Fabrizio Gilardi

The increasing use of artificial intelligence (AI) in news production raises important questions about how audiences perceive and respond to AI-generated journalism. This preregistered survey experiment (N = 599, German-speaking Switzerland) examines (i) perceptions of article quality (measured as credibility, readability, and expertise) across news excerpts that were human-written, AI-assisted, or fully AI-generated, and (ii) self-reported intentions to engage following disclosure of AI involvement. Participants rated two short news excerpts before learning how they had been produced. Articles across all conditions were evaluated similarly in perceived quality. After disclosure, participants in the AI-assisted and AI-generated conditions reported a higher willingness to continue reading their assigned articles compared to the control group, but future willingness to read AI-generated news did not differ across conditions. Overall, the findings suggest that readers assess AI-generated and human-written news comparably in quality, while disclosure of AI use can momentarily increase curiosity or interest without yet changing longer-term reading intentions.

Efforts to curb online hate speech depend on our ability to reliably detect it at scale. Previous studies have highlighted the strong zero-shot classification performance of large language models (LLMs), offering a potential tool to efficiently identify harmful content. Yet for complex and ambivalent tasks like hate speech detection, pre-trained LLMs can be insufficient and carry systemic biases. Domain-specific models fine-tuned for the given task and empirical context could help address these issues, but, as we demonstrate, the quality of data used for fine-tuning decisively matters. In this study, we fine-tuned GPT-4o-mini using a unique corpus of online comments annotated by diverse groups of coders with varying annotation quality: research assistants, activists, two kinds of crowd workers, and citizen scientists. We find that only annotations from those groups of annotators that are better than zero-shot GPT-4o-mini in recognizing hate speech improve the classification performance of the fine-tuned LLM. Specifically, fine-tuning using the highest-quality annotator group – trained research assistants – boosts classification performance by increasing the model’s precision without notably sacrificing the good recall of zero-shot GPT-4o-mini. In contrast, lower-quality annotations do not improve and may even decrease the ability to identify hate speech. By examining tasks reliant on human judgment and context, we offer insights that go beyond hate speech detection.

Hate speech is widely seen as a significant obstacle to constructive online discourse, but the most effective strategies to mitigate its effects remain unclear. We claim that understanding its distribution across users is key to developing and evaluating effective content moderation strategies. We address this missing link by first examining the distribution of hate speech in five original datasets that collect user-generated posts across multiple platforms (social media and online newspapers) and countries (Switzerland and the United States). Across these diverse samples, the vast majority of hate speech is produced by a small fraction of users. Second, results from a pre-registered field experiment on Twitter indicate that counterspeech strategies obtain only small reductions of future hate speech, mainly because this approach proves ineffective against the most prolific contributors of hate. These findings suggest that complementary content moderation strategies may be necessary to effectively address the problem.

A wide range of participation is essential for democracy, as it helps prevent the dominance of extreme views, erosion of legitimacy, and political polarization. However, engagement in online political discussions often features a limited spectrum of views due to high levels of self-selection and the tendency of online platforms to facilitate exchanges primarily among like-minded individuals. This study examines whether an LLM-based bot can widen the scope of perspectives expressed by participants in online discussions through two pre-registered randomized experiments conducted in a chatroom. We evaluate the impact of a bot that actively monitors discussions, identifies missing arguments, and introduces them into the conversation. The results indicate that our bot significantly expands the range of arguments, as measured by both objective and subjective metrics. Furthermore, disclosure of the bot as AI does not significantly alter these effects. These findings suggest that LLM-based moderation tools can positively influence online political discourse.

Living in the Post API age, researchers face unprecedented challenges in obtaining social media data, while users are concerned about how big tech companies use their data. Data donation offers a promising alternative, however, its scalability is limited by low participation and high dropout rates. Research suggests that data marketplaces could be a solution, but its realization remains challenging due to theoretical gaps in treating data as an asset. This paper examines whether data marketplaces can increase individuals willingness to sell their X (Twitter) data package and the minimum price they would accept. It also explores how privacy protections and the type of data buyer may affect these decisions. Results from two preregistered online survey experiments show that a data marketplace increases participants' willingness to sell their X data by 12 to 25 percentage points compared to data donation (depending on treatments), and by 6.8 points compared to onetime purchase offers. Although difference in minimum acceptable prices are not statistically significant, over 64 percentage of participants set their price within the marketplace's suggested range (0.25 to 2), substantially lower than the amounts offered in prior onetime purchase studies. Finally, in the marketplace setting, neither the type of buyer nor the inclusion of a privacy safeguard significantly influenced participants willingness to sell.

As artificial intelligence (AI) systems become deeply embedded in everyday life, calls to align AI development with ethical and societal values have intensified. Interdisciplinary collaboration is often championed as a key pathway for fostering such engagement. Yet it remains unclear whether interdisciplinary research teams are actually leading this shift in practice. This study analyzes over 100,000 AI-related papers published on ArXiv between 2014 and 2024 to examine how ethical values and societal concerns are integrated into technical AI research. We develop a classifier to identify societal content and measure the extent to which research papers express these considerations. We find a striking shift: while interdisciplinary teams remain more likely to produce societally-oriented research, computer science-only teams now account for a growing share of the field's overall societal output. These teams are increasingly integrating societal concerns into their papers and tackling a wide range of domains - from fairness and safety to healthcare and misinformation. These findings challenge common assumptions about the drivers of societal AI and raise important questions. First, what are the implications for emerging understandings of AI safety and governance if most societally-oriented research is being undertaken by exclusively technical teams? Second, for scholars in the social sciences and humanities: in a technical field increasingly responsive to societal demands, what distinctive perspectives can we still offer to help shape the future of AI?

There is broad consensus that AI presents risks, but considerable disagreement about the nature of those risks. These differing viewpoints can be understood as distinct narratives, each offering a specific interpretation of AI’s potential dangers. One narrative focuses on doomsday predictions of AI posing long-term existential risks for humanity. Another narrative prioritizes immediate concerns that AI brings to society today, such as the reproduction of biases embedded into AI systems. A significant point of contention is that the “existential risk” narrative, which is largely speculative, may distract from the less dramatic but real and present dangers of AI. We address this “distraction hypothesis” by examining whether a focus on existential threats diverts attention from the immediate risks AI poses today. In three preregistered, online survey experiments (N = 10,800), participants were exposed to news headlines that either depicted AI as a catastrophic risk, highlighted its immediate societal impacts, or emphasized its potential benefits. Results show that i) respondents are much more concerned with the immediate, rather than existential, risks of AI, and ii) existential risk narratives increase concerns for catastrophic risks without diminishing the significant worries respondents express for immediate harms. These findings provide important empirical evidence to inform ongoing scientific and political debates on the societal implications of AI.

An unresolved aspect of women's underrepresentation in politics is the media portrayal of female candidates. This paper studies how advocacy campaigns may affect potential bias, leveraging the 2019 Swiss federal elections, which were shaped by two nation-wide, cross-party campaigns advocating for gender equality. The empirical analysis compares the 2015 and 2019 election campaigns, relying on an original dataset of the mentions that all candidates (over 3,700 respectively 4,600) received in over 2.2 million news articles. The analysis produces three main results. First, although in both elections male candidates received more media attention than female candidates did, the gender gap was significantly smaller in 2019 than in 2015. Second, in both elections, male and female candidates tended to be mentioned in conjunction with gender-stereotypical topics. Third, the gender gap in media attention before and after a key women's rights event was similar to that between the corresponding periods in 2015. These findings suggest that the differences observed between 2015 and 2019 are linked to the political campaign at large rather than to a specific event, despite its historical dimensions. The results contribute to the understanding of how advocacy campaigns can change bias in media coverage and, methodologically, to measuring and understanding gendered media coverage of politics.

Academics increasingly acknowledge the predictive power of social media for a wide variety of events and, more specifically, for financial markets. Anecdotal and empirical findings show that cryptocurrencies are among the financial assets that have been affected by news and influencers' activities on Twitter. However, the extent to which Twitter crypto influencer's posts about trading signals and their effect on market prices is mostly unexplored. In this paper, we use LLMs to uncover buy and not-buy signals from influencers and news outlets' Twitter posts and use a VAR analysis with Granger Causality tests and cross-correlation analysis to understand how these trading signals are temporally correlated with the top nine major cryptocurrencies' prices. Overall, the results show a mixed pattern across cryptocurrencies and temporal periods. However, we found that for the top three cryptocurrencies with the highest presence within news and influencer posts, their aggregated LLM-detected trading signal over the preceding 24 hours granger-causes fluctuations in their market prices, exhibiting a lag of at least 6 hours. In addition, the results reveal fundamental differences in how influencers and news outlets cover cryptocurrencies.