Misinformation has been the focus of significant public attention following a spate of volatile civic events including elections and national referendums that surfaced the prevalence of influence operations. The persistence of misinformation on online platforms is a testament to the inadequacy of existing solutions. With the democratization of generative models for text, images, and videos that immediately enter online distribution channels, it is important to develop safer governance mechanisms via policies that draw on lessons learned from practitioners, and from research in social science and machine learning. We provide a platform to discuss and present research in these areas, and encourage members to develop collaborative early-stage research ideas by tapping into the community.
A sample of topics that are of interest to us include:
- Generative Models
- Social Network Analysis
- Fake News
- Hate Speech
- Online Toxicity
- Causal Effects and Interventions
- Natural Language Processing
- Agent-based Models
- Multiagent Systems
- Reinforcement Learning
- Transparency and Ethics
- Recommender Systems
- Algorithmic Auditing
- Psychosocial Analysis
This reading group is a space to introduce beginners to research in these fields. It will provide a platform to continue the conversation with a cross-functional community to foster nuanced discussions on how to address the challenges created by global misinformation through the ethical use of technology and effective governance.
Organizers
The AI, Misinformation, and Policy (AMPol) Seminar is organized by Swapneel Mehta (CDS Ph.D. candidate) and Megan Brown (CSMAP Sr. Research Engineer).
Attendance Information
While we had in-person events in the past year, all seminars are at present held virtual-only unless otherwise specified.
Spring 2024
Date & Time: Wednesday, April 17 at 9:00am EST
Speaker & Title: Lynnette Hui Xian Ng (PhD Candidate, CMU), “Geopolitical Narratives of Social Media Bots”
Bio: Lynnette Ng is a PhD Candidate in Societal Computing at Carnegie Mellon University, School of Computer Science. Her research deals with using large-scale social media data to understand patterns of human and bot behavior on social media, motivated by mis/disinformation and information maneuvers. Her work has won Best Paper Awards in top conferences and journals, and has been featured in magazines like New Scientist and Impact Magazine, and have trended on Reddit. Prior to her PhD, Lynnette obtained a Bachelors in Computer Science from National University of Singapore, where she won the NUSS medal for Outstanding Achievement.
Abstract: Social media bot agents raise concern because they can spread disinformation and manipulate elections. This talk will provide a glimpse of bot narratives against a geopolitical backdrop, juxtaposing the bot activities of different regions. I will highlight some of the recent research studies revolving around the actions of bots in the geopolitical landscape, and describe the process of bot detection in several social media platforms. Finally, I discuss the manner in which bots work together in groups, which is known as coordination, and relevant detection and ranking algorithms.
Location: Zoom
Date & Time: Tuesday, January 30, 12:00pm
Speaker & Title: Eric Brichetto (Independent OSINT Researcher), “Exposing Disinformation Campaigns Through Actor-Focused Open Source Intelligence”
Bio: Eric Brichetto is an open-source intelligence (OSINT) investigator and software engineer with 2 years of experience in uncovering disinformation campaigns, propaganda narratives, and the threat actors behind them. Previously at The Toro Company for 5 years, he spearheaded development of 3 internal-facing products & initiatives connected to the MyTurf & MyToro applications and introduced best practices for API versioning. As an OSINT analyst, Eric has partnered with experts in information warfare and cybercrime to provide intelligence to Congress, track the activities of international influence operators, aggregate data on the national influence industry, and fact-check the conclusions of the US Department of the Treasury’s Office of Foreign Asset Control on international sanctions designations. His work combines his software and data engineering skills with advanced OSINT techniques. Eric is a Microsoft-certified Azure Developer Associate, an Atlantic Council Digital Sherlock scholar, and lightly published author on Medium documenting his own investigations into disinformation networks.
Abstract: Disinformation investigations often focus on assessing content, which can sometimes miss the broader ecosystems in which deception spreads. This presentation will demonstrate a powerful paradigm shift: leveraging OSINT to take an actor-focused approach. The speaker will showcase techniques including social media profiling, analyzing business records, and cross-referencing technical & non-technical public data to systematically map connections between various influence operators. A case study will trace from a single suspicious website about French politics to a network of European entities involved in coordinated propaganda campaigns. Audiences will gain new perspective on how both academic and industry practitioners can use actor-focused OSINT to expose the web-based and human infrastructure behind even sophisticated misinformation efforts. The talk aims to equip the next generation of researchers and practitioners with an expanded toolkit to illuminate manipulation in the digital public sphere.
Location: Zoom
Spring 2023
Date & Time: Tuesday, April 18, 3:30pm
Speaker & Title: Nathaniel Lubin, “Accountability Infrastructure for Recommendation Systems: Methods for Identifying and Mitigating Structural Harms Created by System Architecture” (additional resource: “Accountability Infrastructure” website)
Bio: Nathaniel Lubin has spent his career focused on digital strategy, technology, and politics. Recently, his work has centered on developing novel approaches to improving online discourse, building measurement tools, and combating misinformation. Nathaniel is a Rebooting Social Media Fellow at Harvard’s Berkman Klein Center and a Visiting Fellow at the Digital Life Initiative at Cornell Tech. He founded Fellow Americans, a non-profit which creates and tests more effective digital content, focusing on topics like COVID-19 response, civic participation, and improved social trust while working with some of the largest progressive organizations, and CEO of Survey 160, a software product designed to source data for polling and research. As a consultant, he has assisted more than 30 startups, major corporations, foundations, and advocacy organizations working to leverage technology and digital tools to better communicate with key audiences. Nathaniel previously was the Director of the Office of Digital Strategy at the White House where he led a team of strategists and practitioners to modernize the way the White House engaged and communicated with the American public. Before that, he served as Director of Digital Marketing at Obama for America in 2012 where he led the largest paid digital fundraising, persuasion, and outreach programs yet run in politics with a budget of more than $112 million. Originally from New York, Lubin is an honors graduate from Harvard University.
Abstract: Attention capitalism has generated design processes and product development decisions that prioritize platform growth over all other considerations. To the extent limits have been placed on these incentives, interventions have primarily taken the form of content moderation. While moderation is important for what we call “acute harms,” societal-scale harms – such as negative effects on mental health and social trust – require new forms of institutional transparency and scientific investigation, which we group under the name accountability infrastructure.
This is not a new problem. In fact, there are many conceptual lessons and implementation approaches for accountability infrastructure within the history of public health. Channeling these insights, we reinterpret the societal harms generated by technology platforms as a public health problem. To that end, we present a novel mechanism design framework and practical measurement methods for that framework. The proposed approach is iterative and built into the product design process, and is applicable for either internally-motivated (i.e. self regulation by companies) or externally-motivated (i.e. government regulation) interventions.
In doing this, we aim to help shape a research agenda of principles for mechanism design around problem areas on which there is broad consensus and a firm base of support. We offer constructive examples and discussion of potential implementation methods related to these topics, as well as several new data illustrations for potential effects of exposure to online content.
Date & Time: Friday, April 7, 11:15am
Speaker & Title: Dr. Christian Schroeder de Witt, “Bringing Multi-Agent Learning to Societal Impact: From Steganography to Solar Geoengineering”
Recording: View Christian Schroeder de Witt’s Lecture Recording (Form to Request Password to Christian Schroeder de Witt’s Lecture Recording)
Slides: View Christian Schroeder de Witt’s Lecture Slides
Bio: Over the past years, much progress has been made in deep multi-agent learning in recreational games, such as Go or DOTA 2. In this talk, I propose to use these tools to generate societal impact instead. Starting from my recent breakthrough research yielding the world’s first perfectly secure generative steganography algorithm, to robustifying Human-AI systems against illusory attacks, non-linear decision-making in the net-zero transition, and optimizing solar geoengineering discuss how multi-agent learning can raise new questions about some of the biggest societal challenges of our time, as well as power new technology to advance equity, expand opportunity and protect basic rights and liberties.
Abstract: Dr. Christian Schroeder de Witt is an artificial intelligence researcher specialising in fundamental research on multi-agent control in high-dimensional settings. He has authored a variety of highly influential research works, and is pioneering numerous real-world applications of deep multi-agent reinforcement learning, ranging from steganography to climate-economics models and solar geoengineering. As part of various industry collaborations, Christian has previously worked on A.I for autonomous drone control, as well as automated cybersecurity defence systems. Christian currently holds a postdoctoral researcher role at the University of Oxford, UK.
Born and raised in Frankfurt am Main, Germany, Christian holds a DPhil (Ph.D.) and two distinguished master’s degrees from the University of Oxford, ranging from theoretical physics, to computer science, and artificial intelligence. Aside from academia, Christian has experience working in diverse industries, including with Google AI, Man AHL, and as Head of Engineering of a Berlin eCommerce company.
In 2020, Christian was named “30 under 30” youth politician of a UK political party with House of Commons representation. Christian’s political work on climate policy is informed by a long-term research assistantship with Prof. Myles Allen (Coordinating Lead Author of IPCC SR1.5) at Oxford Net Zero, where he has been focusing on decarbonising small and medium-sized companies. In 2022, Christian was selected as a “30 under 35 rising strategist (Europe)” by Schmidt Futures International Strategy Forum. This fellowship allows him to work with the European Council on Foreign Relations.
Date & Time: Friday, March 31, 11:15am
Speaker & Title: Kiran Garimella, “Data donation systems for platform research”
Recording: View Kiran Garimella’s Lecture Recording (Form to Request Password to Kiran Garimella’s Lecture) (captions coming)
Slides: View Kiran Garimella’s Lecture Slides
Bio: Kiran Garimella is an Assistant Professor in the School of Communication and Information at Rutgers University. His research deals with using large-scale data to tackle societal issues such as misinformation, political polarization, and hate speech. Prior to joining Rutgers, Dr. Garimella was the Michael Hammer postdoc at the Institute for Data, Systems and Society at MIT and a postdoc at EPFL, Switzerland. His work on studying and mitigating polarization on social media won the best paper awards at top computer science conferences. Kiran received his Ph.D. in computer science at Aalto University, Finland, and Masters & Bachelors from IIIT Hyderabad, India. Prior to his Ph.D., he worked as a Research Engineer at Yahoo Research, Barcelona, and QCRI, Doha.
Abstract: Data donation systems are emerging as a new way to facilitate research on social media platforms, where access to user data can be restricted due to privacy concerns. These systems allow users to voluntarily donate their data for research purposes, providing researchers with valuable insights into user behavior and platform dynamics. In this talk, I will discuss the potential benefits and challenges of data donation systems for platform research. I will explore the different models of data donation systems, including opt-in and opt-out approaches, and examine the ethical considerations involved in such systems. I will also discuss the technical challenges of implementing data donation systems, including data quality control, data security, and data anonymization. Finally, I will highlight some of the recent research studies that have been conducted using data donation systems and the potential impact of such studies on our understanding of social media platforms. Overall, data donation systems have the potential to provide researchers with unprecedented access to user data, while also protecting user privacy and autonomy. However, the design and implementation of such systems must be carefully considered to ensure that they are transparent, ethical, and technically feasible.
Date & Time: Monday, March 6th, 2:30 pm
Speaker & Title: Emily Saltz, Jigsaw “Using mixed methods to understand the effects of online information interventions: Lessons from UX research in industry”
Recording: View Emily Saltz Lecture Recording (Form to Request Password to Emily Saltz Lecture) (captions coming)
Slides: View Emily Saltz Lecture Slides
Bio: Emily Saltz is a UX Researcher at Google Jigsaw, working on tools for platforms and moderators to address online harms. Before that, she was a UX Researcher at the New York Times R&D Lab, conducting research on topics ranging from media credibility (the News Provenance Project), to NLP Q&A tools. She was a 2020 Fellow at the Partnership on AI, and holds a Master’s in Computer-Interaction from Carnegie Mellon, and a BA in Linguistics from UC Santa Cruz.
Abstract: This talk will provide a glimpse into the experience of studying online harms and information interventions in industry R&D groups such as Google Jigsaw and the New York Times, and how this research is used to inform product decisions deployed at scale. It will also describe the process of working with cross-functional product teams, using in-depth qualitative research alongside larger scale surveys and lab studies to answer questions relevant to both industry and academia, such as user attitudes towards credibility labels across platforms.