arXiv subjects

Samuel P. Fraiberger

Publications and source records attributed to Samuel P. Fraiberger.

11 recordsLinked to original sources

arXiv2026

LLMs Mirror Country-Specific Gender Patterns If Asked, but Skew Male When Generating Media in Local Languages

Large language models (LLMs) are increasingly used to generate media, but whether their content perpetuates gender stereotypes is unknown: standard benchmarks rely on selection-based formats rather than long-form generation, and surveyed baselines for local gender associations are scarce outside the West. We collect gender associations for 22 occupational and domestic roles from 695 respondents across the United States, India, Kenya, and Nigeria, and evaluate eight LLMs under two regimes: direct questioning and media generation. Models track the surveyed associations under direct questioning but skew substantially more male under media generation in major local-language cells, consistent with the male bias documented in human-produced media. Outside the US, the shift is much smaller and non-significant under English prompting, so English-only or country-agnostic evaluation would miss this bias in the languages where these models are most deployed. Instruction prompting reduces the shift directionally, but trades off against alignment with the surveyed associations. Evaluating LLM gender bias for global deployment therefore requires generation-format testing, local-language prompting, and locally-collected human baselines.

cs.CL↗

arXiv2026

The Enforcement and Feasibility of Hate Speech Moderation

Online hate speech is associated with harms ranging from deteriorating mental health to violence, yet how consistently platforms moderate hate, and whether enforcement is feasible at scale, remain poorly understood. We audit hate speech moderation on Twitter (now X) using 540,000 tweets annotated by trained native speakers, representative of a full day on the platform. Five months after posting, 80% of hateful tweets, including violent ones, remained online. Removal was only marginally more likely than for non-hateful tweets, far below scams or adult content, and insensitive to severity and reach. Automated detection could not reliably classify hate but ranked it highly, enabling human triage. Simulating this workflow, current staffing curbed little exposure, yet substantial reductions proved financially feasible, far below applicable regulatory fines. Persistent hate reflects resource allocation, not technical limits.

cs.CY↗

arXiv2026

Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias

Demographic cue-based evaluation is widely used to study how large language models (LLMs) adapt their responses to signaled demographic attributes within and across groups. This approach typically relies on a single cue (e.g., names) as a proxy for group membership, implicitly treating different cues as interchangeable operationalizations of a single underlying identity-conditioned behavior. We test this assumption in realistic advice-seeking interactions spanning 14.8 million prompts, focusing on race and gender in a U.S. context. We find that cues for the same group induce only partially overlapping changes in model responses, yielding inconsistent conclusions about personalization, while bias conclusions are unstable, with both magnitude and direction of group differences varying across cues. We further show that these inconsistencies reflect differences in cue-group association strength and linguistic features bundled within cues that shape model responses. Together, our findings suggest that demographic conditioning in LLMs is not a cue-invariant category-level parameter but depends fundamentally on how identity is cued, reflecting responses to linguistic signals rather than stable demographic categories. We therefore call for multi-cue, mechanism-aware evaluations as a foundation for robust and interpretable claims about demographic variation in LLM responses.

cs.CL↗

arXiv2025

HoWDe: a validated algorithm for Home and Work location Detection

Smartphone location data have become a key resource for understanding urban mobility, yet extracting actionable insights requires robust and reproducible preprocessing pipelines. A central step is the identification of individuals' home and work locations, which underpins analyses of commuting, employment, accessibility, and socioeconomic patterns. However, existing approaches are often ad hoc, data-specific, and difficult to reproduce, limiting comparability across studies and datasets. We introduce HoWDe, an open-source software library for detecting home and work locations from large-scale mobility data. HoWDe implements a transparent, modular pipeline explicitly designed to handle missing data, heterogeneous sampling rates, and differences in data sparsity across individuals. The code allows users to tune a small set of interpretable parameters, enabling to adapt the algorithm to diverse applications and datasets. Using two unique ground truth datasets comprising 5,099 individuals across 68 countries, we show that HoWDe achieves home and work detection accuracies of up to 97% and 88%, respectively, with consistent performance across demographic groups and geographic contexts. We further demonstrate how parameter settings propagate to downstream metrics such as employment estimates and commuting flows, highlighting the importance of transparent methodological choices. By providing a validated, documented, and easily deployable pipeline, HoWDe supports scalable in-house preprocessing and facilitates the sharing of privacy-preserving mobility datasets. Our software and evaluation benchmarks establish methodological standards that enhance the robustness and reproducibility of human mobility research at urban and national scales.

cs.SI↗

arXiv2025

HateDay: Insights from a Global Hate Speech Dataset Representative of a Day on Twitter

To address the global challenge of online hate speech, prior research has developed detection models to flag such content on social media. However, due to systematic biases in evaluation datasets, the real-world effectiveness of these models remains unclear, particularly across geographies. We introduce HateDay, the first global hate speech dataset representative of social media settings, constructed from a random sample of all tweets posted on September 21, 2022 and covering eight languages and four English-speaking countries. Using HateDay, we uncover substantial variation in the prevalence and composition of hate speech across languages and regions. We show that evaluations on academic datasets greatly overestimate real-world detection performance, which we find is very low, especially for non-European languages. Our analysis identifies key drivers of this gap, including models' difficulty to distinguish hate from offensive speech and a mismatch between the target groups emphasized in academic datasets and those most frequently targeted in real-world settings. We argue that poor model performance makes public models ill-suited for automatic hate speech moderation and find that high moderation rates are only achievable with substantial human oversight. Our results underscore the need to evaluate detection systems on data that reflects the complexity and diversity of real-world social media.

cs.CL↗

arXiv2024

Socioeconomic disparities in mobility behavior during the COVID-19 pandemic in developing countries

Mobile phone data have played a key role in quantifying human mobility during the COVID-19 pandemic. Existing studies on mobility patterns have primarily focused on regional aggregates in high-income countries, obfuscating the accentuated impact of the pandemic on the most vulnerable populations. Leveraging geolocation data from mobile-phone users and population census for 6 middle-income countries across 3 continents between March and December 2020, we uncovered common disparities in the behavioral response to the pandemic across socioeconomic groups. Users living in low-wealth neighborhoods were less likely to respond by self-isolating, relocating to rural areas, or refraining from commuting to work. The gap in the behavioral responses between socioeconomic groups persisted during the entire observation period. Among users living in low-wealth neighborhoods, those who commute to work in high-wealth neighborhoods pre-pandemic were particularly at risk of experiencing economic stress, facing both the reduction in economic activity in the high-wealth neighborhood and being more likely to be affected by public transport closures due to their longer commute distances. While confinement policies were predominantly country-wide, these results suggest that, when data to identify vulnerable individuals are not readily available, GPS-based analytics could help design targeted place-based policies to aid the most vulnerable.

physics.soc-ph↗

arXiv2024

NaijaHate: Evaluating Hate Speech Detection on Nigerian Twitter Using Representative Data

To address the global issue of online hate, hate speech detection (HSD) systems are typically developed on datasets from the United States, thereby failing to generalize to English dialects from the Majority World. Furthermore, HSD models are often evaluated on non-representative samples, raising concerns about overestimating model performance in real-world settings. In this work, we introduce NaijaHate, the first dataset annotated for HSD which contains a representative sample of Nigerian tweets. We demonstrate that HSD evaluated on biased datasets traditionally used in the literature consistently overestimates real-world performance by at least two-fold. We then propose NaijaXLM-T, a pretrained model tailored to the Nigerian Twitter context, and establish the key role played by domain-adaptive pretraining and finetuning in maximizing HSD performance. Finally, owing to the modest performance of HSD systems in real-world conditions, we find that content moderators would need to review about ten thousand Nigerian tweets flagged as hateful daily to moderate 60% of all hateful content, highlighting the challenges of moderating hate speech at scale as social media usage continues to grow globally. Taken together, these results pave the way towards robust HSD systems and a better protection of social media users from hateful content in low-resource settings.

cs.CL↗

arXiv2021

Fine-grained prediction of food insecurity using news streams

Anticipating the outbreak of a food crisis is crucial to efficiently allocate emergency relief and reduce human suffering. However, existing food insecurity early warning systems rely on risk measures that are often delayed, outdated, or incomplete. Here, we leverage recent advances in deep learning to extract high-frequency precursors to food crises from the text of a large corpus of news articles about fragile states published between 1980 and 2020. Our text features are causally grounded, interpretable, validated by existing data, and allow us to predict 32% more food crises than existing models up to three months ahead of time at the district level across 15 fragile states. These results could have profound implications on how humanitarian aid gets allocated and open new avenues for machine learning to improve decision making in data-scarce environments.

cs.CL↗

arXiv2020

Uncovering socioeconomic gaps in mobility reduction during the COVID-19 pandemic using location data

Using smartphone location data from Colombia, Mexico, and Indonesia, we investigate how non-pharmaceutical policy interventions intended to mitigate the spread of the COVID-19 pandemic impact human mobility. In all three countries, we find that following the implementation of mobility restriction measures, human movement decreased substantially. Importantly, we also uncover large and persistent differences in mobility reduction between wealth groups: on average, users in the top decile of wealth reduced their mobility up to twice as much as users in the bottom decile. For decision-makers seeking to efficiently allocate resources to response efforts, these findings highlight that smartphone location data can be leveraged to tailor policies to the needs of specific socioeconomic groups, especially the most vulnerable.

physics.soc-ph↗

arXiv2020

Mobile phone data and COVID-19: Missing an opportunity?

This paper describes how mobile phone data can guide government and public health authorities in determining the best course of action to control the COVID-19 pandemic and in assessing the effectiveness of control measures such as physical distancing. It identifies key gaps and reasons why this kind of data is only scarcely used, although their value in similar epidemics has proven in a number of use cases. It presents ways to overcome these gaps and key recommendations for urgent action, most notably the establishment of mixed expert groups on national and regional level, and the inclusion and support of governments and public authorities early on. It is authored by a group of experienced data scientists, epidemiologists, demographers and representatives of mobile network operators who jointly put their work at the service of the global effort to combat the COVID-19 pandemic.

cs.CY↗

arXiv2016

Enhancing Transparency and Control when Drawing Data-Driven Inferences about Individuals

Recent studies have shown that information disclosed on social network sites (such as Facebook) can be used to predict personal characteristics with surprisingly high accuracy. In this paper we examine a method to give online users transparency into why certain inferences are made about them by statistical models, and control to inhibit those inferences by hiding ("cloaking") certain personal information from inference. We use this method to examine whether such transparency and control would be a reasonable goal by assessing how difficult it would be for users to actually inhibit inferences. Applying the method to data from a large collection of real users on Facebook, we show that a user must cloak only a small portion of her Facebook Likes in order to inhibit inferences about their personal characteristics. However, we also show that in response a firm could change its modeling of users to make cloaking more difficult.

stat.ML↗