Language models can be trained to recognize the moral sentiment of text, creating new opportunities study role morality in human life. As interest language and has grown, several ground truth datasets with annotations have been released. However, these vary method data collection, domain, topics, instructions for annotators, etc. Simply aggregating such heterogeneous during training yield that fail generalize well. We describe a fusion framework on multiple improve performance...

Online manipulation is a pressing concern for democracies, but the actions and strategies of coordinated inauthentic accounts, which have been used to interfere in elections, are not well understood. We analyze five million-tweet multilingual dataset related 2017 French presidential election, when major information campaign led by Russia called "#MacronLeaks" took place. utilize heuristics identify accounts detect attitudes, concerns emotions within their tweets, collectively known as...

The rich and dynamic information environment of social media provides researchers, policy makers, entrepreneurs with opportunities to learn about phenomena in a timely manner. However, using these data understand behavior is difficult due heterogeneity topics events discussed the highly online environment. To address challenges, we present method for systematically detecting measuring emotional reactions offline change point detection on time series collective affect, further explaining...

Recent advances in NLP have improved our ability to understand the nuanced worldviews of online communities. Existing research focused on probing ideological stances treats liberals and conservatives as separate groups. However, this fails account for views organically formed communities connections between them. In paper, we study discussions 2020 U.S. election Twitter identify complex interacting Capitalizing interconnectedness, introduce a novel approach that harnesses message passing...

The COVID-19 pandemic has upended daily life around the globe, posing a threat to public health. Intuitively, we expect that surging cases and deaths would lead fear, distress other negative emotions. However, using state-of-the-art methods measure sentiment, emotions, moral concerns in social media messages posted early stage of pandemic, see counter-intuitive rise positive affect. We hypothesize increase positivity is associated with decrease uncertainty emotion regulation. Finally,...

Effective response to pandemics requires coordinated adoption of mitigation measures, like masking and quarantines, curb a virus's spread. However, as the COVID-19 pandemic demonstrated, political divisions can hinder consensus on appropriate response. To better understand these divisions, our study examines vast collection COVID-19-related tweets. We focus five contentious issues: coronavirus origins, lockdowns, masking, education, vaccines. describe weakly supervised method identify...

The COVID-19 pandemic has posed unprecedented challenges to public health world-wide. To make decisions about mitigation strategies and understand the disease dynamics, policy makers epidemiologists must know how is spreading in their communities. Here we analyse confirmed infections deaths over multiple geographic scales show that COVID-19's impact highly unequal: many regions have nearly zero infections, while others are hot spots. We attribute effect a Reed-Hughes-like mechanism which...

Effective response to pandemics requires coordinated adoption of mitigation measures, like masking and quarantines, curb a virus's spread. However, as the COVID-19 pandemic demonstrated, political divisions can hinder consensus on appropriate response. To better understand these divisions, our study examines vast collection COVID-19-related tweets. We focus five contentious issues: coronavirus origins, lockdowns, masking, education, vaccines. describe weakly supervised method identify...

Language models (LMs) are known to represent the perspectives of some social groups better than others, which may impact their performance, especially on subjective tasks such as content moderation and hate speech detection. To explore how LMs different perspectives, existing research focused positional alignment, i.e., closely mimic opinions stances groups, e.g., liberals or conservatives. However, human communication also encompasses emotional moral dimensions. We define problem affective...

User representation learning aims to capture user preferences, interests, and behaviors in low-dimensional vector representations. These representations have widespread applications recommendation systems advertising; however, existing methods typically rely on specific features like text content, activity patterns, or platform metadata, failing holistically model behavior across different modalities. To address this limitation, we propose SoMeR, a Social Media Representation framework that...

Narratives are foundation of human cognition and decision making. Because narratives play a crucial role in societal discourses spread misinformation because the pervasive use social media, narrative dynamics on media can have profound impact. Yet, systematic computational understanding online faces critical challenge scale dynamics; how we reliably automatically extract from massive amount texts? How do emerge, spread, die? Here, propose discovery framework that fill this gap by combining...

Effective communication during health crises is critical, with social media serving as a key platform for public experts (PHEs) to engage the public. However, it also amplifies pseudo-experts promoting contrarian views. Despite its importance, role of emotional and moral language in PHEs' COVID-19 remains under explored. This study examines how PHEs communicated on Twitter pandemic, focusing their engagement political elites. Analyzing tweets from 489 356 January 2020 2021, alongside...

Social scientists use surveys to probe the opinions and beliefs of populations, but these methods are slow, costly, prone biases. Recent advances in large language models (LLMs) enable creating computational representations or "digital twins" populations that generate human-like responses mimicking population's language, styles, attitudes. We introduce Community-Cross-Instruct, an unsupervised framework for aligning LLMs online communities elicit their beliefs. Given a corpus community's...

<sec> <title>BACKGROUND</title> Effective communication is crucial during health crises, and social media has become a prominent platform for public experts to inform engage with the public. At same time, also platforms pseudo-experts who may promote contrarian views. Despite significance of media, key elements such as use moral or emotional language messaging strategy, particularly COVID-19 pandemic, not been explored. </sec> <title>OBJECTIVE</title> This study aims analyze how notable...

Quantifying the effect of textual interventions in social systems, such as reducing anger media posts to see its impact on engagement, poses significant challenges. Direct real-world systems are often infeasible, necessitating reliance observational data. Traditional causal inference methods, typically designed for binary or discrete treatments, inadequate handling complex, high-dimensional nature This paper addresses these challenges by proposing a novel approach, CausalDANN, estimate...

Background Effective communication is crucial during health crises, and social media has become a prominent platform for public experts (PHEs) to share information engage with the public. At same time, also provides pseudoexperts who may spread contrarian views. Despite importance of media, key elements communication, such as use moral or emotional language messaging strategy, particularly emergency phase COVID-19 pandemic, have not been explored. Objective This study aimed analyze how PHEs...

The rich and dynamic information environment of social media provides researchers, policymakers, entrepreneurs with opportunities to learn about phenomena in a timely manner. However, using these data understand behavior is difficult due heterogeneity topics events discussed the highly online environment. To address challenges, we present method for systematically detecting measuring emotional reactions offline change point detection on time series collective affect, further explaining...

The rich and dynamic information environment of social media provides researchers, policy makers, entrepreneurs with opportunities to learn about phenomena in a timely manner. However, using this data understand behavior is difficult due heterogeneity topics events discussed the highly online environment. To address these challenges, we present method for systematically detecting measuring emotional reactions offline change point detection on time series collective affect, further explaining...

Explicit and implicit bias clouds human judgment, leading to discriminatory treatment of disadvantaged groups. A fundamental goal automated decisions is avoid the pitfalls in judgment by developing decision strategies that can be applied all protected Improving fairness interventions via decision-inspired methods, however, has been under-utilized. In this paper, we propose a causal framework learns optimal intervention policies from data subject novel constraints. We define two measures...