Research Papers

Machine learning models in high-stakes applications, such as recidivism prediction and automated personnel selection, often exhibit systematic performance disparities across sensitive subpopulations, raising critical concerns regarding algorithmic bias. Fairness auditing addresses these risks through two primary functions: certification, which verifies adherence to fairness constraints; and flagging, which isolates specific demographic groups experiencing disparate treatment. However, existing auditing techniques are frequently limited by restrictive distributional assumptions or prohibitive computational overhead. We propose a novel empirical likelihood-based (EL) framework that constructs robust statistical measures for model performance disparities. Unlike traditional methods, our approach is non-parametric; the proposed disparity statistics follow asymptotically chi-square or mixed chi-square distributions, ensuring valid inference without assuming underlying data distributions. This framework uses a constrained optimization profile that admits stable numerical solutions, facilitating both large-scale certification and efficient subpopulation discovery. Empirically, the EL methods outperform bootstrap-based approaches, yielding coverage rates closer to nominal levels while reducing computational latency by several orders of magnitude. We demonstrate the practical utility of this framework on the COMPAS dataset, where it successfully flags intersectional biases, specifically identifying a significantly higher positive prediction rate for African-American males under 25 and a systemic under-prediction for Caucasian females relative to the population mean.

AI systems are increasingly used in high-stakes domains such as credit rating, where fairness concerns are critical. Existing fairness assessments are typically conducted by AI experts or regulators using predefined protected attributes and metrics, which often fail to capture the diversity and nuance of fairness notions held by the individuals who are affected by these systems' decisions, such as decision subjects. Recent work has therefore called for involving affected individuals in fairness assessment, yet little empirical evidence exists on how they create their own fairness criteria or what kinds of criteria they produce - knowledge that could not only inform experts' fairness evaluation and mitigation, but also guide the design of AI assessment tools. We address this gap through a qualitative user study with 18 participants in a credit rating scenario. Participants first articulated their fairness notions in their own words. Then, participants turned them into concrete quantified and operationalized fairness criteria, through an interactive prototype we designed. Our findings provide empirical evidence of the process through which people's fairness notions emerge via grounding in model features, and uncover a diverse set of individuals' custom-defined criteria for both outcome and procedural fairness. We provide design implications for processes and tools that support more inclusive and value-sensitive AI fairness assessment.

Innovative HealthTech teams develop Artificial Intelligence (AI) systems in contexts where ethical expectations and organizational priorities must be balanced under severe resource constraints. While Responsible AI practices are expected to guide the design and evaluation of such systems, they frequently remain abstract or poorly aligned with the operational realities of early-stage innovation. At the ecosystem level, this misalignment disproportionately affects disadvantaged projects and founders, therefore limiting the diversity of problem-areas under consideration, solutions, stakeholder perspectives, and population datasets represented in AI-enabled healthcare systems. Visualization provides a practical mechanism for supporting decision-making across the AI lifecycle. When developed via a rigorous and collaborative design process, structured on domain knowledge and designed around real-world constraints, visual interfaces can operate as effective sociotechnical governance artifacts enabling responsible decision-making. Grounded in innovation-oriented Human-Centered Computing methodologies, we synthesize insights from a series of design studies conducted via a longitudinal visualization research program, a case study centered on governance dashboard design in a translational setting, and a survey of a cohort of early-stage HealthTech startups. Based on these findings, we articulate design process implications for governance-oriented visualization systems: co-creation with stakeholders, alignment with organizational maturity and context, and support for heterogeneous roles and tasks among others. This work contributes actionable guidance for designing Responsible AI governance dashboards that support decision-making and accountability in early-stage health innovation, and suggests that ecosystem-level coordination can enable more scalable and diverse AI innovation in healthcare.

As climate-related hazards intensify, conventional early warning systems (EWS) disseminate alerts rapidly but often fail to trigger timely protective actions, leading to preventable losses and inequities. We introduce Climate RADAR (Risk-Aware, Dynamic, and Action Recommendation system), a generative AI-based reliability layer that reframes disaster communication from alerts delivered to actions executed. It integrates meteorological, hydrological, vulnerability, and social data into a composite risk index and employs guardrail-embedded large language models (LLMs) to deliver personalized recommendations across citizen, volunteer, and municipal interfaces. Evaluation through simulations, user studies, and a municipal pilot shows improved outcomes, including higher protective action execution, reduced response latency, and increased usability and trust. By combining predictive analytics, behavioral science, and responsible AI, Climate RADAR advances people-centered, transparent, and equitable early warning systems, offering practical pathways toward compliance-ready disaster resilience infrastructures.

As financial institutions increasingly adopt machine learning for credit risk assessment, the persistence of algorithmic bias remains a critical barrier to equitable financial inclusion. This study provides a comprehensive audit of structural gender bias within the Taiwan Credit Default dataset, specifically challenging the prevailing doctrine of "fairness through blindness." Despite the removal of explicit protected attributes and the application of industry standard fairness interventions, our results demonstrate that gendered predictive signals remain deeply embedded within non-sensitive features. Utilizing SHAP (SHapley Additive exPlanations), we identify that variables such as Marital Status, Age, and Credit Limit function as potent proxies for gender, allowing models to maintain discriminatory pathways while appearing statistically fair. To mathematically quantify this leakage, we employ an adversarial inverse modeling framework. Our findings reveal that the protected gender attribute can be reconstructed from purely non-sensitive financial features with an ROC AUC score of 0.65, demonstrating that traditional fairness audits are insufficient for detecting implicit structural bias. These results advocate for a shift from surface-level statistical parity toward causal-aware modeling and structural accountability in financial AI.

The climate crisis requires responses that integrate scientific, ethical, social, and technological perspectives. Artificial intelligence (AI) has emerged as a powerful tool in climate modelling, environmental monitoring, and energy optimisation, yet its growing use also raises critical environmental, ethical, legal, and social questions. This contribution examines the ambivalent role of AI in the ecological crisis, addressing both its promises and its risks. On the one hand, AI supports improvements in climate forecasting, renewable energy management, and real-time detection of environmental degradation. On the other hand, the energy demands of data centres, resource-intensive hardware production, algorithmic bias, corporate concentration of power, and technocratic decision-making reveal contradictions that challenge its sustainability. The discussion explores these issues through interdisciplinary lenses, including environmental ethics, philosophy of technology, and legal governance, and concludes with recommendations for socially just, ecologically responsible, and democratically accountable uses of AI. Rather than assuming AI as an inherently sustainable solution, this analysis argues that its contribution to climate action depends fundamentally on the values, institutions, and power structures that shape its development.

From school playgrounds to corporate boardrooms, status hierarchies -- rank orderings based on respect and perceived competence -- are universal features of human social organization. Language models trained on human-generated text inevitably encounter these hierarchical patterns embedded in language, raising the question of whether they might reproduce such dynamics in multi-agent settings. This thesis investigates when and how language models form status hierarchies by adapting Berger et al.'s (1972) expectation states framework. I create multi-agent scenarios where separate language model instances complete sentiment classification tasks, are introduced with varying status characteristics (e.g., credentials, expertise), then have opportunities to revise their initial judgments after observing their partner's responses. The dependent variable is deference, the rate at which models shift their ratings toward their partner's position based on status cues rather than task information. Results show that language models form significant status hierarchies when capability is equal (35 percentage point asymmetry, p < .001), but capability differences dominate status cues, with the most striking effect being that high-status assignments reduce higher-capability models' deference rather than increasing lower-capability models' deference. The implications for AI safety are significant: status-seeking behavior could introduce deceptive strategies, amplify discriminatory biases, and scale across distributed deployments far faster than human hierarchies form organically. This work identifies emergent social behaviors in AI systems and highlights a previously underexplored dimension of the alignment challenge.

Artificial Intelligence (AI) readiness in the Global South extends beyond infrastructure to include curriculum design, workforce development, and cross-sector collaboration. Bangladesh, ranked 82nd in the 2023 Oxford Insights AI Readiness Index, exhibits significant deficits in technology capacity and research ecosystems, despite strong governmental visions. While HCI and ICTD research have explored digital inclusion and responsible AI, little empirical work examines how educational, industrial, and policy domains intersect to shape readiness. We present a multi-method qualitative study of AI readiness in Bangladesh, combining institutional analyses, 59 stakeholder interviews, and curriculum benchmarking against global exemplars. Findings reveal outdated curricula, limited faculty upskilling, inadequate computing resources, entrenched gender disparities, and the near-total absence of AI ethics instruction. We contribute empirical mapping of current practices, identification of structural and cultural barriers, and actionable pathways for embedding human-centered, inclusive, and responsible AI practices into national agendas, advancing equitable innovation in emerging AI ecosystems.

Digital platforms have an ever-expanding user base, and act as a hub for communication, business, and connectivity. However, this has also allowed for the spread of hate speech and misogyny. Artificial intelligence models have emerged as an effective solution for countering online hate speech but are under explored for low resource and code-mixed languages and suffer from a lack of interpretability. Explainable Artificial Intelligence (XAI) can enhance transparency in the decisions of deep learning models, which is crucial for a sensitive domain such as hate speech detection. In this paper, we present a multi-modal and explainable web application for detecting misogyny in text and memes in code-mixed Hindi and English. The system leverages state-of-the-art transformer-based models that support multilingual and multimodal settings. For text-based misogyny identification, the system utilizes XLM-RoBERTa (XLM-R) and multilingual Bidirectional Encoder Representations from Transformers (mBERT) on a dataset of approximately 4,193 comments. For multimodal misogyny identification from memes, the system utilizes mBERT + EfficientNet, and mBERT + ResNET trained on a dataset of approximately 4,218 memes. It also provides feature importance scores using explainability techniques including Shapley Additive Values (SHAP) and Local Interpretable Model Agnostic Explanations (LIME). The application aims to serve as a tool for both researchers and content moderators, to promote further research in the field, combat gender based digital violence, and ensure a safe digital space. The system has been evaluated using human evaluators who provided their responses on Chatbot Usability Questionnaire (CUQ) and User Experience Questionnaire (UEQ) to determine overall usability.

Asymptotic Universal Alignment: A New Alignment Framework via Test-Time Scaling

Yang Cai, Weiqiang Zheng

Aligning large language models (LLMs) to serve users with heterogeneous and potentially conflicting preferences is a central challenge for personalized and trustworthy AI. We formalize an ideal notion of universal alignment through test-time scaling: for each prompt, the model produces $k\ge 1$ candidate responses and a user selects their preferred one. We introduce $(k,f(k))$-robust alignment, which requires the $k$-output model to have win rate $f(k)$ against any other single-output model, and asymptotic universal alignment (U-alignment), which requires $f(k)\to 1$ as $k\to\infty$. Our main result characterizes the optimal convergence rate: there exists a family of single-output policies whose $k$-sample product policies achieve U-alignment at rate $f(k)=\frac{k}{k+1}$, and no method can achieve a faster rate in general. We show that popular post-training methods, including Nash learning from human feedback (NLHF), can fundamentally underutilize the benefits of test-time scaling. Even though NLHF is optimal for $k=1$, sampling from the resulting (often deterministic) policy cannot guarantee win rates above $\tfrac{1}{2}$ except for an arbitrarily small slack. This stems from a lack of output diversity: existing alignment methods can collapse to a single majority-preferred response, making additional samples redundant. In contrast, our approach preserves output diversity and achieves the optimal test-time scaling rate. In particular, we propose a family of symmetric multi-player alignment games and prove that any symmetric Nash equilibrium policy of the $(k+1)$-player alignment game achieves the optimal $(k,\frac{k}{k+1})$-robust alignment. Finally, we provide theoretical convergence guarantees for self-play learning dynamics in these games and extend the framework to opponents that also generate multiple responses.

Ways in which people's opinions change are, without a doubt, subject to a rich tapestry of differing influences. Factors that affect how one arrives at an opinion reflect how they have been shaped by their environment throughout their lives, education, material status, what belief systems are they subscribed to, and what socio-economic minorities are they a part of. This already complex system is further expanded by the ever-changing nature of one's social network. It is therefore no surprise that many models have a tendency to perform best for the majority of the population and discriminating those people who are members of various marginalized groups . This bias and the study of how to counter it are subject to a rapidly developing field of Fairness in Social Network Analysis (SNA). The focus of this work is to look into how a state-of-the-art model discriminates certain minority groups and whether it is possible to reliably predict for whom it will perform worse. Moreover, is such prediction possible based solely on one's demographic or topological features? To this end, the NetSense dataset, together with a state-of-the-art CoDiNG model for opinion prediction have been employed. Our work explores how three classifier models (Demography-Based, Topology-Based, and Hybrid) perform when assessing for whom this algorithm will provide inaccurate predictions. Finally, through a comprehensive analysis of these experimental results, we identify four key patterns of algorithmic bias. Our findings suggest that no single paradigm provides the best results and that there is a real need for context-aware strategies in fairness-oriented social network analysis. We conclude that a multi-faceted approach, incorporating both individual attributes and network structures, is essential for reducing algorithmic bias and promoting inclusive decision-making.

Sentiment analysis focuses on identifying the emotional polarity expressed in textual data, typically categorized as positive, negative, or neutral. Hate speech detection, on the other hand, aims to recognize content that incites violence, discrimination, or hostility toward individuals or groups based on attributes such as race, gender, sexual orientation, or religion. Both tasks play a critical role in online content moderation by enabling the detection and mitigation of harmful or offensive material, thereby contributing to safer digital environments. In this study, we examine the performance of three transformer-based models: BERT-base-multilingual-cased, RoBERTa-base, and XLM-RoBERTa-base with the first eight layers frozen, for multilingual sentiment analysis and hate speech detection. The evaluation is conducted across five languages: English, Korean, Japanese, Chinese, and French. The models are compared using standard performance metrics, including accuracy, precision, recall, and F1-score. To enhance model interpretability and provide deeper insight into prediction behavior, we integrate the Local Interpretable Model-agnostic Explanations (LIME) framework, which highlights the contribution of individual words to the models decisions. By combining state-of-the-art transformer architectures with explainability techniques, this work aims to improve both the effectiveness and transparency of multilingual sentiment analysis and hate speech detection systems.