Carnegie Mellon University

Center for Informed Democracy & Social - cybersecurity (IDeaS)

CMU's center for disinformation, hate speech and extremism online

IDeaS Center for Informed Democracy & Social-cybersecurity
Efficient Over-time Analysis of Social Network Data Sets
Rick Carley
Tuesday, June 15 1:00-2:00 EDT

Analysis of over-time Social Network Data is extremely important in understanding how a given social discourse is developing. This can include changes in network structure (the connections between agents) and in the emotions and words being communicated. In this session, we will focus on over-time analysis of changes in the network structure by making use of a variety of social network metrics. In part, the changes in time of key social network metrics can indicate that a significant change in the society is happening. We will explain how the ability to detect such changes can be enhanced through the use of advanced signal processing techniques. In addition, we will explore the use of social network metrics for extracting various types of social groups (often called the community structure). We will discus the computational problem that arises in the detection of groups, given dynamic network data. In particular, we will consider the challenge of detecting groups given time variant incremental data with as little delay as possible given vast quantities of network data.

Signal Detection in Cyberspace
Alex Davis
Tuseday, June 15 2:45-4:15 EDT

In this session I'll discuss the use of signal detection theory for understanding both normative and descriptive decision-making in cyber environments. Normative theories define bounds on the ability of any system to reduce errors in decisions that separate signal from noise. Descriptive theories fit statistical models to human data to understand both their performance on signal detection tasks, as well as characterize the reasons for human deviation from optimal performance. Phishing email detection and Twitter bot detection are two examples where normative and descriptive aspects of signal detection theory can help us understand optimal behavior, human performance, and possible interventions that improve public welfare.

Decision Science for Social Cyber-Security
Baruch Fischhoff
Thursday, June 10 4:45-6:15 EDT

The performance of social cyber-security systems depends on the decisions made by people involved with them.  Those include the people who use them (“should I trust this message?”), the people who analyze them (“what is the risk of ransomware?”), and the people who manage them (“how much should we spend on training?”).  Decision science offers a systematic approach to analyzing and potentially improving the operation of technical systems.  The talk will present that approach, illustrated by practical examples from social cyber-security and other domains.

Applying Social Psychology to Cybersecurity
Jason Hong
Friday, June 11 4:45-6:15 EDT

This talk would present the results of our work in applying social psychology to influence people's cybersecurity behaviors, as well as factors leading to adoption (or not) of cybersecurity features on Facebook. Would also talk about the Security Sensitivity stack that we've developed of awareness, ability,
and motivation. Would wrap up with SA-6 and SA-13 survey instruments that we've developed to measure security attitudes, plus some ongoing work.

Success in the Marketplace of Ideas 
Danny Oppenheimer
Monday, June 14 2:45-4:15 EDT

Some ideas spread widely; others don't.  Whether an idea spreads has little relation to whether or not it is true.  In this workshop, I review the psychology literature on what makes ideas believable, memorable, and contagious, with the ultimate goal of creating a more informed society. 

Modeling and Analysis of Information Spread over Multi-layer Networks
Osman Yağan
Friday, June 11 3:00-4:30 EDT

In this session, we will present recent advances in mathematical modeling and analysis of spreading processes over networks. The main focus will be on information and misinformation spread over social networks, but we will also talk about how these tools and concepts can be applied more broadly to other spreading processes including viral spread. Specifically, we will present results on spreading processes over multi-layer networks with an eye towards revealing the impact of multiple social media platforms on the speed and extent of (mis)information spread. Secondly, we will discuss the impact of mutations (e.g., modifications made on the information during the propagation) on the spread of (mis)information.

Technical Day- Saturday, June 12

Hate Speech on Social Networks
Joshua Uyheng

Hate speech represents a significant issue on social media platforms, as it targets vulnerable populations and links to real-world violence. To address this problem, extensive efforts have sought to develop methods for detecting hate speech in an automated fashion. Yet hate speech is more than a linguistic phenomenon; it is also a social one. This talk discusses how computational tools may be integrated with social scientific theory to go from merely identifying hate speech, to understanding its community-driven dynamics. Furthermore, from a social cyber-security perspective, I show how attention to online hate networks enables a richer characterization of the activities and impacts of bots and trolls through information operations. Through this work, I introduce novel theoretical contributions for understanding online hate speech, general methodologies for characterizing its dynamics in social networks, and empirical results on case studies of national elections and the global Covid-19 pandemic. I conclude with directions for future work and potential interventions arising from insights gained from this research.

*Additional descriptions will be added