搜索结果：Ethnicity

Large language models (LLMs) increasingly operate in high-stakes settings including healthcare and medicine, where demographic attributes such as race and ethnicity may be explicitly stated or implicitly inferred from text. However, existing studies primarily document outcome-level disparities, offering limited insight into internal mechanisms underlying these effects. We present a mechanistic study of how race and ethnicity are represented and operationalized within LLMs. Using two publicly available datasets spanning toxicity-related generation and clinical narrative understanding tasks, we analyze three open-source models with a reproducible interpretability pipeline combining probing, neuron-level attribution, and targeted intervention. We find that demographic information is distributed across internal units with substantial cross-model variation. Although some units encode sensitive or stereotype-related associations from pretraining, identical demographic cues can induce qualitatively different behaviors. Interventions suppressing such neurons reduce bias but leave substantial residual effects, suggesting behavioral rather than representational change and motivating more sys

In many consumer virtual reality (VR) applications, users embody predefined characters that offer minimal customization options, frequently emphasizing storytelling over user choice. We explore whether matching a user's physical characteristics, specifically ethnicity and gender, with their virtual self-avatar affects their sense of embodiment in VR. We conducted a 2 x 2 within-subjects experiment (n=32) with a diverse user population to explore the impact of matching or not matching a user's self-avatar to their ethnicity and gender on their sense of embodiment. Our results indicate that matching the ethnicity of the user and their self-avatar significantly enhances sense of embodiment regardless of gender, extending across various aspects, including appearance, response, and ownership. We also found that matching gender significantly enhanced ownership, suggesting that this aspect is influenced by matching both ethnicity and gender. Interestingly, we found that matching ethnicity specifically affects self-location while matching gender specifically affects one's body ownership.

Numerous studies have shown that existing Face Recognition Systems (FRS), including commercial ones, often exhibit biases toward certain ethnicities due to under-represented data. In this work, we explore ethnicity alteration and skin tone modification using synthetic face image generation methods to increase the diversity of datasets. We conduct a detailed analysis by first constructing a balanced face image dataset representing three ethnicities: Asian, Black, and Indian. We then make use of existing Generative Adversarial Network-based (GAN) image-to-image translation and manifold learning models to alter the ethnicity from one to another. A systematic analysis is further conducted to assess the suitability of such datasets for FRS by studying the realistic skin-tone representation using Individual Typology Angle (ITA). Further, we also analyze the quality characteristics using existing Face image quality assessment (FIQA) approaches. We then provide a holistic FRS performance analysis using four different systems. Our findings pave the way for future research works in (i) developing both specific ethnicity and general (any to any) ethnicity alteration models, (ii) expanding suc

Language models have been shown to propagate social bias through their output, particularly in the representation of gender and ethnicity. This paper investigates gender and ethnicity biases in AI-generated occupational stories. Representation biases are measured before and after applying our proposed mitigation strategy, Bias Analysis and Mitigation through Explanation (BAME), revealing improvements in demographic representation ranging from 2% to 20%. BAME leverages model-generated explanations to inform targeted prompt engineering, effectively reducing biases without modifying model parameters. By analyzing stories generated across 25 occupational groups, three large language models (Claude 3.5 Sonnet, Llama 3.1 70B Instruct, and GPT-4 Turbo), and multiple demographic dimensions, we identify persistent patterns of overrepresentation and underrepresentation linked to training data stereotypes. Our findings demonstrate that guiding models with their own internal reasoning mechanisms can significantly enhance demographic parity, thereby contributing to the development of more transparent generative AI systems.

Matching avatar characteristics to a user can impact sense of embodiment (SoE) in VR. However, few studies have examined how participant demographics may interact with these matching effects. We recruited a diverse and racially balanced sample of 78 participants to investigate the differences among participant groups when embodying both demographically matched and unmatched avatars. We found that participant ethnicity emerged as a significant factor, with Asian and Black participants reporting lower total SoE compared to Hispanic participants. Furthermore, we found that user ethnicity significantly influences ownership (a subscale of SoE), with Asian and Black participants exhibiting stronger effects of matched avatar ethnicity compared to White participants. Additionally, Hispanic participants showed no significant differences, suggesting complex dynamics in ethnic-racial identity. Our results also reveal significant main effects of matched avatar ethnicity and gender on SoE, indicating the importance of considering these factors in VR experiences. These findings contribute valuable insights into understanding the complex dynamics shaping VR experiences across different demographi

AI fairness measurements, including tests for equal treatment, often take the form of disaggregated evaluations of AI systems. Such measurements are an important part of Responsible AI operations. These measurements compare system performance across demographic groups or sub-populations and typically require member-level demographic signals such as gender, race, ethnicity, and location. However, sensitive member-level demographic attributes like race and ethnicity can be challenging to obtain and use due to platform choices, legal constraints, and cultural norms. In this paper, we focus on the task of enabling AI fairness measurements on race/ethnicity for \emph{U.S. LinkedIn members} in a privacy-preserving manner. We present the Privacy-Preserving Probabilistic Race/Ethnicity Estimation (PPRE) method for performing this task. PPRE combines the Bayesian Improved Surname Geocoding (BISG) model, a sparse LinkedIn survey sample of self-reported demographics, and privacy-enhancing technologies like secure two-party computation and differential privacy to enable meaningful fairness measurements while preserving member privacy. We provide details of the PPRE method and its privacy guara

Globalization and the world wide web has resulted in academia and science being an international and multicultural community forged by researchers and scientists with different ethnicities. How ethnicity shapes the evolution of membership, status and interactions of the scientific community, however, is not well understood. This is due to the difficulty of ethnicity identification at the large scale. We use name ethnicity classification as an indicator of ethnicity. Based on automatic name ethnicity classification of 1.7+ million authors gathered from Web, the name ethnicity of computer science scholars is investigated by population size, publication contribution and collaboration strength. By showing the evolution of name ethnicity from 1936 to 2010, we discover that ethnicity diversity has increased significantly over time and that different research communities in certain publication venues have different ethnicity compositions. We notice a clear rise in the number of Asian name ethnicities in papers. Their fraction of publication contribution increases from approximately 10% to near 50% from 1970 to 2010. We also find that name ethnicity acts as a homophily factor on coauthor n

We use place of birth information from the Social Security Administration linked to earnings data from the Longitudinal Employer-Household Dynamics Program and detailed race and ethnicity data from the 2010 Census to study how long-term earnings differentials vary by place of birth for different self-identified race and ethnicity categories. We focus on foreign-born persons from countries that are heavily Hispanic and from countries in the Middle East and North Africa (MENA). We find substantial heterogeneity of long-term earnings differentials within country of birth, some of which will be difficult to detect when the reporting format changes from the current two-question version to the new single-question version because they depend on self-identifications that place the individual in two distinct categories within the single-question format, specifically, Hispanic and White or Black, and MENA and White or Black. We also study the USA-born children of these same immigrants. Long-term earnings differences for the 2nd generation also vary as a function of self-identified ethnicity and race in ways that changing to the single-question format could affect.

During the COVID-19 pandemic, Aotearoa followed an elimination strategy followed by a mitigation strategy, which saw high success and kept health impact low. However, there were inequities in health outcomes, notably that Māori and Pacific Peoples had lower vaccine coverage and experienced higher age-standardised rates of hospitalisation and death. Models provide predictions of disease spread and burden, which can effectively inform policy, but are often less good at including inequities/heterogeneity. Despite the differences in health outcomes, most models have not explicitly considered ethnic heterogeneities as factors. We developed such a model to investigate the first Omicron wave of the COVID-19 pandemic in Aotearoa, which was the first widespread community transmission of SARS-CoV-2. We analysed three models for contact patterns within and between ethnicities: proportionate, assortative, and unconstrained mixing, which were fit using ethnicity-specific data on reported cases and spatially disaggregated population counts. We found that Māori, Pacific, and Asian transmission rates were between 1.08-2.46, 1.50-3.89, and 0.80-0.92 times the European rates, respectively. We then f

There is active debate over whether to consider patient race and ethnicity when estimating disease risk. By accounting for race and ethnicity, it is possible to improve the accuracy of risk predictions, but there is concern that their use may encourage a racialized view of medicine. In diabetes risk models, despite substantial gains in statistical accuracy from using race and ethnicity, the gains in clinical utility are surprisingly modest. These modest clinical gains stem from two empirical patterns: first, the vast majority of individuals receive the same screening recommendation regardless of whether race or ethnicity are included in risk models; and second, for those who do receive different screening recommendations, the difference in utility between screening and not screening is relatively small. Our results are based on broad statistical principles, and so are likely to generalize to many other risk-based clinical decisions.

Precision medicine aims to create biomedical solutions tailored to specific factors that affect disease risk and treatment responses within the population. The success of the genomics era and recent widespread availability of electronic health records (EHR) has ushered in a new wave of genomic biobanks connected to EHR databases (EHR-linked biobanks). This perspective aims to discuss how race, ethnicity, and genetic ancestry are currently utilized to study common disease variation through genetic association studies. Although genetic ancestry plays a significant role in shaping the genetic landscape underlying disease risk in humans, the overall risk of a disease is caused by a complex combination of environmental, sociocultural, and genetic factors. When using EHR-linked biobanks to interrogate underlying disease etiology, it is also important to be aware of how the biases associated with commonly used descent-associated concepts such as race and ethnicity can propagate to downstream analyses. We intend for this resource to support researchers who perform or analyze genetic association studies in the EHR-linked biobank setting such as those involved in consortium-wide biobanking e

This study builds on person perception and human AI interaction (HAII) theories to investigate how content and source cues, specifically race, ethnicity, and nationality, affect judgments of AI-generated content in a high-stakes self-presentation context: college applications. Results of a pre-registered experiment with a nationally representative U.S. sample (N = 644) show that content heuristics, such as linguistic style, played a dominant role in AI detection. Source heuristics, such as nationality, also emerged as a significant factor, with international students more likely to be perceived as using AI, especially when their statements included AI-sounding features. Interestingly, Asian and Hispanic applicants were more likely to be judged as AI users when labeled as domestic students, suggesting interactions between racial stereotypes and AI detection. AI attribution led to lower perceptions of personal statement quality and authenticity, as well as negative evaluations of the applicant's competence, sociability, morality, and future success.

We examine whether large language models (LLMs) exhibit race- and gender-based name discrimination in hiring decisions, similar to classic findings in the social sciences (Bertrand and Mullainathan, 2004). We design a series of templatic prompts to LLMs to write an email to a named job applicant informing them of a hiring decision. By manipulating the applicant's first name, we measure the effect of perceived race, ethnicity, and gender on the probability that the LLM generates an acceptance or rejection email. We find that the hiring decisions of LLMs in many settings are more likely to favor White applicants over Hispanic applicants. In aggregate, the groups with the highest and lowest acceptance rates respectively are masculine White names and masculine Hispanic names. However, the comparative acceptance rates by group vary under different templatic settings, suggesting that LLMs' race- and gender-sensitivity may be idiosyncratic and prompt-sensitive.

Large Language Models (LLMs) offer a lucrative promise for scalable content moderation, including hate speech detection. However, they are also known to be brittle and biased against marginalised communities and dialects. This requires their applications to high-stakes tasks like hate speech detection to be critically scrutinized. In this work, we investigate the robustness of hate speech classification using LLMs particularly when explicit and implicit markers of the speaker's ethnicity are injected into the input. For explicit markers, we inject a phrase that mentions the speaker's linguistic identity. For the implicit markers, we inject dialectal features. By analysing how frequently model outputs flip in the presence of these markers, we reveal varying degrees of brittleness across 3 LLMs and 1 LM and 5 linguistic identities. We find that the presence of implicit dialect markers in inputs causes model outputs to flip more than the presence of explicit markers. Further, the percentage of flips varies across ethnicities. Finally, we find that larger models are more robust. Our findings indicate the need for exercising caution in deploying LLMs for high-stakes tasks like hate spee

A Music Recommendation System based on Emotion, Age, and Ethnicity is developed in this study, using FER-2013 and ``Age, Gender, and Ethnicity (Face Data) CSV'' datasets. The CNN architecture, which is extensively used for this kind of purpose has been applied to the training of the models. After adding several appropriate layers to the training end of the project, in total, 3 separate models are trained in the Deep Learning side of the project: Emotion, Ethnicity, and Age. After the training step of these models, they are used as classifiers on the web application side. The snapshot of the user taken through the interface is sent to the models to predict their mood, age, and ethnic origin. According to these classifiers, various kinds of playlists pulled from Spotify API are proposed to the user in order to establish a functional and user-friendly atmosphere for the music selection. Afterward, the user can choose the playlist they want and listen to it by following the given link.

Anonymized electronic medical records are an increasingly popular source of research data. However, these datasets often lack race and ethnicity information. This creates problems for researchers modeling human disease, as race and ethnicity are powerful confounders for many health exposures and treatment outcomes; race and ethnicity are closely linked to population-specific genetic variation. We showed that deep neural networks generate more accurate estimates for missing racial and ethnic information than competing methods (e.g., logistic regression, random forest). RIDDLE yielded significantly better classification performance across all metrics that were considered: accuracy, cross-entropy loss (error), and area under the curve for receiver operating characteristic plots (all $p < 10^{-6}$). We made specific efforts to interpret the trained neural network models to identify, quantify, and visualize medical features which are predictive of race and ethnicity. We used these characterizations of informative features to perform a systematic comparison of differential disease patterns by race and ethnicity. The fact that clinical histories are informative for imputing race and et