The Boston Naming Test (BNT) is a confrontation naming test which is used to measure naming ability. The primary purpose of this study was to identify whether cultural bias negatively affects South African’s performance on the Boston Naming Test (BNT). More specifically the study aimed to identify the exact items of the BNT on which South Africans perform poorly because of cultural bias. The research identified alternate responses given by respondents in terms of a percentage. The study further aimed to explore whether there was a significant difference in performance when comparing English, Zulu and Sotho respondents in terms of item response. This investigation also intended to discover whether being bilingual would affect South African’s performance on the BNT. Finally, the study aimed to explore whether there was a significant difference in the performance on the BNT when comparing male and female respondents. A significant difference was found between the South African and the Canadian sample in terms of item response. 40 items were revealed as problematic in a South African sample. Significant differences were found when comparing English respondents to Zulu respondents as well as when comparing Sotho respondents to English respondents. Although differences were found between male and female performance, the difference was not significant. Ultimately, no significant difference was found between monolingual and bilingual respondents.

A Thesis submitted to The University of Arizona College of Medicine - Phoenix in partial fulfillment of the requirements for the Degree of Doctor of Medicine. / Purpose: compare the Southwestern Indigenous Cognitive Assessment (SWICA), a novel tool for screening AI older adults in Arizona, with The Montreal Cognitive Assessment (MoCA), a commonly used memory screening tool, for comparison of cultural bias. Methods: Cultural bias was assessed by retrospectively comparing coded participant responses to 16 questions about their cultural context. Intrasample variation on MoCA and SWICA tests was controlled by using the participants as their own controls. Data were analyzed using a multiple regression general linear model on SPSS software. Results: Scores on the SWICA test were independently associated with English use in the home (Beta = .396, p = .026), years of education (Beta = 335, p = .027), and ease of learning (Beta = .361, p = .029), but not age (Beta = .366, p = .054). Scores on the MoCA test were independently associated with age (Beta = ‐.491, p = .001), English use in the home (Beta = ‐.320, p = .039) , and years of education (Beta = ‐.284. p = .030), but not ease of learning (Beta = ‐.267, p = .067). Conclusions: Scores were similar on both tests (t=3.934, p=.001), and were independently associated with English use in the home and years of education. SWICA was uniquely associated with ease of learning and MoCA was uniquely associated with age. This preliminary comparison demonstrates the usefulness of SWICA, and validation of this tool is recommended.

Culture has always had a substantial influence on how art is perceived and executed. Artists have, more often than not, let their own backgrounds and experiences influence the way their art is produced; those who merely view art form opinions about works through their own cultural understanding. What makes art and what their own backgrounds allow them to distinguish art as is often defined by cultural origins. My observation is that, in this new age, there are distinct cultural biases, particularly within the U.S., that create social pressures to produce certain types of art, and anyone who operates outside that realm is disadvantaged. I have created imagery to highlight the distress that cultural biases have caused in my own life—as an artist who follows a style outside my culture—and in the lives of other artists who share my struggles, in an allegorical and comical sense.

abstract: Exabytes of data are created online every day. This deluge of data is no more apparent than it is on social media. Naturally, finding ways to leverage this unprecedented source of human information is an active area of research. Social media platforms have become laboratories for conducting experiments about people at scales thought unimaginable only a few years ago. Researchers and practitioners use social media to extract actionable patterns such as where aid should be distributed in a crisis. However, the validity of these patterns relies on having a representative dataset. As this dissertation shows, the data collected from social media is seldom representative of the activity of the site itself, and less so of human activity. This means that the results of many studies are limited by the quality of data they collect. The finding that social media data is biased inspires the main challenge addressed by this thesis. I introduce three sets of methodologies to correct for bias. First, I design methods to deal with data collection bias. I offer a methodology which can find bias within a social media dataset. This methodology works by comparing the collected data with other sources to find bias in a stream. The dissertation also outlines a data collection strategy which minimizes the amount of bias that will appear in a given dataset. It introduces a crawling strategy which mitigates the amount of bias in the resulting dataset. Second, I introduce a methodology to identify bots and shills within a social media dataset. This directly addresses the concern that the users of a social media site are not representative. Applying these methodologies allows the population under study on a social media site to better match that of the real world. Finally, the dissertation discusses perceptual biases, explains how they affect analysis, and introduces computational approaches to mitigate them. The results of the dissertation allow for the discovery and removal of different levels of bias within a social media dataset. This has important implications for social media mining, namely that the behavioral patterns and insights extracted from social media will be more representative of the populations under study. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2017

No description available.

No description available.

Generative models have rapidly evolved to produce coherent text, realistic images, and functional code. Yet these remarkable capabilities also expose critical vulnerabilities -- ranging from subtle adversarial attacks to harmful stereotypes -- that pose both technical and societal challenges. This research investigates these challenges across three modalities (code, text, and vision) before focusing on strategies to mitigate biases specifically in generative language models. First, we reveal how programming language (PL) models rely on a `natural channel' of code, such as human-readable tokens and structure, that adversaries can exploit with minimal perturbations. These attacks expose the fragility of state-of-the-art PL models, highlighting how superficial patterns and hidden assumptions in training data can lead to unanticipated vulnerabilities. Extending this analysis to textual and visual domains, we show how over-reliance on patterns seen in training data manifests as ingrained biases and harmful stereotypes. To enable more inclusive and globally representative model evaluations, we introduce SeeGULL, a large-scale benchmark of thousands of stereotypes spanning diverse cultures and identity groups worldwide. We also develop ViSAGe, a benchmark for identifying visual stereotypes at scale in text-to-image (T2I) models, illustrating the persistence of stereotypes in generated images even when prompted otherwise. Building on these findings, we propose two complementary approaches to mitigate stereotypical outputs in language models. The first is an explicit method that uses fairness constraints for model pruning, ensuring essential bias-mitigating features remain intact. The second is an implicit bias mitigation framework that makes a crucial distinction between comprehension failures and inherently learned stereotypes. This approach uses instruction tuning on general-purpose datasets and mitigates stereotypes implicitly without relying on targeted debiasing techniques. Extensive evaluations on state-of-the-art models demonstrate that our methods substantially reduce harmful stereotypes across multiple identity dimensions, while preserving downstream performance. / Doctor of Philosophy / AI systems, especially generative models that create text, images, and code, have advanced rapidly. They can write essays, generate realistic pictures, and assist with programming. However, these impressive capabilities also come with vulnerabilities that pose both technical and societal challenges. Some of these models can be subtly manipulated into making errors, while others unknowingly reinforce harmful stereotypes present in their training data. This research examines these challenges across three types of generative models: those that generate code, text, and images. First, we investigate how generative models that generate code rely on human-readable patterns that attackers can subtly manipulate, revealing hidden weaknesses in even the most advanced models. Extending this analysis to text and image generation, we show how these models often over-rely on patterns from their training data, leading to harmful stereotypes. To systematically study these issues, we introduce two large-scale benchmarks: SeeGULL, a dataset that identifies stereotypes across cultures and identity groups in AI-generated text, and ViSAGe, a dataset that uncovers hidden biases in AI-generated images. Building on these insights, we propose two complementary solutions to reduce biases in generative language models. The first method explicitly removes biased patterns from compressed AI models by introducing filtering techniques that ensure fairness while keeping the model's accuracy intact. The second takes an implicit approach by improving how generative models interpret instructions, making them less likely to generate biased responses in under-informative scenarios. By improving models' general-purpose understanding, this method helps reduce biases without relying on direct debiasing techniques. Our evaluations show that these strategies significantly reduce harmful stereotypes across multiple identity dimensions, making AI systems more fair and reliable while ensuring they remain effective in real-world applications.

The economic sustainability of an area is largely dependent on the education level of its population, yet little is known about the role cultural racism may play in academic success. The purpose of this correlational study was to evaluate the theory of cultural racism, defined as, the establishment of cultural institutions by whites/Europeans to the detriment of non-white people, as it relates to academic success at the college level. Data were collected from 100 participants from 3 predominately African American high schools in the Atlanta, Georgia area to explore whether the presence of cultural racism existed from the perspective of the participants, and the impact of cultural racism, income, and status as a first generation college student on self-reported academic success. Data were collected through a web-based survey which included the Index of Race-Related Stress questions and analyzed using logistic regression. Study results indicated a statistically significant relationship (p < .01) between the elements of cultural racism and academic success, suggesting that students who experienced cultural racism also experienced poor academic performance. Other variables, including income and whether the student was a first generation college student, also contributed to the overall collegiate academic achievement among this population. Indicators of positive social change stemming from this study include recommendations to policy makers at all levels of government to enhance diversity training for students and educators about the implications of cultural racism in order to ameliorate its negative effects, thereby promoting more economically stable and diverse communities.

The dissertation consists of an introduction and three empirical articles. The introduction gives the theoretical background, integrates the three articles, and elaborates on future research questions. The first article investigates the important decision in the lives of German and Indian university students to identify the important areas of decision making. The second article examines the structure of real life decision making and reveals the underlying factors of five major decision areas. The third article investigates the similarities and differences on decision making between German and Indian students using prior qualitative findings in a large quantitative survey.

The dissertation consists of an introduction and three empirical articles. The introduction gives the theoretical background, integrates the three articles, and elaborates on future research questions. The first article investigates the important decision in the lives of German and Indian university students to identify the important areas of decision making. The second article examines the structure of real life decision making and reveals the underlying factors of five major decision areas. The third article investigates the similarities and differences on decision making between German and Indian students using prior qualitative findings in a large quantitative survey.