there are few venues to gather scientists from different disciplines together to present and discuss progress and challenges in leveraging social media data for emergency response. The goal of this workshop is to provide a stage for scientists working on different fields relevant to leveraging social media data for emergency response. While the focus of the workshop will be on emergency response, the actual sub-topics will be broad, allowing researchers to present state-of-the-art results that might be relevant to this important topic, while providing the space for presenting new, innovative ideas and perspectives from researchers in technical and non-technical fields. No formal report was filed by the organizes for this workshop.
News Media and Computational Journalism (W6)
The main goal of the workshop is to bring together media practitioners and technologists to discuss new opportunities and obstacles that arise in the modern era of information diffusion. This year’s theme is: Misinformation: new directions in automation, real-world applications, and interventions. No formal report was filed by the organizes for this workshop.
Data for the Wellbeing of Most Vulnerable (W7)
The scale, reach, and real-time nature of the Internet is opening new frontiers for understanding the vulnerabilities in our societies, including inequalities and fragility in the face of a changing world. From tracking seasonal illnesses like the flu across countries and populations, to understanding the context of mental conditions such as anorexia and bulimia, web data has the potential to capture the struggles and wellbeing of diverse groups of people. Vulnerable populations including children, elderly, racial or ethnic minorities, socioeconomically disadvantaged, underinsured or those with certain medical conditions, are often absent in commonly used data sources. The very absence of these populations in data can reveal areas of concern, indicating potential lack of access to vital technologies, and potentially being overlooked by algorithms trained on such data. The recent developments around COVID-19 epidemic makes these issues even more urgent, with an unequal share of both disease and economic burden among various populations.
Thus, the aim of this workshop is to encourage the community to use new sources of data to study the wellbeing of vulnerable populations including children, elderly, racial or ethnic minorities, socioeconomically disadvantaged, underinsured or those with certain medical conditions. The selection of appropriate data sources, identification of vulnerable groups, and ethical considerations in the subsequent analysis are of great importance in the extension of the benefits of big data revolution to these populations. As such, the topic is highly multidisciplinary, bringing together researchers and practitioners in computer science, epidemiology, demography, linguistics, and many others. No formal report was filed by the organizes for this workshop.
Cody Buntain is an Assistant Professor for the School of Information Studies at the University of Maryland, College Park.
Yelena Mejova is a Senior Research Scientist at the ISI Foundation in Turin, Italy, working in the area of Data Science for Social Impact and Sustainability.
Diyi Yang is a computer scientist and Assistant Professor in the Georgia Institute of Technology School of Interactive Computing.