arXiv subjects

Joana Gonçalves-Sá

Publications and source records attributed to Joana Gonçalves-Sá.

8 recordsLinked to original sources

arXiv2026

Evidence of political bias in search engines and language models before major elections

Search engines (SEs) and large language models (LLMs) are central to political information access, yet their algorithmic decisions and potential underlying biases remain underexplored. We developed a standardized, privacy-preserving, bot-and-proxy methodology to audit four SEs and two LLMs before the 2024 European Parliament and US presidential elections. We collected answers to approximately 4,360 queries related to elections in five EU countries and 15 US counties, identified political entities and topics in those answers, and mapped them to ideological positions (EU) or issue associations (US). In Europe, SE results disproportionately mentioned far-right entities beyond levels expected from polls, past elections, or media salience. In the US, Google strongly favored topics more important to Republican voters, while other search engines favored issues more relevant to Democrats. LLMs responses were more balanced, although there is evidence of overrepresentation of far-right (and Green) entities. These results show evidence of bias and open important discussions on how even small skews in widely used platforms may influence democratic processes, calling for systematic audits of their outputs.

cs.CY↗

arXiv2026

Cross-National Evidence of Disproportionate Media Visibility for the Radical Right in the 2024 European Elections

This study provides a systematic comparative analysis of media visibility of different political families during the 2024 European Parliament elections. We analyzed close to 21,500 unique news from leading national outlets in Austria, Germany, Ireland, Poland, and Portugal - countries with diverse political contexts and levels of media trust. Combining computational and human classification, we identified parties, political leaders, and groups from the article's URLs and titles, and clustered them according to European Parliament political families and broad political leanings. Cross-country comparison shows that the Mainstream and the Radical Right were mentioned more often than the other political groups. Moreover, the Radical Right received disproportionate attention relative to electoral results (from 2019 or 2024) and electoral projections, particularly in Austria, Germany, and Ireland. This imbalance increased in the final weeks of the campaign, when media influence on undecided voters is greatest. Outlet-level analysis shows that coverage of right-leaning entities dominated across news sources, especially those generating the highest traffic, suggesting a structural rather than outlet-specific pattern. Media visibility is a central resource, and this systematic mapping of online coverage highlights how traditional media can contribute to structural asymmetries in democratic competition.

cs.CY↗

arXiv2025

Digital Gatekeeping: An Audit of Search Engine Results shows tailoring of queries on the Israel-Palestine Conflict

Search engines, often viewed as reliable gateways to information, tailor search results using customization algorithms based on user preferences, location, and more. While this can be useful for routine queries, it raises concerns when the topics are sensitive or contentious, possibly limiting exposure to diverse viewpoints and increasing polarization. To examine the extent of this tailoring, we focused on the Israel-Palestine conflict and developed a privacy-protecting tool to audit the behavior of three search engines: DuckDuckGo, Google and Yahoo. Our study focused on two main questions: (1) How do search results for the same query about the conflict vary among different users? and (2) Are these results influenced by the user's location and browsing history? Our findings revealed significant customization based on location and browsing preferences, unlike previous studies that found only mild personalization for general topics. Moreover, queries related to the conflict were more customized than unrelated queries, and the results were not neutral concerning the conflict's portrayal.

cs.CY↗

arXiv2023

Digital Epidemiology after COVID-19: impact and prospects

Epidemiology and Public Health have increasingly relied on structured and unstructured data, collected inside and outside of typical health systems, to study, identify, and mitigate diseases at the population level. Focusing on infectious disease, we review how Digital Epidemiology (DE) was at the beginning of 2020 and how it was changed by the COVID-19 pandemic, in both nature and breadth. We argue that DE will become a progressively useful tool as long as its potential is recognized and its risks are minimized. Therefore, we expand on the current views and present a new definition of DE that, by highlighting the statistical nature of the datasets, helps in identifying possible biases. We offer some recommendations to reduce inequity and threats to privacy and argue in favour of complex multidisciplinary approaches to tackling infectious diseases.

cs.CY↗

arXiv2023

Some scientific knowledge is a dangerous thing: overconfidence grows non-linearly with knowledge

Overconfidence is a prevalent problem and particularly consequential in its relation with scientific knowledge: being unaware of one`s own ignorance can affect behaviours and threaten public policies and health. We introduce both analytical and methodological changes to the study of confidence in science knowledge and in attitudes towards science and apply them to four large surveys, spanning 30 years in Europe and the USA. We propose a new indirect confidence metric that does not rely on self-reporting or peer comparison, and study how knowledge and confidence vary across their full scale. We find that confidence grows much faster than knowledge, giving rise to a non-linear relationship, with the largest confidence gaps appearing at intermediate knowledge levels. These high-confidence \textbackslash intermediate-knowledge groups also display the least positive attitudes towards science, with important consequences for science communication. These results are contrary to current models, including the predictions of the Dunning-Kruger effect, and we discuss how our model, if correct, can guide future research and communication policies.

physics.ed-ph↗

arXiv2021

Learning from pandemics: using extraordinary events can improve disease now-casting models

Online searches have been used to study different health-related behaviours, including monitoring disease outbreaks. An obvious caveat is that several reasons can motivate individuals to seek online information and models that are blind to people's motivations are of limited use and can even mislead. This is particularly true during extraordinary public health crisis, such as the ongoing pandemic, when fear, curiosity and many other reasons can lead individuals to search for health-related information, masking the disease-driven searches. However, health crisis can also offer an opportunity to disentangle between different drivers and learn about human behavior. Here, we focus on the two pandemics of the 21st century (2009-H1N1 flu and Covid-19) and propose a methodology to discriminate between search patterns linked to general information seeking (media driven) and search patterns possibly more associated with actual infection (disease driven). We show that by learning from such pandemic periods, with high anxiety and media hype, it is possible to select online searches and improve model performance both in pandemic and seasonal settings. Moreover, and despite the common claim that more data is always better, our results indicate that lower volume of the right data can be better than including large volumes of apparently similar data, especially in the long run. Our work provides a general framework that can be applied beyond specific events and diseases, and argues that algorithms can be improved simply by using less (better) data. This has important consequences, for example, to solve the accuracy-explainability trade-off in machine-learning.

cs.LG↗

arXiv2020

PTPARL-D: Annotated Corpus of 44 years of Portuguese Parliament debates

In a representative democracy, some decide in the name of the rest, and these elected officials are commonly gathered in public assemblies, such as parliaments, where they discuss policies, legislate, and vote on fundamental initiatives. A core aspect of such democratic processes are the plenary debates, where important public discussions take place. Many parliaments around the world are increasingly keeping the transcripts of such debates, and other parliamentary data, in digital formats accessible to the public, increasing transparency and accountability. Furthermore, some parliaments are bringing old paper transcripts to semi-structured digital formats. However, these records are often only provided as raw text or even as images, with little to no annotation, and inconsistent formats, making them difficult to analyze and study, reducing both transparency and public reach. Here, we present PTPARL-D, an annotated corpus of debates in the Portuguese Parliament, from 1976 to 2019, covering the entire period of Portuguese democracy.

cs.CL↗

arXiv2017

Human Sexual Cycles are Driven by Culture and Match Collective Moods

It is a long-standing question whether human sexual and reproductive cycles are affected predominantly by biology or culture. The literature is mixed with respect to whether biological or cultural factors best explain the reproduction cycle phenomenon, with biological explanations dominating the argument. The biological hypothesis proposes that human reproductive cycles are an adaptation to the seasonal cycles caused by hemisphere positioning, while the cultural hypothesis proposes that conception dates vary mostly due to cultural factors, such as vacation schedule or religious holidays. However, for many countries, common records used to investigate these hypotheses are incomplete or unavailable, biasing existing analysis towards primarily Christian countries in the Northern Hemisphere. Here we show that interest in sex peaks sharply online during major cultural and religious celebrations, regardless of hemisphere location. This online interest, when shifted by nine months, corresponds to documented human birth cycles, even after adjusting for numerous factors such as language, season, and amount of free time due to holidays. We further show that mood, measured independently on Twitter, contains distinct collective emotions associated with those cultural celebrations, and these collective moods correlate with sex search volume outside of these holidays as well. Our results provide converging evidence that the cyclic sexual and reproductive behavior of human populations is mostly driven by culture and that this interest in sex is associated with specific emotions, characteristic of, but not limited to, major cultural and religious celebrations.

cs.SI↗