No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Resolving content moderation dilemmas between free speech and harmful misinformation
Abstract
In online content moderation, two key values may come into conflict: protecting freedom of expression and preventing harm. Robust rules based in part on how citizens think about these moral dilemmas are necessary to deal with this conflict in a principled way, yet little is known about people's judgments and preferences around content moderation. We examined such moral dilemmas in a conjoint survey experiment where US respondents (N = 2, 564) indicated whether they would remove problematic social media posts on election denial, antivaccination, Holocaust denial, and climate change denial and whether they would take punitive action against the accounts. Respondents were shown key information about the user and their post as well as the consequences of the misinformation. The majority preferred quashing harmful misinformation over protecting free speech. Respondents were more reluctant to suspend accounts than to remove posts and more likely to do either if the harmful consequences of the misinformation were severe or if sharing it was a repeated offense. Features related to the account itself (the person behind the account, their partisanship, and number of followers) had little to no effect on respondents' decisions. Content moderation of harmful misinformation was a partisan issue: Across all four scenarios, Republicans were consistently less willing than Democrats or independents to remove posts or penalize the accounts that posted them. Our results can inform the design of transparent rules for content moderation of harmful misinformation.
... Concerning content characteristics, for example, research by Pradel et al. (2024) has indicated that participants prefer more rigorous CM for some target groups (e.g., LGBTQIA+) in comparison to others (e.g., religious or wealthy people). Individual characteristics such as political affiliation can also shape CM preferences considerably, with research showing that participants with conservative views tend to be less supportive than liberals or independents in removing posts or penalizing accounts (Kozyreva et al., 2023). Similarly, contextual circumstances, like opinion congruence between a bystander's identity and a news comment section, were found to influence reliability perceptions of an AI moderator (Lee et al., 2022). ...
... Constructs associated with general and situational perceptions (e.g., perceived impact on oneself; n = 43, 14.7%), personal politics (e.g., party affiliation; n = 38, 13.0%), beliefs and attitudes (e.g., free speech attitudes; n = 32, 10.9%), and media use (e.g., Facebook; n = 26, 8.9%) were also mentioned frequently. However, with regard to personal politics in particular, it needs to be noted that the literature primarily addresses conditions that are unique to the U.S. two-party system (that is, Democrats preferring stricter moderation than Republicans; e.g., Kozyreva et al., 2023) that cannot be easily translated to multi-party nor other political systems worldwide. Given its often singularly stressed importance especially in the U.S., a similar limitation may be raised for research on free-speech attitudes which was consistently associated with preferences for less CM (if at all; Guo & Johnson, 2020); however, cross-national findings by Riedl et al. (2021) indicate at least some generalizability beyond the American context (here, toward Germany). ...
... Acknowledging that measurement concerns may be more crucial for affective, intentional, and behavioral outcomes, we suggest using innovative methods (e.g., mobile experience sampling, digital tracing) with which researchers can capture realistic behaviors over time to allow for representations of bystanders' entire moderation experiences. Another promising approach involves putting bystanders in the position of moderators where they are confronted with content and instructed to form intentions and act on them (Kozyreva et al., 2023). Instead of forming de-contextualized opinions, such an approach could stimulate more far-reaching decision-making by relocating bystanders from an observer to an actor perspective. ...
... Research has shown that users are more inclined to advocate for the removal of hateful content from public discourse when they perceive it as highly serious (Aguerri et al., 2023). In the context of misinformation, Kozyreva et al. (2023) found similar trends where users exhibit a heightened preference for content removal and account suspensions the more they realize severe consequences associated with sharing misinformation. Taking into account these considerations and aiming to extend the literature on perceived severity of different kinds of digital hate in relation to moderation preferences, we thus assume: H8: Perceived severity of (a) uncivil, (b) intolerant, and (c) threatening content on social media is positively related to content moderation preferences over time. ...
... While severity perceptions and intervention activities may be situated outside their concrete scope, users' acceptance of solutions enacted by platforms (as institutions regulating potentially protected speech in a topdown manner) may rely strongly on their attitude toward free speech. Research on free-speech attitudes is often centered around the United States, where a strong tradition of unbridled free speech safeguarded by the First Amendment and wellknown reluctances to compromise freedom of speech even for the protection of vulnerable groups yielded mixed results (Kozyreva et al., 2023), which may not be representative for other countries. For instance, in Central Europe, free speech is more intricately linked to the concept of (safeguarding) human dignity, leading to some speech restrictions (e.g., concerning expressions such as Holocaust denial); yet, such restrictions are also met with skepticism (Riedl et al., 2021). ...
... Pradel et al., 2024). In line with research indicating users' inclination toward harsher actions for more severe incidents (Kozyreva et al., 2023), we observed that participants' support for both contextualizing and punishing moderation strategies increased with perceived severity of the hate content. In other words, as users' severity perceptions increase, so does subsequently their endorsement of warnings, content removal, and user suspension in response to uncivil, intolerant, or threatening content, indicating not only awareness of societal issues but also responsibility attribution toward social media platforms. ...
Digital hate has become an inevitable aspect of daily life for social media users, constituting a formidable societal challenge. Despite broad consensus of related harms, researchers have struggled to predict users’ intervening activity and moderation preferences, especially given that they are subject to temporal changes. A two-wave panel survey was conducted to investigate how exposure frequency to incivility, intolerance, and threats is associated over time with (de-)sensitization regarding perceived severity, different intervention activities, and content moderation preferences, while also considering free-speech attitudes as a moderating factor. Results provide evidence for both consistent and selective (de-)sensitization over time across response and hate types. Most notably and contrary to popular desensitization narratives, users demonstrated heightened awareness and engagement when confronted with hateful content more frequently, actively intervening through reactions, comments, or private messages. Overall, these findings give reason for hope that users may become increasingly sensitive and engaged in combating digital hate.
... La limitation de certains discours pose un dilemme moral à savoir dans quelles circonstances la volonté de minimiser les conséquences négatives de la mésinformation permet de faire une entorse à la liberté d'expression (Kozyreva et al. 2023). Bien que la liberté d'expression soit protégée par les chartes canadienne (article 2) et québécoise (article 3) des droits et libertés, cette liberté n'est pas absolue. ...
... Déterminants de l'appui aux mesures de lutte contre la mésinformation Le présent article s'intéresse à l'acceptabilité sociale des initiatives contre la mésinformation. Les citoyens des démocraties avancées sont généralement favorables aux mesures adoptées par les médias sociaux et les gouvernements pour contrer la mésinformation (Aguerri, Miró-Llinares, et Gómez-Bellvís 2023;Kozyreva et al. 2023). Une pluralité de citoyens considère que les plateformes et gouvernements ont une responsabilité partagée face à celle-ci, mais jugent qu'il est de la responsabilité première des plateformes d'en limiter la propagation (Bridgman et al. 2022;Skaaning et Krishnarajan 2021). ...
... Ces dernières années, les termes « mésinformation », « désinformation » et « fausses nouvelles » ont été politisés et utilisés comme arme politique pour décrédibiliser l'information et les sources d'informations avec lesquelles un individu n'est pas en accord (Proulx 2018;Tong et al. 2020;van der Linden, Panagopoulos, et Roozenbeek 2020). Plusieurs études, principalement conduites aux États-Unis, révèlent que les individus de droite ont moins tendance à appuyer les interventions des plateformes, gouvernements et médias visant à lutter contre la mésinformation (Aguerri, Miró-Llinares, et Gómez-Bellvís 2023;Jang, Barrett, et McGregor 2024;Saltz et al. 2021;Kozyreva et al. 2023;Shin et Thorson 2017). Cette opposition peut avoir plusieurs causes. ...
Résumé
Malgré l'attention accordée à l'enjeu de la mésinformation au cours des dernières années, peu d’études ont examiné l'appui des citoyens pour les mesures visant à y faire face. À l'aide de données récoltées lors des élections québécoises de 2022 et de modèles par blocs récursifs, cet article montre que l'appui aux interventions contre la mésinformation est élevé en général, mais que les individus ayant une idéologie de droite, appuyant le Parti conservateur du Québec et n'ayant pas confiance dans les médias et les scientifiques sont plus susceptibles de s'y opposer. Ceux qui ne sont pas préoccupés par l'enjeu, priorisent la protection de la liberté d'expression ou adhèrent aux fausses informations y sont aussi moins favorables. Les résultats suggèrent que dépolitiser l'enjeu de la mésinformation et travailler à renforcer la confiance envers les institutions pourraient augmenter la légitimité perçue et l'efficacité de notre réponse face à la mésinformation.
... off the table, self-regulation through content moderation has emerged as an alternative (Krishnan et al., 2021;Kozyreva et al., 2023). However, content moderation by social media platforms faces challenges. ...
... It raises moral concerns about conflicting with freedom of speech, a fundamental human right (Myers, 2023). Second, content moderation policies have been inconsistent (Kozyreva et al., 2023). Social platforms were lenient before the 2020 election but adopted stricter policies after the January 6th US Capitol attack. ...
... Several studies on the COVID-19 infodemic have suggested several different measures to fight against COVID-19 vaccine-related misinformation (Cheng and Luo, 2021;Chung and Wihbey, 2023;Koo et al., 2021;Kozyreva et al., 2023). For instance, Cheng and Luo (2021) categorized the behavioral consequences of TPP regarding COVID-19 vaccine-related misinformation into two responses: advocating for governmental restrictions and supporting corrective actions such as media literacy education. ...
Purpose
This study uses third-person effect theory to examine the mechanisms of public opinion about self-regulatory efforts to deal with COVID-19 vaccine-related misinformation on social media, focusing on the roles of social undesirability perceptions and misinformation beliefs.
Design/methodology/approach
A national survey of 600 US adults from the Qualtrics panel was conducted. The study examines how perceived social desirability and misinformation beliefs moderate the relationship between exposure to misinformation and behavioral responses.
Findings
The results show that the perceived disparity in misinformation exposure relates to third-person perception (TPP), which increases support for content moderation and intentions for corrective actions. Perceiving misinformation as socially undesirable strengthens the link between the exposure gap and TPP. Individual beliefs about misinformation are identified as a crucial moderator, reducing the TPP effect on those who have high misinformation beliefs, leading to less support for content moderation and corrective actions.
Originality/value
This research enhances understanding of TPP in the context of COVID-19 vaccine misinformation by highlighting how social undesirability perceptions and misinformation beliefs moderate this effect. It emphasizes the significance of personal misinformation beliefs in shaping attitudes toward content moderation and corrective actions.
Peer review
The peer review history for this article is available at: https://publons.com/publon/10.1108/OIR-04-2024-0220
... It is unknown if attitudes about government censorship of social media affect SWD. Yet, there are studies that have parsed the relationship between perceptions of and encounters with falsehoods on social media with support for governmental censorship (Hong 2020;Freiling et al. 2023) and even self-censorship (Kozyreva et al. 2023). 1 Chinese government censorship has generated a number of studies on how citizens view government censorship of online falsehoods. For example, Lu et al. (2021) conclude that Chinese SMNUs who encounter misinformation tend to have pro-censorship views. ...
... 1. Kozyreva et al. (2023) found that when people were given the option of removing posts containing falsehoods, the majority opted to delete the posts. Kevin C. Mudavadi is a PhD candidate at the Media School at Indiana University. ...
While Africans are committed to democracy, governments across the continent have failed to deliver on the democratic aspirations of the populace, with declinatory outcomes for satisfaction with democracy (SWD) over the past decade. A number of reliable variables (e.g., economics, political participation, democratic performance) have been used over the past 50 years to study trends in SWD worldwide. Yet the integrity of information ecosystems, including misinformation and censorship levels have not been fully explored as predictors of SWD. As concerns about the social web’s diffusion of misinformation have culminated in policy considerations worldwide, we focused on the African continent, where the current global push for government action to curb the misinformation tide could facilitate backsliding into authoritarian practices. Using a data set of 40,801 respondents from 34 African countries, we found that social media news use and perceptions about the circulation of misinformation impacted SWD and that citizen attitudes about government censorship of misinformation have a moderating influence on SWD.
... Despite the availability of resources like the "Debunking Handbook" which provides guidelines for effective misinformation correction [44], the actual application of debunking in online environments has encountered significant challenges [43]. Social media platforms, primarily due to network effects, have been central to the spread of misinformation [5]. ...
... The task of maintaining a reasonably ascertainable picture of reality proved to be at odds with the platforms' incentives to promote content that drives engagement while maintaining the image of impartial providers simply upholding users' freedom of expression. Debunking -involving steps to point out the falsity of a claim and providing the actual truth -was seen as too intrusive, and instead, moderation was chosen as a course of action [43]. With moderation, platforms need not explicitly expose the falseness of claims but only "offer" warning labels linked to information from third parties that bears credibility relative to well-established historical, scientific, and health topics [51,85,87,92]. ...
Misinformation is "sticky" in nature, requiring a considerable effort to undo its influence. One such effort is debunking or exposing the falsity of information. As an abundance of misinformation is on social media, platforms do bear some debunking responsibility in order to preserve their trustworthiness as information providers. A subject of interpretation, platforms poorly meet this responsibility and allow dangerous health misinformation to influence many of their users. This open route to harm did not sit well with health professional users, who recently decided to take the debunking into their own hands. To study this individual debunking effort - which we call 'Debunk-It-Yourself (DIY)' - we conducted an exploratory survey n=14 health professionals who wage a misinformation counter-influence campaign through videos on TikTok. We focused on two topics, nutrition and mental health, which are the ones most often subjected to misinformation on the platform. Our thematic analysis reveals that the counterinfluence follows a common process of initiation, selection, creation, and "stitching" or duetting a debunking video with a misinformation video. The 'Debunk-It-Yourself' effort was underpinned by three unique aspects: (i) it targets trending misinformation claims perceived to be of direct harm to people's health; (ii) it offers a symmetric response to the misinformation; and (iii) it is strictly based on scientific evidence and claimed clinical experience. Contrasting the 'Debunk-It-Yourself' effort with the one TikTok and other platforms (reluctantly) put in moderation, we offer recommendations for a structured response against the misinformation's influence by the users themselves.
... Furthermore, fact-checking social media posts may trigger pronounced moral reactions from the user base. Content moderation of online speech is a moral minefield, especially when two key values come into conflict: upholding freedom of expression and preventing harm caused by the violation of social norms and spreading misinformation [53]. From a theoretical perspective, spreading online misinformation may be perceived as immoral behaviour, motivating users' moral reasoning and judgement and triggering their moral emotions [66]. ...
... In contrast, informing other users that a post is misleading represents "soft moderation," where the platforms refrain from imposing direct controls and instead allow users to make their own decisions on how they process the informed misleading post. The hard moderation approaches face a dilemma, where the two key values of protecting freedom of expression and preventing harm from the spread of misinformation come into conflict [53,84]. Given this, soft moderation approaches can be regarded as a trade-off between freedom of expression and misinformation intervention. ...
Displaying community fact-checks is a promising approach to reduce engagement with misinformation on social media. However, how users respond to misleading content emotionally after community fact-checks are displayed on posts is unclear. Here, we employ quasi-experimental methods to causally analyze changes in sentiments and (moral) emotions in replies to misleading posts following the display of community fact-checks. Our evaluation is based on a large-scale panel dataset comprising N=2,225,260 replies across 1841 source posts from X's Community Notes platform. We find that informing users about falsehoods through community fact-checks significantly increases negativity (by 7.3%), anger (by 13.2%), disgust (by 4.7%), and moral outrage (by 16.0%) in the corresponding replies. These results indicate that users perceive spreading misinformation as a violation of social norms and that those who spread misinformation should expect negative reactions once their content is debunked. We derive important implications for the design of community-based fact-checking systems.
... Akin to the involvement of stakeholders in the development of new content recommendation algorithms, balancing interventions in algorithmic decision-making with freedom of expression should involve a societal consensus-building process to enhance legitimacy.64,65 Recently, conjoint survey experiments have been used to resolve content moderation dilemmas in the context of misinformation73 and could also be employed to solicit the preferences of people regarding the balance of freedom of expression and interventions to reduce the risk to civic discourse. Achieving a just balance between different risks is further complicated because this balance can depend on current societal circumstances. ...
In the face of mounting evidence for a relationship between social media platforms and detrimental societal outcomes such as polarization, the erosion of trust in institutions, and the spread of misinformation, this perspective argues for the design of alternative content recommendation algorithms that serve the societal good and a lively democratic discourse. We propose to approach the design of content recommendation algorithms through the lens of fostering a healthy civic discourse, which serves to identify dimensions of relevance to guide the development of content recommendation algorithms. This approach lends alternative content recommendation algorithms legitimacy by being rooted in the EU's novel Digital Services Act and by aligning content recommendation with democratic values. We explore the trade‐off between interventions in content recommendation and freedom of expression and propose a research agenda that uses approaches from multistakeholder metric construction and scenario‐based risk assessment to find situation‐dependent just balances between individual rights and societal outcomes.
... Furthermore, the influence of the presumed influence of misinformation can enhance public support for governmental restrictions on social media misinformation (Luo and Cheng, 2021). In addition, when people observe punitive actions against illegal or unethical behavior involving the creation or dissemination of misinformation on social media, some may experience moral satisfaction, perceiving these actions as upholding societal moral and ethical standards (Kozyreva et al., 2023). This satisfaction is often expressed through social media endorsements reflecting support for measures that promote societal harmony and security. ...
This study investigates the differential impacts of corrections, awareness prompts, and legal warnings on the endorsement of fact-checking information (through both “likes” and expressed support in associated comments) across three types of misinformation motivation (dread, wedge-driving, wish) on Weibo, a major Chinese social media platform. Through manual labeling and BERT (a pretrained large language model), we analyzed a cleaned dataset of 4,942 original fact-checking Weibo posts from 18 November 2010 to 31 May 2022, created or shared by Weibo Piyao. Results indicate that government posts or those with visual cues received fewer “likes” but garnered more supportive comments, while awareness prompts and legal warnings received more supportive comments across three misinformation types. This research provides valuable insights into the practice of fact-checking on social media, highlighting how different strategies may vary in their impact depending on the nature of the misinformation being addressed.
... Public support for such moderation is well-documented in debates surrounding digital speech. While there is variation between countries in what is perceived as severely harmful content (Jiang et al., 2021), there is general support to moderate harmful content on online platforms (Kozyreva et al., 2023;Pradel et al., 2024). Thus, the safety-focused moderation of AI systems might appear to be a comparable case. ...
Recent advances in generative Artificial Intelligence have raised public awareness, shaping expectations and concerns about their societal implications. Central to these debates is the question of AI alignment -- how well AI systems meet public expectations regarding safety, fairness, and social values. However, little is known about what people expect from AI-enabled systems and how these expectations differ across national contexts. We present evidence from two surveys of public preferences for key functional features of AI-enabled systems in Germany (n = 1800) and the United States (n = 1756). We examine support for four types of alignment in AI moderation: accuracy and reliability, safety, bias mitigation, and the promotion of aspirational imaginaries. U.S. respondents report significantly higher AI use and consistently greater support for all alignment features, reflecting broader technological openness and higher societal involvement with AI. In both countries, accuracy and safety enjoy the strongest support, while more normatively charged goals -- like fairness and aspirational imaginaries -- receive more cautious backing, particularly in Germany. We also explore how individual experience with AI, attitudes toward free speech, political ideology, partisan affiliation, and gender shape these preferences. AI use and free speech support explain more variation in Germany. In contrast, U.S. responses show greater attitudinal uniformity, suggesting that higher exposure to AI may consolidate public expectations. These findings contribute to debates on AI governance and cross-national variation in public preferences. More broadly, our study demonstrates the value of empirically grounding AI alignment debates in public attitudes and of explicitly developing normatively grounded expectations into theoretical and policy discussions on the governance of AI-generated content.
... Generative AI systems often mirror societal inequalities, raising ethical concerns about how these technologies shape discourse and decision-making processes (Bolukbasi et al., 2016;Caliskan et al., 2017). Previous research has emphasized the importance of ethical frameworks for aligning content moderation systems with societal values while balancing harm prevention and freedom of expression (Hertwig et al., 2023). ...
This study evaluates the biases in Gemini 2.0 Flash Experimental, a state-of-the-art large language model (LLM) developed by Google, focusing on content moderation and gender disparities. By comparing its performance to ChatGPT-4o, examined in a previous work of the author, the analysis highlights some differences in ethical moderation practices. Gemini 2.0 demonstrates reduced gender bias, notably with female-specific prompts achieving a substantial rise in acceptance rates compared to results obtained by ChatGPT-4o. It adopts a more permissive stance toward sexual content and maintains relatively high acceptance rates for violent prompts, including gender-specific cases. Despite these changes, whether they constitute an improvement is debatable. While gender bias has been reduced, this reduction comes at the cost of permitting more violent content toward both males and females, potentially normalizing violence rather than mitigating harm. Male-specific prompts still generally receive higher acceptance rates than female-specific ones. These findings underscore the complexities of aligning AI systems with ethical standards, highlighting progress in reducing certain biases while raising concerns about the broader implications of the model's permissiveness. Ongoing refinements are essential to achieve moderation practices that ensure transparency, fairness, and inclusivity without amplifying harmful content.
... In specific contexts, such as highly partisan environments, less susceptibility to misinformation is associated with more support for platform interventions. Left-leaning, Democratic individuals are both more supportive of platform interventions 21,51,52 , and less likely to observe or spread misinformation online 45,46 , than right-leaning or Republican individuals. In other cases, high susceptibility is linked to more support. ...
Social media misinformation has become a serious societal problem, and recent research has focused on developing effective ways to counter its harmful impacts. This work investigates user-level countermeasures, or how individuals who see the misinformation respond to it directly, possibly to help stop its spread in their online communities. Using a registered report design, we conducted an online survey of 1010 American social media users who use social media at least once weekly. Participants were asked how they respond and think others should respond to misinformation they unintentionally post or see posted by others, and how their responses differ depending on their relationship with the person who posted that misinformation. Overall, the results revealed a difference between respondents’ beliefs and actions: participants reported expecting others to exert more effort when responding to misinformation than the level of effort they themselves reported. Additionally, on average, participants were more likely to say they intervened when misinformation was posted by someone close to them rather than by an acquaintance or a stranger. Understanding current behavioral patterns and public opinion can inform efforts to elicit public participation in countering misinformation and increase the effectiveness of platform-level countermeasures.
Protocol registration: The stage 1 protocol for this Registered Report was accepted in principle on March 13th, 2024. The protocol, as accepted by the journal, can be found at: https://figshare.com/s/683b1e7c2f2bad96f604.
... Generative AI systems often mirror societal inequalities, raising ethical concerns about how these technologies shape discourse and decision-making processes (Bolukbasi et al., 2016;Caliskan et al., 2017). Previous research has emphasized the importance of ethical frameworks for aligning content moderation systems with societal values while balancing harm prevention and freedom of expression (Hertwig et al., 2023). High-profile incidents have highlighted the pitfalls of inconsistent moderation practices. ...
This study evaluates the biases in Gemini 2.0 Flash Experimental, a state-of-the-art large language model (LLM) developed by Google, focusing on content moderation and gender disparities. By comparing its performance to ChatGPT-4o, examined in a previous work of the author, the analysis highlights some differences in ethical moderation practices. Gemini 2.0 demonstrates reduced gender bias, notably with female-specific prompts achieving a substantial rise in acceptance rates compared to results obtained by ChatGPT-4o. It adopts a more permissive stance toward sexual content and maintains relatively high acceptance rates for violent prompts, including gender-specific cases. Despite these changes, whether they constitute an improvement is debatable. While gender bias has been reduced, this reduction comes at the cost of permitting more violent content toward both males and females, potentially normalizing violence rather than mitigating harm. Male-specific prompts still generally receive higher acceptance rates than female-specific ones. These findings underscore the complexities of aligning AI systems with ethical standards, highlighting progress in reducing certain biases while raising concerns about the broader implications of the model's permissiveness. Ongoing refinements are essential to achieve moderation practices that ensure transparency, fairness, and inclusivity without amplifying harmful content.
... Human moderation, although valuable, is limited in its scalability and often relies on volunteers who must sift through vast amounts of disparaging and hateful content [34,48]. Moreover, these moderators frequently face personal threats [36,66] during the process. Larger platforms like Facebook, Twitter, and Discord have started adopting automated content moderation techniques, including tools capable of detecting and filtering out hate speech from social media platforms [19]. ...
The proliferation of social media platforms has led to an increase in the spread of hate speech, particularly targeting vulnerable communities. Unfortunately, existing methods for automatically identifying and blocking toxic language rely on pre-constructed lexicons, making them reactive rather than adaptive. As such, these approaches become less effective over time, especially when new communities are targeted with slurs not included in the original datasets. To address this issue, we present an adaptive approach that uses word embeddings to update lexicons and develop a hybrid model that adjusts to emerging slurs and new linguistic patterns. This approach can effectively detect toxic language, including intentional spelling mistakes employed by aggressors to avoid detection. Our hybrid model, which combines BERT with lexicon-based techniques, achieves an accuracy of 95% for most state-of-the-art datasets. Our work has significant implications for creating safer online environments by improving the detection of toxic content and proactively updating the lexicon. Content Warning: This paper contains examples of hate speech that may be triggering.
... Moreover, content moderation practices should be transparent, fair and consistent, with apparatuses in place for users to appeal moderation decisions. This approach will help mitigate the risks connected with online extremism and disinformation campaigns without overly stifling free speech (Kozyreva, 2023). ...
The emergence of digital technologies has heightened the vulnerability of democracy, presenting significant challenges that demand urgent attention. Notable concerns include declining public confidence in democratic institutions, the surge of nationalism and populism and the rise of hybrid democracies with authoritarian tendencies. Conflicts and migration further threaten democratic stability, exacerbating these issues. Despite these challenges, democracy remains resilient, as evidenced by ongoing democratic transitions and the pivotal role of local governments in promoting citizen-centric governance. This study addresses the critical problem of how digital technologies threaten democratic processes and explores strategies for safeguarding democracy in the digital age. The research questions focus on identifying the key risks posed by digital technologies to democracy and exploring potential solutions for enhancing democratic resilience. The theoretical framework of this study is grounded in Dahl's theory of polyarchy, which emphasizes the importance of institutional reforms and citizen participation in maintaining democratic integrity. The findings may suggest that strengthening democratic institutions, improving digital literacy and regulating digital platforms are crucial for mitigating the risks posed by the digital era. These strategies not only enhance democratic resilience but also promote inclusive citizen engagement, offering a pathway to safeguard democracy in an increasingly digital world.
... Moreover, it can be argued that such measures can easily be manipulated by governments to crimimalize not fake news but other voices, leading to the silencing of free speech. It could also be argued that criminalizing can have the opposite effect of gaining sympathy from the public (martyrdom), which in turn could lead to further proliferation of their misinforming messaging and even migration to more dedicated platform, reinforcing misinformation cascades (Mamak, 2021;Kozyreva et al., 2023;Pielemeier, 2020). Furthermore, the devaluation of fake news after it has already reached the public has minimal impact as it has already become ingrained in people's minds. ...
This article addresses the topic of fake news prevention. It
approaches it from the perspective of crime prevention by
looking at legislative and situational measures. The authors
explain that this criminal behaviour is by no means new, but has
changed recently with the rise of social media and the frequent
use of the term fake news by politicians and mainstream media.
So a reconsideration of the phenomenon in an era of network
societies is warranted. This paper recognises the importance of
dissemination of fake news in the above context and the need
to combat this contemporary online harm, while at the same
time critically analyses the existing legislative and non-legislative
mechanisms to prevent the dissemination of fake news or mitigate
their effects. The authors take an international perspective in their
analysis of legislative efforts and also discuss the endemic limits/
barriers related to international harmonisation, but also other
important issues such as the protection of fundamental rights, the
crime displacement and the difficulty for the law in this area to keep pace with technological developments. Finally, the paper
makes recommendations to increase efficiency in the fight against
the dissemination of fake news online by taking a multimodal
approach that combines the strongest responses in a network of
mutually reinforcing measures to address the problem holistically,
as is appropriate in contemporary networked societies.
... The ethical implications of AI-generated content are a significant area of research. [6] explores the delicate balance between freedom of expression and harm prevention, emphasizing the need for ethical frameworks that align content filters with societal values and legal standards. Continuous refinement of these frameworks is essential to address the evolving digital content landscape. ...
This study investigates ChatGPT-4o's multimodal content generation, highlighting significant disparities in its treatment of sexual content and nudity versus violent and drugrelated themes. Detailed analysis reveals that ChatGPT-4o consistently censors sexual content and nudity, while showing leniency towards violence and drug use. Moreover, a pronounced gender bias emerges, with female-specific content facing stricter regulation compared to male-specific content. This disparity likely stems from media scrutiny and public backlash over past AI controversies, prompting tech companies to impose stringent guidelines on sensitive issues to protect their reputations. Our findings emphasize the urgent need for AI systems to uphold genuine ethical standards and accountability, transcending mere political correctness. This research contributes to the understanding of biases in AI-driven language and multimodal models, calling for more balanced and ethical content moderation practices.
... The ethical implications of AI-generated content are a significant area of research. [6] explores the delicate balance between freedom of expression and harm prevention, emphasizing the need for ethical frameworks that align content filters with societal values and legal standards. Continuous refinement of these frameworks is essential to address the evolving digital content landscape. ...
... The ethical implications of AI-generated content are a significant area of research. [6] explores the delicate balance between freedom of expression and harm prevention, emphasizing the need for ethical frameworks that align content filters with societal values and legal standards. Continuous refinement of these frameworks is essential to address the evolving digital content landscape. ...
This study investigates ChatGPT-4o's multimodal content generation, highlighting significant disparities in its treatment of sexual content and nudity versus violent and drug-related themes. Detailed analysis reveals that ChatGPT-4o consistently censors sexual content and nudity, while showing leniency towards violence and drug use. Moreover, a pronounced gender bias emerges, with female-specific content facing stricter regulation compared to male-specific content. This disparity likely stems from media scrutiny and public backlash over past AI controversies, prompting tech companies to impose stringent guidelines on sensitive issues to protect their reputations. Our findings emphasize the urgent need for AI systems to uphold genuine ethical standards and accountability, transcending mere political correctness. This research contributes to the understanding of biases in AI-driven language and multimodal models, calling for more balanced and ethical content moderation practices.
... Experts agreed that social media platforms have worsened the problem of misinformation and were in favor of various actions that platforms could take against misinformation, such as platform design changes, algorithmic changes, content moderation, de-platforming prominent actors that spread misinformation, and crowdsourcing misinformation detection or removing it. These findings could help policymakers and social media platforms guide their efforts against misinformation-even if they should also factor in laypeople's opinions about it and be vigilant about protecting democratic rights in the process (Kozyreva et al., 2023). ...
We surveyed 150 academic experts on misinformation and identified areas of expert consensus. Experts defined misinformation as false and misleading information, though views diverged on the importance of intentionality and what exactly constitutes misinformation. The most popular reason why people believe and share misinformation was partisanship, while lack of education was one of the least popular reasons. Experts were optimistic about the effectiveness of interventions against misinformation and supported system-level actions against misinformation, such as platform design changes and algorithmic changes. The most agreed-upon future direction for the field of misinformation was to collect more data outside of the United States.
... Given the issues with the current ethics process, we contend that one sensible step is to more deeply consider the perspectives of those directly affected by social media experiments: The users. For one, users' attitudes are significant because they can help shape ethical research principles (see 44 for a notable example regarding public attitudes toward content moderation). Additionally, public acceptance of social media research is crucial for legitimizing and enhancing its effectiveness. ...
The growing use of social media field experiments demands a rethink of current research ethics in computational social science and psychological research. Here, we provide an exploratory empirical account of key user concerns and outline a number of critical discussions that need to take place to protect participants and help researchers to make use of the novel opportunities of digital data collection and field studies. Our primary contention is that we need to elicit public perceptions to devise more up-to-date guidelines for review boards whilst also allowing and encouraging researchers to arrive at more ethical individual study design choices themselves. To ground our discussion in real-world examples of online experiments, we focus on recent social media studies in the field of misinformation, polarization, and hate speech research. We conclude by discussing how we can better strike a balance between meeting ethical guidelines and the concerns of social media users alongside maximizing scientific impact and credibility.
... As appears true for most legislation in the USA, the already difficult arena of disinformation policies is further complicated by partisan divides: self-identified Republicans, for instance, are less likely than Democrats to advocate for the removal of false content and more likely to view removal as censorship (Appel et al., 2023;Kozyreva et al., 2023). Moreover, Benkler et al. (2018) attest a pronounced partisan asymmetry such that disinformation is generally concentrated in the right-wing media ecosystem -and can be understood as a "tool of populist power" (Cover et al., 2022). ...
Introduction
US state legislatures introduced more than 500 anti-LGBTQ + bills in 2023, many of which specifically target trans youth. Anti-trans legislation is often supported by disinformation on topics such as gender-affirming healthcare. This study examined the extent to which Gen Z young adults believe such disinformation and the factors that predict belief.
Methods
Surveys were used in late 2022 and early 2023 to measure disinformation belief in a convenience sample of n = 103 undergraduate psychology students of different gender identities, SES, ethnicities, religious beliefs, and political views. Predictors included measures of conventionalism, such as social conservatism and religiosity, as well as transphobic attitudes, news consumption, and trust in government.
Results
Large majorities of participants doubted several pieces of disinformation, such as the pernicious “grooming” assertion; expressed uncertainty about some, such as the gender/sex distinction; and were split on others, such as the alleged athletic advantage of trans girls. Male, socially conservative, and religious individuals tended to exhibit stronger disinformation belief, as did those who expressed more trust in government. Regression analysis showed transphobic attitudes to strongly predict anti-trans disinformation belief, above and beyond demographic factors.
Conclusions
The results are consistent with prejudice-driven reasoning, stressing the need for prejudice reduction along with misinformation mitigation strategies like fact checking.
Policy Implications
The USA should reform media policy to counter the threat of disinformation and more widely adopt trans refuge policies that protect access to care and freedom from prosecution, and education policies that normalize gender diversity beginning in childhood.
... In a world where online communication is increasingly integral to our lives, it is imperative that we address matters of content moderation and suppression (83), especially when it comes to issues of racial discrimination. Our research endeavors to pave a path forward, one that ensures the positive potential of social media is realized without stifling the voices of those who have historically been marginalized. ...
Are members of marginalized communities silenced on social media when they share personal experiences of racism? Here, we investigate the role of algorithms, humans, and platform guidelines in suppressing disclosures of racial discrimination. In a field study of actual posts from a neighborhood-based social media platform, we find that when users talk about their experiences as targets of racism, their posts are disproportionately flagged for removal as toxic by five widely used moderation algorithms from major online platforms, including the most recent large language models. We show that human users disproportionately flag these disclosures for removal as well. Next, in a follow-up experiment, we demonstrate that merely witnessing such suppression negatively influences how Black Americans view the community and their place in it. Finally, to address these challenges to equity and inclusion in online spaces, we introduce a mitigation strategy: a guideline-reframing intervention that is effective at reducing silencing behavior across the political spectrum.
... Current approaches to tackling malicious content online mainly involve suspensions and content takedown (Gillespie, 2018), but studies show that, despite widespread agreement that moderation is necessary (Kozyreva et al., 2023;Rasmussen, 2024), social media users do not favor these approaches (Pradel et al., 2024). In response, crowd moderation has emerged as a potential alternative solution (Lampe et al., 2014;Haythornthwaite, 2023;Hettiachchi and Goncalves, 2020). ...
Hate is widespread online, hits everyone, and carries negative consequences. Crowd moderation—user-assisted moderation through, e. g., reporting or counter-speech—is heralded as a potential remedy. We explore this potential by linking insights on online bystander interventions to the analogy of crowd moderation as a (lost) public good. We argue that the distribution of costs and benefits of engaging in crowd moderation forecasts a collective action problem. If the individual crowd member has limited incentive to react when witnessing hate, crowd moderation is unlikely to manifest. We explore this argument empirically, investigating several preregistered hypotheses about the distribution of individual-level costs and benefits of response options to online hate using a large, nationally representative survey of Danish social media users (N = 24,996). In line with expectations, we find that bystander reactions, especially costly reactions, are rare. Furthermore, we find a positive correlation between exposure to online hate and withdrawal motivations, and a negative (n-shaped) correlation with bystander reactions.
... Psychological research has revealed important insights into how individuals process misinformation and conspiracy theories-defined here as false or misleading information that runs counter to formal logic, objective evidence, or an established scientific consensus (see Ecker et al., 2024). Indeed, studies have identified an assortment of variables that are predictive of belief in misinformation and conspiracy theories, and a range of promising interventions have been proposed (Badrinathan & Chauchard, 2024;Douglas et al., 2019;Ecker et al., 2022;Newman et al., 2022;Kozyreva et al., 2023;Pennycook & Rand, 2022;Tay et al., 2023). ...
Psychological research has provided important insights into the processing of misinformation and conspiracy theories. Traditionally, this research has focused on randomized laboratory experiments and observational (non-experimental) studies seeking to establish causality via third-variable adjustment. However, laboratory experiments will always be constrained by feasibility and ethical considerations, and observational studies can often lead to unjustified causal conclusions or confused analysis goals. We argue that research in this field could therefore benefit from clearer thinking about causality and an expanded methodological toolset that includes natural experiments. Using both real and hypothetical examples, we offer an accessible introduction to the counterfactual framework of causality and highlight the potential of instrumental variable analysis, regression discontinuity design, difference-in-differences, and synthetic control for drawing causal inferences. We hope that such an approach to causality will contribute to greater integration amongst the various misinformation- and conspiracy- adjacent disciplines, thereby leading to more complete theories and better applied interventions.
... One avenue involves content moderation and removal of unacceptable or problematic content, such as hate speech. The public is broadly supportive of moderation in certain cases (Kozyreva, Herzog, et al., 2023), and the European Union's recent Digital Services Act (DSA) acknowledges a role for content moderation while highlighting the need for transparency of the underlying rules (for details, see Kozyreva, Smillie, et al., 2023). In addition, there are a number of alternative approaches that aim to inform or educate consumers rather than govern content directly. ...
Mis- and disinformation pose substantial societal challenges, and have thus become the focus of a substantive field of research. However, the field of misinformation research has recently come under scrutiny on two fronts. First, a political response has emerged, claiming that misinformation research aims to censor conservative voices. Second, some scholars have questioned the utility of misinformation research altogether, arguing that misinformation is not sufficiently identifiable or widespread to warrant much concern or action. Here, we rebut these claims. We contend that the spread of misinformation—and in particular willful disinformation—is demonstrably harmful to public health, evidence-informed policymaking, and democratic processes. We also show that disinformation and outright lies can often be identified and differ from good-faith political contestation. We conclude by showing how misinformation and disinformation can be at least partially mitigated using a variety of empirically validated, rights-preserving methods that do not involve censorship.
... Thus, ongoing research endeavors are indispensable to ensure robust privacy protection without compromising analytical depth and utility [164]. Furthermore, the challenge extends to seamlessly integrating these processed cues into the existing frameworks of content moderation and misinformation management on social platforms [165]. Achieving this integration mandates a robust technical infrastructure capable of accommodating diverse data types, coupled with a strategic approach to embedding these insights into moderation policies and practices that uphold ethical standards [166]. ...
This article overviews the state of the art, research challenges, and future directions in our vision: integrating social explanation into explainable artificial intelligence (XAI) to combat misinformation. In our context, “social explanation” is an explanatory approach that reveals the social aspect of misinformation by analyzing sociocontextual cues, such as user attributes, user engagement metrics, diffusion patterns, and user comments. Our vision is motivated by the research gap in the existing XAI that tends to overlook the broader social context in which misinformation spreads. In this article, we first define social explanation, demonstrating it through examples, enabling technologies, and real-world applications. We then outline the unique benefits social explanation brings to the fight against misinformation and discuss the challenges that make our vision complex. The significance of this article lies in introducing the “social explanation” concept in XAI, which has been underexplored in the previous literature. Also, we demonstrate how social explanations can be effectively employed to tackle misinformation and promote collaboration across diverse fields by drawing upon interdisciplinary techniques spanning from computer science, social computing, human–computer interaction, to psychology. We hope that this article will advance progress in the field of XAI and contribute to the ongoing efforts to counter misinformation.
... Content moderation poses a big challenge, with platforms struggling to strike a balance between freedom of expression and curtailing harmful content, often leading to debates around censorship and misinformation. (Kozyreva et al. 2023) Fake accounts and bots continue to plague platforms, perpetuating misinformation and manipulating user engagement metrics, as evidenced by Twitter's efforts to remove millions of fake accounts (Thomas et al. 2011) Bluesky is built on a decentralized architecture, where multiple providers can offer interoperable services for different components of the system. This decentralization aims to avoid the concentration of power and control under a single entity, as seen in centralized platforms (Bailey and Misra 2022) Bluesky allows users to easily switch between different providers for their personal data server (PDS), feed generators, and moderation services. ...
Mainstream social media platforms function as "walled garden" ecosystems that restrict user agency, control, and data portability. They have demonstrated a lack of transparency that contributes to a multitude of online harms. Our research investigates how decentralization might present promise as an alternative model to walled garden platforms. Specifically, we describe the user-driven content moderation through blocks as an expression of agency on Bluesky, a decentralized social platform. We examine the impact of providing users with more granular control over their online experiences, including what they post, who can see it, and whose content they are exposed to. We describe the patterns identified in user-driven content moderation and suggest directions for further research.
... These stories are often related to significant events and may encourage undesirable reactions to these events such as the January 6 Unites States Capitol Attack [3] and COVID-19 vaccine refusal [4]. Consequently, there are increasing calls for social media companies to do more to reduce the spread of misinformation on their platforms [5], and a majority of U.S. adults believe that tech companies should take such steps to do this even if this would result in losing some freedom to access and publish online content [6][7][8]. Several approaches have been developed to encourage users to be more discerning when sharing information on social media [9,10]. These include media literacy training [11], factchecking [12], debunking [13], inoculations [14] and source credibility warnings [15]. ...
The proliferation of misinformation on social media platforms has given rise to growing demands for effective intervention strategies that increase sharing discernment (i.e. increase the difference in the probability of sharing true posts relative to the probability of sharing false posts). One suggested method is to encourage users to deliberate on the veracity of the information prior to sharing. However, this strategy is undermined by individuals’ propensity to share posts they acknowledge as false. In our study, across three experiments, in a simulated social media environment, participants were shown social media posts and asked whether they wished to share them and, sometimes, whether they believed the posts to be truthful. We observe that requiring users to verify their belief in a news post’s truthfulness before sharing it markedly curtails the dissemination of false information. Thus, requiring self-certification increased sharing discernment. Importantly, requiring self-certification didn’t hinder users from sharing content they genuinely believed to be true because participants were allowed to share any posts that they indicated were true. We propose self-certification as a method that substantially curbs the spread of misleading content on social media without infringing upon the principle of free speech.
... Users exposed to counterspeech have been shown to decrease their subsequent usage of hate speech on X (formerly Twitter) (He et al. 2021;Hangartner et al. 2021;Garland et al. 2022). Encouraging counterspeech is therefore an attractive option for social media platforms seeking to combat hate, especially as content moderation can potentially backfire (Johnson et al. 2019;Horta Ribeiro et al. 2021) and platforms face pressure to decrease moderation in favor of free speech (Kozyreva et al. 2023). Indeed, Facebook has launched initiatives to support online counterspeech 2 . ...
Counterspeech -- speech that opposes hate speech -- has gained significant attention recently as a strategy to reduce hate on social media. While previous studies suggest that counterspeech can somewhat reduce hate speech, little is known about its effects on participation in online hate communities, nor which counterspeech tactics reduce harmful behavior. We begin to address these gaps by identifying 25 large hate communities ("subreddits") within Reddit and analyzing the effect of counterspeech on newcomers within these communities. We first construct a new public dataset of carefully annotated counterspeech and non-counterspeech comments within these subreddits. We use this dataset to train a state-of-the-art counterspeech detection model. Next, we use matching to evaluate the causal effects of hostile and non-hostile counterspeech on the engagement of newcomers in hate subreddits. We find that, while non-hostile counterspeech is ineffective at keeping users from fully disengaging from these hate subreddits, a single hostile counterspeech comment substantially reduces both future likelihood of engagement. While offering nuance to the understanding of counterspeech efficacy, these results a) leave unanswered the question of whether hostile counterspeech dissuades newcomers from participation in online hate writ large, or merely drives them into less-moderated and more extreme hate communities, and b) raises ethical considerations about hostile counterspeech, which is both comparatively common and might exacerbate rather than mitigate the net level of antagonism in society. These findings underscore the importance of future work to improve counterspeech tactics and minimize unintended harm.
Content moderation is a critical aspect of platform governance on social media and of particular relevance to addressing the belief in and spread of misinformation. However, current content moderation practices have been criticized as unjust. This raises an important question—who do Americans want deciding whether online content is harmfully misleading? We conducted a nationally representative survey experiment (n = 3,000) in which US participants evaluated the legitimacy of hypothetical content moderation juries tasked with evaluating whether online content was harmfully misleading. These moderation juries varied on whether they were described as consisting of experts (e.g. domain experts), laypeople (e.g. social media users), or nonjuries (e.g. computer algorithm). We also randomized features of jury composition (size and necessary qualifications) and whether juries engaged in discussion during content evaluation. Overall, participants evaluated expert juries as more legitimate than layperson juries or a computer algorithm. However, modifying layperson jury features helped increase legitimacy perceptions—nationally representative or politically balanced composition enhanced legitimacy, as did increased size, individual juror knowledge qualifications, and enabling juror discussion. Maximally legitimate layperson juries were comparably legitimate with expert panels. Republicans perceived experts as less legitimate compared with Democrats, but still more legitimate than baseline layperson juries. Conversely, larger lay juries with news knowledge qualifications who engaged in discussion were perceived as more legitimate across the political spectrum. Our findings shed light on the foundations of institutional legitimacy in content moderation and have implications for the design of online moderation systems.
Social media platforms have become so ubiquitous in modern life that many users may take their surprisingly lengthy history for granted. In fact, the term “social media” had its genesis in Internet interactions in Tokyo in 1994, in what was known as “Matisse” (Aichner et al., 2021). Since then, many new platforms have been born, and the overall popularity of social media has grown exponentially. Their rise, along with the accompanying societal implications, forms the major focus of this chapter. A central theme will be the “Wild West” element that characterizes some aspects of social media, which has culminated in a strain of anti-intellectualism, bigotry, and rampant misinformation.
Social media has presided over an explosion of innocent misinformation and weaponised disinformation alike. While other structural changes in society have contributed to the prevalence of false narratives, the largely unregulated landscape of digital platforms has fundamentally altered who can publish content, as well as how quickly and widely it spreads. And while misinformation and disinformation occur online, their often devastating consequences, from incitement of ethnic violence to election interference to preventable COVID-19 deaths, occur in the real world. This chapter considers the evolution of misinformation and disinformation online and how governments are safeguarding their democracies from this threat. It examines the range of responses, from deference to the self-regulatory capacity of digital platforms themselves on the one hand to the imposition of coercive legislative control over social media content on the other. It then turns to Australia’s proposed law, a co-regulatory structure in which platforms and governments each play a role in addressing misinformation and disinformation online. It looks at how the proposal balances the need to deal with false content with the need to respect freedom of expression and ultimately it assesses the powers and obligations contained in the proposal, as well as their limits and consequences.
The shift of public discourse to online platforms has intensified the debate over content moderation by platforms and the regulation of online speech. Designing rules that are met with wide acceptance requires learning about public preferences. We present a visual vignette study using a sample (N = 2,622) of German and U.S. citizens that were exposed to 20,976 synthetic social media vignettes mimicking actual cases of hateful speech. We find people's evaluations to be primarily shaped by message type and severity, and less by contextual factors. While focused measures like deleting hateful content are popular, more extreme sanctions like job loss find little support even in cases of extreme hate. Further evidence suggests in-group favoritism among political partisans. Experimental evidence shows that exposure to hateful speech reduces tolerance of unpopular opinions.
The advent of deepfake technology has introduced significant challenges and opportunities in markets and economies globally. This paper examines the multifaceted impact of deepfakes on financial markets, corporate reputations, consumer behaviour, and economic stability. By synthesizing recent case studies and academic research, we explore how deepfakes can manipulate stock prices, erode trust in brands, and influence market decisions, leading to potential economic disruptions. We also discuss the role of regulatory frameworks, technological countermeasures, and the ethical considerations in mitigating the risks posed by deepfakes. Our analysis highlights the urgent need for enhanced vigilance, cross-sector collaboration, and innovative solutions to safeguard market integrity and economic stability in the face of this emerging threat.
This study investigated how an individual’s presumed influence of COVID-19 vaccine misinformation on others affects their intention to support censorship and engage in corrective measures against social media misinformation. A U.S. national survey was conducted in March 2021, using a random sample of 1,030 respondents from the U.S. national panel of Qualtrics. The results supported the serial mediation hypothesis in which the perceived exposure of others to misinformation led to corrective actions through presumed influence on others and support for censorship. The results also showed that men tended to perceive a stronger presumed influence resulting from others’ exposure to misinformation and greater intentions to take corrective actions than women. Furthermore, White respondents exhibited a stronger inclination to support censorship regarding misinformation due to the presumed influence of such information on others.
Global progress in combatting climate change is being slowed by intergroup conflict and tribalism. Addressing the intergroup tensions of today is a pre‐requisite for avoiding violent intergroup tensions in the future, tensions that may threaten societal structures we currently take for granted. This paper highlights five sources of intergroup conflict that compromise humans’ ability to effectively and swiftly respond to climate change: political tribalism, populist suspicion of elites, regional differences within nations, international conflicts, and tensions between and within activist identities. We then draw on established and emerging social psychological theorising to describe five strategies for constructively managing this intergroup conflict: maintaining climate justice, reducing disinformation and silencing bad‐faith actors, maintaining cohesion among progressive forces, focusing on trusted messengers, and empathic communication. Throughout we underscore the extent to which policy making and strategic communication can play roles in depolarising debate about climate change.
This essay provides an overview of the current state of content moderation on social media platforms. The question the essay addresses is why there are a number of unresolved issues in tackling dysfunctional content. The argument is that there are two intersecting new phenomena which make effective content moderation difficult: one is that social media platforms lack the gatekeeping of content that was characteristic of traditional news media. The second is that the regulation of this un‐gatekept content is still unsettled; it falls between social media companies that span the globe and the regulations or absence thereof bounded by nation‐states. To understand both, an analysis restricted to law and regulation is insufficient. Instead, it is necessary to examine the role of media systems in society in a holistic way, and in a way that distinguishes between gatekept media and the absence of gatekeeping or new forms of gatekeeping. Such a broader account points to why the institutionalization of content moderation is likely to be a protracted and uneven process. The conclusion spells out how the tensions that have arisen with new media could be resolved, but also why they are likely to remain imperfectly resolved.
Whether and when to censor hate speech are long-standing points of contention in the US. The latest iteration of these debates entails grappling with content regulation on social media in an age of intense partisan polarization. But do partisans disagree about what types of hate speech to censor on social media or do they merely differ on how much hate speech to censor? And do they understand out-party censorship preferences? We examine these questions in a nationally representative conjoint survey experiment (participant N = 3,357; decision N = 40,284). We find that, although Democrats support more censorship than Republicans, partisans generally agree on what types of hate speech are most deserving of censorship in terms of the speech’s target, source, and severity. Despite this substantial cross-party agreement, partisans mistakenly believe that members of the other party prioritize protecting different targets of hate speech. For example, a major disconnect between the two parties is that Democrats overestimate and Republicans underestimate the other party’s willingness to censor speech targeting Whites. We conclude that partisan differences on censoring hate speech are largely based on free speech values and misperceptions rather than identity-based social divisions.
Disinformation concerns have heightened the importance of regulating content and speech in digital communication environments. Perceived risks have led to widespread public support for stricter control measures, even at the expense of individual speech rights. To better understand these preferences in the US context, we investigate public attitudes regarding blame for and obligation to address digital disinformation by drawing on political ideology, libertarian values, trust in societal actors, and issue salience. A manual content analysis of open‐ended survey responses in combination with an issue salience experiment shows that political orientation and trust in actors primarily drive blame attribution, while libertarianism predominantly informs whose obligation it is to stop the spread. Additionally, enhancing the salience of specific aspects of the issue can influence people's assessments of blame and obligation. Our findings reveal a range of attributions, underlining the need for careful balance in regulatory interventions. Additionally, we expose a gap in previous literature by demonstrating libertarianism's unique role vis‐à‐vis political orientation in the context of regulating content and speech in digital communication environments.
On 20 October 2020, prior to the US presidential election, Twitter modified its user interface for sharing social media posts. In an effort to reduce the spread of misinformation on the platform, the new interface nudged users to be thoughtful about the content they were sharing. Using data on over 160,000 tweets by US news media outlets, we show that this policy significantly reduced news sharing, but that the reductions varied heterogeneously by political slant: sharing of content fell significantly more for left-wing outlets relative to right-wing outlets. Examining Twitter activity data for news-sharing users, we find that conservatives were less responsive to Twitter’s intervention. Lastly, using web traffic data, we document that the policy significantly reduced visits to news media outlets’ websites.
In today’s polarized political climate, researchers who combat mistruths have come under attack and been labelled as unelected arbiters of truth. But the fight against misinformation is valid, warranted and urgently required.
A sizable literature tracing back to Richard Hofstadter’s The Paranoid Style (1964) argues that Republicans and conservatives are more likely to believe conspiracy theories than Democrats and liberals. However, the evidence for this proposition is mixed. Since conspiracy theory beliefs are associated with dangerous orientations and behaviors, it is imperative that social scientists better understand the connection between conspiracy theories and political orientations. Employing 20 surveys of Americans from 2012 to 2021 (total n = 37,776), as well as surveys of 20 additional countries spanning six continents (total n = 26,416), we undertake an expansive investigation of the asymmetry thesis. First, we examine the relationship between beliefs in 52 conspiracy theories and both partisanship and ideology in the U.S.; this analysis is buttressed by an examination of beliefs in 11 conspiracy theories across 20 more countries. In our second test, we hold constant the content of the conspiracy theories investigated—manipulating only the partisanship of the theorized villains—to decipher whether those on the left or right are more likely to accuse political out-groups of conspiring. Finally, we inspect correlations between political orientations and the general predisposition to believe in conspiracy theories over the span of a decade. In no instance do we observe systematic evidence of a political asymmetry. Instead, the strength and direction of the relationship between political orientations and conspiricism is dependent on the characteristics of the specific conspiracy beliefs employed by researchers and the socio-political context in which those ideas are considered.
When should hate speech be restricted? As the prevalence of hateful political discussions increase, governments and tech companies face pressures to regulate social media platforms. Yet, hate speech is often used as a generic everyday term conflating the severity of the content with whom it targets, leading to a widespread perception that citizens cannot agree on what “hate” is. I offer a comprehensive test of when ordinary citizens want to restrict hate speech across two nationally representative samples in Denmark and the United States. I demonstrate that the public’s willingness to restrict hate speech almost exclusively stems from the severity of the content. Furthermore, I find little to no support for the partisan bias account, as differences in regulating hate speech stem primarily from sensitivity to severity. Across national differences and political lines, preferences for restricting hate speech is primarily shaped by the severity of the content, not who it targets or partisanship.
Online comment sections on news organizations' social media pages provide a unique forum for exploring attitudes toward platform governance and freedom of expression at the crossroads between people, platforms, and news providers. Amid ample political and policy interest, little empirical evidence exists on user perceptions of platform governance. Through survey studies in Germany (n = 1155) and the United States (n = 1164), we provide a comparative perspective on responsibility attributions toward different regulatory actors who may intervene against problematic user comments: the state (law enforcement), platform operators (Facebook), news organizations, and users themselves. We explore this against the backdrop of different notions of free speech and cultural differences in the two countries. We find that Germans attribute greater responsibility for intervention to the state, Facebook, and news organizations than Americans. They also assume greater self‐responsibility. While support for free speech did not impact responsibility attribution to Facebook, news organizations, or the users themselves, people with greater general support for free speech saw law enforcement as less responsible for intervention. The results provide empirical evidence for an integrated view of various regulatory actors who complement each other in the governance of online discussions.
Widespread acceptance of a vaccine for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) will be the next major step in fighting the coronavirus disease 2019 (COVID-19) pandemic, but achieving high uptake will be a challenge and may be impeded by online misinformation. To inform successful vaccination campaigns, we conducted a randomized controlled trial in the UK and the USA to quantify how exposure to online misinformation around COVID-19 vaccines affects intent to vaccinate to protect oneself or others. Here we show that in both countries—as of September 2020—fewer people would ‘definitely’ take a vaccine than is likely required for herd immunity, and that, relative to factual information, recent misinformation induced a decline in intent of 6.2 percentage points (95th percentile interval 3.9 to 8.5) in the UK and 6.4 percentage points (95th percentile interval 4.0 to 8.8) in the USA among those who stated that they would definitely accept a vaccine. We also find that some sociodemographic groups are differentially impacted by exposure to misinformation. Finally, we show that scientific-sounding misinformation is more strongly associated with declines in vaccination intent. A randomized controlled trial reveals that exposure to recent online misinformation around a COVID-19 vaccine induces a decline in intent to vaccinate among adults in the UK and the USA.
Researchers are often interested in analysing conditional treatment effects. One variant of this is ‘causal moderation’, which implies that intervention upon a third (moderator) variable would alter the treatment effect. This study considers the conditions under which causal moderation can be identified and presents a generalized framework for estimating causal moderation effects given randomized treatments and non‐randomized moderators. As part of the estimation process, it allows researchers to implement their preferred method of covariate adjustment, including parametric and non‐parametric methods, or alternative identification strategies of their choosing. In addition, it provides a set‐up whereby sensitivity analysis designed for the average treatment effect context can be extended to the moderation context. To illustrate the methods, the study presents two applications: one dealing with the effect of using the term ‘welfare’ to describe public assistance in the United States, and one dealing with the effect of asylum seekers’ religion on European attitudes towards asylum seekers.
As government pressure on major technology companies builds, both firms and legislators are searching for technical solutions to difficult platform governance puzzles such as hate speech and misinformation. Automated hash-matching and predictive machine learning tools – what we define here as algorithmic moderation systems – are increasingly being deployed to conduct content moderation at scale by major platforms for user-generated content such as Facebook, YouTube and Twitter. This article provides an accessible technical primer on how algorithmic moderation works; examines some of the existing automated tools used by major platforms to handle copyright infringement, terrorism and toxic speech; and identifies key political and ethical issues for these systems as the reliance on them grows. Recent events suggest that algorithmic moderation has become necessary to manage growing public expectations for increased platform responsibility, safety and security on the global stage; however, as we demonstrate, these systems remain opaque, unaccountable and poorly understood. Despite the potential promise of algorithms or ‘AI’, we show that even ‘well optimized’ moderation systems could exacerbate, rather than relieve, many existing problems with content policy as enacted by platforms for three main reasons: automated moderation threatens to (a) further increase opacity, making a famously non-transparent set of practices even more difficult to understand or audit, (b) further complicate outstanding issues of fairness and justice in large-scale sociotechnical systems and (c) re-obscure the fundamentally political nature of speech decisions being executed at scale.
Researchers need to select high-quality research designs and communicate those designs clearly to readers. Both tasks are difficult. We provide a framework for formally “declaring” the analytically relevant features of a research design in a demonstrably complete manner, with applications to qualitative, quantitative, and mixed methods research. The approach to design declaration we describe requires defining a model of the world ( M ), an inquiry ( I ), a data strategy ( D ), and an answer strategy ( A ). Declaration of these features in code provides sufficient information for researchers and readers to use Monte Carlo techniques to diagnose properties such as power, bias, accuracy of qualitative causal inferences, and other “diagnosands.” Ex ante declarations can be used to improve designs and facilitate preregistration, analysis, and reconciliation of intended and actual analyses. Ex post declarations are useful for describing, sharing, reanalyzing, and critiquing existing designs. We provide open-source software, DeclareDesign, to implement the proposed approach.
With the rapid development of artificial intelligence have come concerns about how machines will make moral decisions, and the major challenge of quantifying societal expectations about the ethical principles that should guide machine behaviour. To address this challenge, we deployed the Moral Machine, an online experimental platform designed to explore the moral dilemmas faced by autonomous vehicles. This platform gathered 40 million decisions in ten languages from millions of people in 233 countries and territories. Here we describe the results of this experiment. First, we summarize global moral preferences. Second, we document individual variations in preferences, based on respondents’ demographics. Third, we report cross-cultural ethical variation, and uncover three major clusters of countries. Fourth, we show that these differences correlate with modern institutions and deep cultural traits. We discuss how these preferences can contribute to developing global, socially acceptable principles for machine ethics. All data used in this article are publicly available.
Codes of conduct in autonomous vehicles
When it becomes possible to program decision-making based on moral principles into machines, will self-interest or the public good predominate? In a series of surveys, Bonnefon et al. found that even though participants approve of autonomous vehicles that might sacrifice passengers to save others, respondents would prefer not to ride in such vehicles (see the Perspective by Greene). Respondents would also not approve regulations mandating self-sacrifice, and such regulations would make them less willing to buy an autonomous vehicle.
Science , this issue p. 1573 ; see also p. 1514
Survey experiments are a core tool for causal inference. Yet, the design of classical survey experiments prevents them from identifying which components of a multidimensional treatment are influential. Here, we show how conjoint analysis, an experimental design yet to be widely applied in political science, enables researchers to estimate the causal effects of multiple treatment components and assess several causal hypotheses simultaneously. In conjoint analysis, respondents score a set of alternatives, where each has randomly varied attributes. Here, we undertake a formal identification analysis to integrate conjoint analysis with the potential outcomes framework for causal inference. We propose a new causal estimand and show that it can be nonparametrically identified and easily estimated from conjoint data using a fully randomized design. We then demonstrate the value of these techniques through empirical applications to voter decision-making and attitudes toward immigrants.
Technological advances are enabling roles for machines that present novel ethical challenges. The study of 'AI ethics' has emerged to confront these challenges, and connects perspectives from philosophy, computer science, law, and economics. Less represented in these interdisciplinary efforts is the perspective of cognitive science. We propose a framework - computational ethics - that specifies how the ethical challenges of AI can be partially addressed by incorporating the study of human moral decision-making. The driver of this framework is a computational version of reflective equilibrium (RE), an approach that seeks coherence between considered judgments and governing principles. The framework has two goals: (i) to inform the engineering of ethical AI systems, and (ii) to characterize human moral judgment and decision-making in computational terms. Working jointly towards these two goals will create the opportunity to integrate diverse research questions, bring together multiple academic communities, uncover new interdisciplinary research topics, and shed light on centuries-old philosophical questions.
Since 2015, there has been a huge increase in laws that ostensibly seek to counter misinformation. Since the pandemic began, this trend has only accelerated. Both authoritarian and democratic governments have introduced more new policies to fight misinformation in 2019 and in 2020. In authoritarian states pandemic-related misinformation provided a new justification for repressive policies. Questions of political motivations aside, as the continuing problem of pandemic misinformation illustrates, it’s unclear how effective these laws are.
Experimental political science has changed. In two short decades, it evolved from an emergent method to an accepted method to a primary method. The challenge now is to ensure that experimentalists design sound studies and implement them in ways that illuminate cause and effect. Ethical boundaries must also be respected, results interpreted in a transparent manner, and data and research materials must be shared to ensure others can build on what has been learned. This book explores the application of new designs; the introduction of novel data sources, measurement approaches, and statistical methods; the use of experiments in more substantive domains; and discipline-wide discussions about the robustness, generalizability, and ethics of experiments in political science. By exploring these novel opportunities while also highlighting the concomitant challenges, this volume enables scholars and practitioners to conduct high-quality experiments that will make key contributions to knowledge.
Liars are causing devastating problems. They are endangering public health. They are threatening self-government. They are destroying the reputation of good people—and inflating the reputation of people who are not so good. Nonetheless, most falsehoods ought not to be censored or regulated, even if they are lies. In general, free societies allow them. Public officials should not be allowed to act as the truth police. A key reason is that we cannot trust officials to separate truth from falsehood; their own judgments are unreliable, and their own biases get in the way. If officials are licensed to punish falsehoods, they will end up punishing dissent. The best response to falsehoods is usually to correct them, rather than to punish or censor them. But there is an exception to this proposition: governments should have the power to regulate the most harmful lies and falsehoods. False statements are not constitutionally protected if the government can show that they threaten to cause serious harm. Public officials should be able to restrict and punish lies and falsehoods that pose serious threats to public health and safety. To protect the democratic process, public officials should be able to restrict certain lies and falsehoods. They should be able to safeguard people’s reputations. Private institutions, including television networks, magazines, and newspapers, and social media platforms such as Facebook and Twitter, should be doing much more to slow or stop the spread of lies and falsehoods.
Conjoint analysis is a common tool for studying political preferences. The method disentangles patterns in respondents’ favorability toward complex, multidimensional objects, such as candidates or policies. Most conjoints rely upon a fully randomized design to generate average marginal component effects (AMCEs). They measure the degree to which a given value of a conjoint profile feature increases, or decreases, respondents’ support for the overall profile relative to a baseline, averaging across all respondents and other features. While the AMCE has a clear causal interpretation (about the effect of features), most published conjoint analyses also use AMCEs to describe levels of favorability. This often means comparing AMCEs among respondent subgroups. We show that using conditional AMCEs to describe the degree of subgroup agreement can be misleading as regression interactions are sensitive to the reference category used in the analysis. This leads to inferences about subgroup differences in preferences that have arbitrary sign, size, and significance. We demonstrate the problem using examples drawn from published articles and provide suggestions for improved reporting and interpretation using marginal means and an omnibus F-test. Given the accelerating use of these designs in political science, we offer advice for best practice in analysis and presentation of results.
Conjoint analysis is a common tool for studying political preferences. The method disentangles patterns in respondents' favorability toward complex, multidimensional objects, such as candidates or policies. Most conjoints rely upon a fully randomized design to generate average marginal component effects (AMCEs). These measure the degree to which a given value of a conjoint profile feature increases, or decreases, respondents' support for the overall profile relative to a baseline, averaging across all respondents and other features. While the AMCE has a clear causal interpretation (about the effect of features), most published conjoint analyses also use AMCEs to describe levels of favorability. This often means comparing AM-CEs among respondent subgroups. We show that using conditional AMCEs to describe the degree of subgroup agreement can be misleading as regression interactions are sensitive to the reference category used in the analysis. This leads to inferences about subgroup differences in preferences that have arbitrary sign, size, and significance. We demonstrate the problem using examples drawn from published articles and provide suggestions for improved reporting and interpretation using marginal means and an omnibus F-test. Given the accelerating use of these designs in political science, we offer advice for best practice in analysis and presentation of results.
Where does morality come from? Why are moral judgments often so similar across cultures, yet sometimes so variable? Is morality one thing, or many? Moral Foundations Theory (MFT) was created to answer these questions. In this chapter, we describe the origins, assumptions, and current conceptualization of the theory and detail the empirical findings that MFT has made possible, both within social psychology and beyond. Looking toward the future, we embrace several critiques of the theory and specify five criteria for determining what should be considered a foundation of human morality. Finally, we suggest a variety of future directions for MFT and moral psychology.
Censorship of social media in China
Figuring out how many and which social media comments are censored by governments is difficult because those comments, by definition, cannot be read. King et al. have posted comments to social media sites in China and then waited to see which of these never appeared, which appeared and were then removed, and which appeared and survived. About 40% of their submissions were reviewed by an army of censors, and more than half of these never appeared. By varying the content of posts across topics, they conclude that any mention of collective action is selectively suppressed.
Science , this issue 10.1126/science.1251722
Issues of freedom of expression, intellectual property and data protection dominate debates concerning Internet governance, and the legal responsibility of gatekeepers is often at the centre of such discussions. A focused analysis is needed on what is meant by the term and how to identify and differentiate between the various gatekeepers. This article traces the historical development of the term gatekeeper and shows how traditional conceptions of gatekeeping are inadequate for the context of the Internet where gatekeeping primarily involves control over the flow, content and accessibility of information. A particular type of gatekeeper will be identified, termed ‘Internet Information Gatekeepers’, which are those gatekeepers that as a result of their control of the flow of information, control deliberation and participation in democratic culture. This article will then propose a human rights driven framework for identifying and differentiating between the various gatekeepers and their levels of responsibility.
The abstract for this document is available on CSA Illumina.To view the Abstract, click the Abstract button above the document title.
Where does morality come from? Why are moral judgments often so similar across cultures, yet sometimes so variable? Is morality one thing, or many? Moral Foundations Theory (MFT) was created to answer these questions. In this chapter we describe the origins, assumptions, and current conceptualization of the theory, and detail the empirical findings that MFT has made possible, both within social psychology and beyond. Looking toward the future, we embrace several critiques of the theory, and specify five criteria for determining what should be considered a foundation of human morality. Finally, we suggest a variety of future directions for MFT and for moral psychology.
The dramatic surge in regional integration schemes over the past two decades has been one of the most important developments in world politics. Virtually all countries are now members of at least one regional grouping. South Asia is no exception to this trend. In December 1985, seven South Asian countries came together to establish South Asian Association for Regional Cooperation (SAARC) to address issues of peace and development in the region. This book examines regionalism in South Asia, exploring the linkages between institutional structures, government capabilities, and domestic actors' preferences to explain the dynamics of regional cooperation. It considers the formation and evolution of SAARC, explaining why its growth in terms of institutional developments and program implementation has remained modest and slow over the past two decades. It also addresses the impact of important issues such as the acquisition of nuclear capabilities by India and Pakistan, the unending conflicts in Kashmir, the war against global terror in Afghanistan, and India's growing economy. Drawing on a wealth of empirical research, including elite interviews and trade transaction data, this book sheds new light on the main cooperation issues in South Asia today and provides important information on the trends and prospects for regional cooperation in future years.
The long-standing rationalist tradition in moral psychology emphasizes the role of reason in moral judgment. A more recent trend places increased emphasis on emotion. Although both reason and emotion are likely to play important roles in moral judgment, relatively little is known about their neural correlates, the nature of their interaction, and the factors that modulate their respective behavioral influences in the context of moral judgment. In two functional magnetic resonance imaging (fMRI) studies using moral dilemmas as probes, we apply the methods of cognitive neuroscience to the study of moral judgment. We argue that moral dilemmas vary systematically in the extent to which they engage emotional processing and that these variations in emotional engagement influence moral judgment. These results may shed light on some puzzling patterns in moral judgment observed by contemporary philosophers.
Covid lies cost lives-we have a duty to clamp down on them
- G Monbiot
G. Monbiot, Covid lies cost lives-we have a duty to clamp down on them (2021). https:
//www.theguardian.com/commentisfree/2021/jan/27/covid-lies-cost-lives-right-clamp-downmisinformation.
Standing for voice and free expression
- M Zuckerberg
M. Zuckerberg, Standing for voice and free expression (2019). https://about.fb.com/news/2019/
10/mark-zuckerberg-stands-for-voice-and-free-expression/.
Proposal for a regulation of the European Parliament and of the Council on a single market for digital services
European Commission, Proposal for a regulation of the European Parliament and of the Council
on a single market for digital services (Digital Services Act) and amending Directive 2000/31/EC
(COM/2020/825 final) (2020). https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:
52020PC0825&from=en.
Trump suspended for two years
- N Clegg
N. Clegg, In response to Oversight Board, Trump suspended for two years; will only be reinstated
if conditions permit (2021). https://about.fb.com/news/2021/06/facebook-response-to-oversightboard-recommendations-trump/.
Permanent suspension of @realDonaldTrump
Twitter, Permanent suspension of @realDonaldTrump (2021). https://blog.twitter.com/en_us/
topics/company/2020/suspension.
A lot of you have asked what we’re doing about misinformation so I wanted to give an update
- M Zuckerberg
M. Zuckerberg, A lot of you have asked what we're doing about misinformation, so I wanted to give
an update (2016). https://www.facebook.com/zuck/posts/10103269806149061.
Removing Holocaust denial content
- M Bickert
M. Bickert, Removing Holocaust denial content (2020). https://about.fb.com/news/2020/10/
removing-holocaust-denial-content/.
COVID-19 and vaccine policy updates and protections
Instagram, COVID-19 and vaccine policy updates and protections (n.d.). https://help.instagram.
com/697825587576762. Accessed 27 December 2021.
An update on our work to keep people informed and limit misinformation about COVID-19
- G Rosen
G. Rosen, An update on our work to keep people informed and limit misinformation about COVID-19 (2020). https://about.fb.com/news/2020/04/covid-19-misinfo-update/#removing-more-falseclaims. Accessed 04 January 2022.
COVID-19 misleading information policy (n.d
Twitter, COVID-19 misleading information policy (n.d.). https://help.twitter.com/en/rules-andpolicies/medical-misinformation-policy. Accessed 26 December 2021.
Community guidelines (n.d
Pinterest, Community guidelines (n.d.). https://policy.pinterest.com/en/community-guidelines.
Accessed 26 April 2022.
Menace to public health': 270 experts criticise Spotify over Joe Rogan's podcast
- M Yang
M. Yang, 'Menace to public health': 270 experts criticise Spotify over Joe Rogan's podcast (2022).
https://www.theguardian.com/technology/2022/jan/14/spotify-joe-rogan-podcast-open-letter.
Cohen Sacha Baron Cohen’s keynote address at ADL’s 2019 Never Is Now summit on anti-Semitism and hate
- S Baron
S. Baron Cohen, Sacha Baron Cohen's keynote address at ADL's 2019 Never Is Now summit on
anti-Semitism and hate (2019). https://www.adl.org/news/article/sacha-baron-cohens-keynoteaddress-at-adls-2019-never-is-now-summit-on-anti-semitism.
Censorious governments are abusing
The Economist, Censorious governments are abusing "fake news" laws (2021). https://www.
economist.com/international/2021/02/11/censorious-governments-are-abusing-fake-news-laws.
Rush to pass ‘fake news’ laws during Covid-19 intensifying global media freedom challenges
- J Wiseman
J. Wiseman, Rush to pass 'fake news' laws during Covid-19 intensifying global media freedom
challenges (2020). https://ipi.media/rush-to-pass-fake-news-laws-during-covid-19-intensifyingglobal-media-freedom-challenges.
A guide to anti-misinformation actions around the world
- D Funke
- D Flamini
D. Funke, D. Flamini, A guide to anti-misinformation actions around the world (n.d.). https:
//www.poynter.org/ifcn/anti-misinformation-actions/. Accessed 27 January 2022.
Social media's struggle with self-censorship
The Economist, Social media's struggle with self-censorship (2020). https://www.economist.com/
briefing/2020/10/22/social-medias-struggle-with-self-censorship.
- E Awad
E. Awad et al., Computational ethics. Trends Cognit. Sci. 26, 388-405 (2022).
When Do the Public Support Hate Speech Restrictions? Symmetries and Asymmetries across Partisans in Denmark and the United States
- J Rasmussen
J. Rasmussen, When Do the Public Support Hate Speech Restrictions? Symmetries and
Asymmetries across Partisans in Denmark and the United States. psyarxiv [Preprint] (2022).
https://psyarxiv.com/j4nuc (Accessed 25 June 2022).
Building rules in public: Our approach to synthetic & manipulated media
- Y Roth
- A Achuthan
Y. Roth, A. Achuthan, Building rules in public: Our approach to synthetic & manipulated media
(2020). https://blog.twitter.com/en_us/topics/company/2020/new-approach-to-synthetic-andmanipulated-media.
Walker More Americans now say government should take steps to restrict false information online than
- A Mitchell
A. Mitchell, M. Walker, More Americans now say government should take steps to restrict false information online than in 2018 (2021). https://www.pewresearch.org/fact-tank/2021/08/18/moreamericans-now-say-government-should-take-steps-to-restrict-false-information-onlinethan-in-2018/.
Global support for principle of free expression, but opposition to some forms of speech
- R Wike
- K Simmons
R. Wike, K. Simmons, Global support for principle of free expression, but opposition to some forms
of speech (2015). https://www.pewresearch.org/global/2015/11/18/global-support-for-principleof-free-expression-but-opposition-to-some-forms-of-speech/.