Speakers
The rise of language models: shaping communication in the era of disinformation
Modern language models are becoming increasingly popular and one of key tools in the field of artificial intelligence. These models are trained using vast amounts of text data and can generate credible, human-like text. Prominent examples of language models include GPT-3, GPT-4 and ChatGPT. Their popularity is due to their versatility in natural language processing (NLP) tasks, including stories generation, language translation, summarization, chatbots, sentiment analysis, and more. During this presentation, we will explore the intricacies of language models, discussing their fundamental principles and showcasing interesting examples of their usage. We will also highlight the importance of language prompt techniques in effectively leveraging these models for specific tasks. Our focus will be on the issue of disinformation, where we will examine the dangers, usage types, and possible countermeasures to prevent language models from being exploited as a low-cost tool to pollute public debates on the Web and on Social Media.
Tiziano Fagni
Tiziano Fagni is a researcher at IIT-CNR’s Cyber Intelligence Research Unit. He holds an Msc in computer science and a PhD in information engineering from the University of Pisa. His research focuses on machine learning and data mining, specifically natural language processing (NLP) on big data and social media. He uses deep learning algorithms and neural language modeling techniques to address social media issues such as hate speech detection, deepfake text generation and detection, and user stance detection on political content. He has published several peer-reviewed papers on major international conferences and journals and is an active reviewer for several important journals.
Watermarking: Defense and Hazards
digital watermarking allows to hide information within a digital carrier, such as text, video, and network traffic. For instance, cloaked data can be used to check the integrity of a software, track the diffusion of digital media, or to enforce the intellectual property. With the diffusion of AI-generated contents, the ability of developing advanced watermarking schemes has become a prime research topic. In fact, watermarking enables to protect the code generated through large language models or to understand whether an image has been created by a human or a machine. Unfortunately, the availability of techniques to conceal data within other data also opens to many security issues, including an emerging class of threats defined as steganographic malware. Therefore, this course briefly introduces the core concepts of digital watermarking and shows its most cutting-edge applications, such as the defense against deepfakes. It also outlines the main research questions to be faced to make watermarking techniques capable of handling future digital contents and support ethical needs.
Luca Caviglione
Luca Caviglione (male) is a Senior Research Scientist at the Institute for Applied Mathematics and Information Technologies of the National Research Council of Italy. He holds a Ph.D. in Electronic and Computer Engineering from the University of Genoa, Italy. His research interests include optimization of large-scale computing frameworks, traffic analysis, wireless and heterogeneous communication architectures, and network security. He is the author or co-author of about 200 academic publications, and several patents in the field of p2p and energy-aware computing. He has been involved in Research Projects and Network of Excellences funded by the ESA, the EU and the MIUR. He is also a contract professor in the field of networking/security and a board member for the PhD program in Security, Risk and Vulnerability of the University of Genoa (Cybersecurity and Reliable AI). He is the head of the IMATI Research Unit of the National Inter-University Consortium for Telecommunications, part of the Steering Committee of the Criminal Use of Information Hiding initiative within the European Cybercrime Center.
NLP & Network Science for Modeling Vaccine Hesitancy
Vaccine hesitancy has been a recent battleground for public health practitioners, political pundits, and social media influencers. It presents a unique opportunity to study the role of communication technologies on opinion formation, the expression of the underlying values, the algorithmically-induced echo chambers, and the effect of persuasive techniques, including mis- and disinformation. In this session, we will learn about the latest theories and tools in Natural Language Processing, Network Science, and Computational Social Science to extract and model the above signals in order to explore the dynamics of vaccine hesitancy and similar controversial topics.
Yelena Mejova
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. Specializing in social media analysis and mining, her work concerns the quantification of health and wellbeing signals in social media, as well as tracking of social phenomena, including politics and news consumption. Since 2023, she is a co-Editor-in-Chief of EPJ Data Science. Previously as a scientist at the Qatar Computing Research Institute, Yelena was a part of the Social Computing Group working on computational social science, especially as applied to tracking real-life health signals. Her papers received best paper awards at numerous ACM conferences, including GoodIT (2023), CHI (2021), UMAP (2019), and CIKM (2013).
AI for Multimedia: Enhancing Security and Securing Content
Artificial Intelligence has an increasing role in both multimedia analysis and creation. The tasks in which AI is used range from retrieval to generation, from recognition to rescue. This course will begin by revisiting the representation learning aspect of deep learning, followed by analyzing the opportunities and risks it presents. Then, it will address various recent research topics in multimedia related to enhancing security and securing content through AI.
Fabrizio Falchi
Fabrizio Falchi is a senior researcher at the Artificial Intelligence for Multimedia and Humanities laboratory (AIMH) of the Information Science and Technologies Institute (ISTI) of CNR in Pisa, where he coordinates the activities of the Computer Vision and Deep Learning research group. He has a Ph.D. in Information Engineering from the University of Pisa and a Ph.D. in Informatics from the Faculty of Informatics of Masaryk University of Brno. He also received an M.B.A. from Scuola Superiore Sant’Anna, to which he is also associated. He is a member of the scientific council of DIITET-CNR, delegated for the DIITET node in the Artificial Intelligence and Intelligent Systems national laboratory of CINI, and he has been co-chair of the national virtual lab on Artificial Intelligence of CNR. His research interests include computer vision, deep learning, artificial intelligence, and multimedia information retrieval.
Sentiment analysis from social media platforms
Every day, millions of people use social media platforms, generating a vast amount of opinion-rich data. This data can be exploited to extract valuable information about human dynamics and behaviors. In this context, we will discuss how topic discovery, opinion mining, and emotion analysis techniques can be applied to social media data to uncover political polarization among people. As a case study, we will describe a sentiment analysis methodology applied to the 2020 US presidential election. In our case study, we utilized a clustering-based technique to extract the main discussion topics and monitor their weekly impact on social media conversations. Subsequently, we employed a neural-based opinion mining technique to determine the political orientation of social media users by analyzing their posts. We also investigated the temporal dynamics of online discussions by studying how users’ publishing behavior correlates with their political alignment. Finally, we will describe how sentiment analysis and text mining techniques can be combined to discover the relationship between user polarity and the sentiments expressed towards different candidates.
Paolo Trunfio
Paolo Trunfio is an Associate Professor of Computer Engineering at University of Calabria, Italy. He currently serves as an associate editor of the Journal of Big Data, ACM Computing Surveys and is a member of the editorial boards of several scientific journals, including Future Generation Computer Systems, Big Data and Cognitive Computing, the International Journal of Web Information Systems, and the International Journal of Parallel, Emergent and Distributed Systems. He is a senior member of IEEE and ACM.
Polarization on social networks
The advent of online social networks has amplified the long-standing phenomenon of opinion polarization. This lecture will explore the interdisciplinary research on opinion formation, focusing on mathematical models developed in sociology and related fields. We will delve into prominent models, such as those based on social influence, homophily, and confirmation bias, highlighting their strengths and limitations. The lecture will examine the historical context and contemporary manifestations of polarization in online environments, presenting key models and their underlying assumptions, and discussing their ability to explain real-world polarization phenomena. It will also address the computational complexities in simulating and analyzing these models, particularly in large-scale networks, and explore empirical studies that have tested their validity. Finally, the lecture will consider the implications of these models for understanding social dynamics, designing interventions to mitigate polarization, and predicting future trends. This lecture is intended for researchers, students, and practitioners interested in the intersection of social networks, opinion dynamics, and computational social science. It will provide a comprehensive overview of the current state of research and highlight exciting avenues for future investigation.
Chiara Boldrini
Chiara Boldrini is a Senior Researcher at IIT-CNR. Her research interests are in computational social sciences, decentralized AI, mobile and ubiquitous systems. She has published 60+ papers on these topics. She is the IIT-CNR co-PI for the National Extended Partnership in Artificial Intelligence FAIR, H2020 SoBigData++ and H2020 HumaneE-AI-Net projects, and was involved in several EC projects since FP7. She currently holds the position of Editor-in-Chief for Special Issues at Elsevier Computer Communications and is a member of the Editorial Board of Elsevier Pervasive and Mobile Computing. She served as TPC chair of IEEE PerCom’24 and, over the years, has been on the organizing committee of several IEEE and ACM conferences/workshops, including IEEE PerCom and ACM MobiHoc.
How I got my ERC Grant
This talk covers my experience with the ERC Grants: one of the most challenging and prestigious funding schemes in Europe. I will discuss my journey with the ERC, including my initial failure and the subsequent victory of an ERC Starting Grant. The talk will briefly explain what an ERC Grant is, how to apply for such a grant, and how ERC proposals are made and evaluated. Next, I will discuss how I prepared my winning proposal, focusing not on the scientific content of the proposal itself, but rather on the critical choices that I made before and during the writing of the proposal. I will analyze similarities and differences between my failed and successful attempts. Finally, I will talk about the ERC interview, a very peculiar and feared feature of the ERC evaluation process. Overall, this talk provides basic information for all young researchers possibly interested in the ERC Grants, and key insights for mastering an ERC proposal preparation.
Stefano Cresci
Stefano Cresci received the Ph.D. degree in Information Engineering from the University of Pisa. He is a Researcher at IIT-CNR, Italy. His interests lay at the intersection of web science and data science, with a focus on online harms, content moderation, and coordinated online behavior. For his achievements he received multiple awards, including an ERC grant to develop data-driven, user-centered, and personalized content moderation (DEDUCE), the ERCIM Cor Baayen Young Researcher Award, the IEEE Next-Generation Data Scientist Award, and the IEEE Computer Society Italy Section Chapter PhD Thesis Award.