Speakers & Presenters

Opening Remarks

About the Speaker

Natasha M. Dartigue is the Maryland State Public Defender and a nationally recognized architect of justice reform. Appointed in 2022, she is the first person of color to lead the Maryland Office of the Public Defender, building on a career that began with a clerkship for the late Judge Roger W. Brown and rose through the ranks as a trial attorney, felony trial supervisor, and Deputy District Public Defender for Baltimore City. A 1995 graduate of Howard University School of Law, she added to that record of civic leadership as a member of the Leadership Maryland Class of 2025. That depth of frontline experience now anchors her role as a national voice for the defense bar: Ms. Dartigue was elected to the Board of Directors of the National Association of Criminal Defense Lawyers (NACDL) in 2024, and in 2026 she assumed the presidency of the Maryland State Bar Association, after serving a term as president-elect.

Ms. Dartigue's leadership extends well beyond her office, reshaping how Maryland approaches systemic injustice. In 2023, she co-founded and co-chaired the Maryland Equitable Justice Collaborative (MEJC) alongside Attorney General Anthony Brown, an unprecedented alliance between prosecution and defense that produced 18 landmark recommendations to dismantle mass incarceration and reduce racial disparities in Maryland's justice system. She then translated that research into action, launching the Maryland Justice Partnership (MJP) in January 2026, a statewide implementation framework, not a study group, designed to turn MEJC's recommendations into enacted law and measurable reform. She also expanded the state's capacity to correct wrongful convictions by establishing the second Innocence Project Clinic at the University of Maryland Francis King Carey School of Law, giving Maryland public defenders innocence-work partnerships with both of the state's law schools for the first time.

A daughter of Haitian immigrants, Ms. Dartigue describes her historic appointment as one that “redefines what leadership looks like” and offers hope to children of immigrants too often overlooked. Her record of sustained excellence has been repeatedly recognized: she was named to The Daily Record's Top 100 Women in both 2018 and 2024, earning induction into its Circle of Excellence, reserved for honorees recognized multiple times for sustained achievement, and was placed on the 2024 Criminal Law Power List for driving change and impact across the field. From the courtroom to the halls of the General Assembly, Natasha Dartigue stands as a model of how a career centered on principled leadership can drive systemic change.

Equitable AI in the post SFFA SCOTUS decision in Admissions

About the Speaker

Dr. Juan E. Gilbert is the Andrew Banks Family Preeminence Endowed Professor and a Distinguished Professor at the University of Florida where he leads the Computing for Social Good Lab. He is a Fellow of the ACM, IEEE, and American Association for the Advancement of Science. He is a member of the American Academy of Arts and Sciences and the National Academy of Inventors. In 2012, he received the Presidential Award for Excellence in Science, Mathematics, and Engineering Mentoring from President Barack Obama. In 2023, he received the National Medal of Technology and Innovation from President Joe Biden.

Must AI Mean What We Say?

About the Speaker

Jeffrey Bardzell is Peterson Family Dean of the School of Arts and Sciences at Worcester Polytechnic Institute, where he provides leadership in higher education strategy, liberal arts in polytechnic contexts, and AI in teaching and learning. Internationally recognized for his interdisciplinary scholarship, Bardzell was originally trained in comparative literature and philosophy and subsequently transitioned to informatics, human-computer interaction, and design research. Before joining WPI, Bardzell served as both dean and vice provost at the University of North Carolina at Chapel Hill, with a portfolio that bridged institutional planning, research-driven academic strategy, and curriculum-wide AI implementation. Bardzell is also an active writer and public scholar, maintaining a Substack that explores higher education leadership, AI in academia, and the future of liberal arts education.

Human-AI Collaboration in Federal Performance Management: Integrating Relational Competencies into AI-Supported Appraisal Systems

Abstract

Artificial intelligence is increasingly being integrated into public-sector work, yet many government performance appraisal systems remain designed to evaluate technical outputs rather than the relational and judgment-based competencies that AI cannot easily replicate. This presentation examines how AI may reshape federal performance management by changing the relative value of different employee competencies.

Using the U.S. Department of Commerce's CD-541 performance appraisal framework as an illustrative case, the presentation explores the extent to which existing appraisal systems emphasize measurable technical performance while providing limited mechanisms for evaluating collaboration, trust-building, communication, ethical judgment, and other relational competencies. As AI assumes a greater role in drafting, data analysis, and routine administrative tasks, these human-centered competencies may become increasingly important to organizational effectiveness.

The presentation proposes a framework for integrating relational competencies into AI-supported appraisal systems and discusses practical approaches to measuring and evaluating these competencies within existing administrative constraints. The discussion also considers implications for accountability, workforce development, and the future of public-sector performance management.

The session contributes to ongoing conversations about human-AI collaboration by examining not only how AI can support performance evaluation, but also how appraisal systems themselves may need to evolve to recognize the distinctly human contributions that remain essential in public service.

About the Presenter

Omololu Adeyemi is a Doctoral Student in Public Administration at the University of Baltimore and a U.S. Army Veteran. His research focuses on performance management, organizational effectiveness, culturally competent leadership, and the implications of artificial intelligence for public-sector workforce management. His current work examines how federal performance appraisal systems can be redesigned to better capture relational competencies, human judgment, and accountability in AI-enabled government workplaces.

The Augmented Scientist: Human-AI Collaboration in Scientific Discovery

Abstract

Artificial Intelligence has long been embedded in scientific research through modeling, image classification, and computational analysis. However, generative AI marks a fundamental shift in how scientists can approach the research cycle. Specialized platforms and thoughtful prompting can provide tools to help researchers with ideation, literature synthesis and gap analysis, experimental design, data analysis, and ultimately discovery. This rapidly evolving relationship between human and AI begs the question: How can scientists leverage AI to enhance discovery while preserving human agency, expertise, and the all-important critical thinking and judgment?

The Augmented Scientist is a human-centered framework for integrating generative AI into scientific workflows. Rather than centering AI as the knowledge generator, this framework integrates AI into a research team as a collaborator capable of supporting tasks such as pattern identification, code generation, and increasing the pace and scope of discovery. But the team lead is always the human scientist, serving as director and principal investigator, with the domain knowledge and decision-making authority.

In this talk I will explore the practical applications of generative AI across the research lifecycle, highlight various tools and their applications, and demonstrate how these platforms can strengthen scientific workflows when used critically and intentionally. I will also present real-world case studies from ecological research, from elevated CO2 x nutrient studies to AI-supported seedbank platforms, that illustrate the promise and limitations of genAI in scientific inquiry.

Finally, this talk will address broader issues of trust, over-reliance, hallucinations, and access, emphasizing the need for AI literacy and pathways for broader participation in research. In an era where AI is rapidly accelerating discovery, it is paramount that we ensure equitable access to the resources necessary so that the next generation of scientists can fully participate in the future of augmented discovery.

About the Presenter

Dr. Jere Boudell is Associate Dean of the College of STEM, Director of the Applied AI Institute, and Professor of Biology. As a plant ecologist and technologist, she works at the intersection of environment, artificial intelligence, and human learning. Her research and leadership focus on using AI to augment scientific discovery, strengthen education, and support ecological stewardship. Through faculty and student workshops, student innovation programs, and interdisciplinary collaborations, she helps build more inclusive and human-centered futures for science and technology.

Measuring the Local and Global Impact of Corporate Social Responsibility Using Artificial Intelligence

Abstract

Co-authors: Denis Gracanin, Associate Professor of Computer Science, Virginia Tech; Rachel Soloveichik, Economist, Bureau of Economic Analysis; Majid Behravan, Postdoctoral Researcher, Morgan State University; Sezer Dumen, Graduate Research Assistant, Virginia Tech

This study introduces an AI-powered methodology that uses Natural Language Processing (NLP) and Large Language Models (LLMs) to construct a comprehensive database of U.S. firms’ corporate social responsibility (CSR) activities from 2000-2025. The database extracts and standardizes nearly 300,000 observations, including 94,655 firm-donation-year records, from 1.2 billion news articles, SEC filings, sustainability reports, and social media. The approach also enables valuation of non-monetary contributions, including approximately $9 billion in employee volunteering across more than 12,000 records, using replacement-cost methods. Overall, CSR donations increased from under $1 billion in 2000 to over $2.5 billion by 2025, driven in part by a post-2018 surge in non-monetary activities. Contributions are highly concentrated among industries. Statistical analysis further shows that donations decline significantly with distance (i.e., geographic proximity to the recipient), especially for foundation-backed firms and for local categories such as health and education. By overcoming longstanding fragmentation in CSR data, this scalable approach enables more robust empirical analysis of key drivers, including firm size, governance, and market power. The database is useful for our national accounting system and policy design.

About the Presenter

Dina El Mahdy is a Professor of Accounting at Morgan State University. Her research spans accounting education and technology, corporate social responsibility, and corporate governance. Her work has been cited by the Securities and Exchange Commission, the Public Company Accounting Oversight Board, Forbes, The New York Post, and Bloomberg. Her publications have appeared in top journals such as Auditing: A Journal of Practice and Theory, Accounting and Finance, Review of Quantitative Finance and Accounting, Issues in Accounting Education, Journal of Risk and Financial Management, International Journal of Accounting and Information Management, and Journal of Forensic and Investigative Accounting.

Mediated Recognition: Perceived AI Appreciation and Evaluative Preference

Abstract

As artificial intelligence (AI) increasingly assumes evaluative roles once reserved for humans, understanding who prefers AI over human judgement, and why, has become a critical question for social science. Existing research on AI preferences explains variation through individual dispositions or demographic traits, often neglecting the broader social structures and cultural contexts in which such preferences are formed. This paper introduces the concept of perceived AI appreciation-the extent to which individuals believe their personal qualities can be detected, interpreted, and given evaluative weight by AI systems-to recast AI preference as a problem of mediated recognition. Treating AI as a mediated communicative agent, we argue that people anticipate which of their qualities could be mediated and rewarded by the evaluator and orient toward the system they expect to appreciate their particular advantages. Drawing on Bourdieu’s theory of capital, we contend that forms of capital differ in how much their recognition depends on interpersonal interaction: economic capital is readily recognized with little human mediation, whereas cultural and social capital are sustained through ongoing interpersonal recognition and are therefore perceived as less appreciable by AI. Using survey data from urban China we find that individuals with higher economic capital prefer AI evaluation, while those with more omnivorous cultural capital or political social capital prefer human evaluators. By reframing AI preference as a problem of perceived mediated recognition, we offer a relational account of human-AI interaction and shows how preference for AI is structured by social positionality rather than fixed individual traits.

About the Presenter

Bio coming soon.

AI Governance for Human-AI Collaboration in Higher Education

Abstract

The rapid proliferation of Generative Artificial Intelligence (GenAI) across higher education has created new opportunities for teaching, learning, research, and institutional effectiveness while simultaneously exposing limitations in current governance approaches, which were designed for more stable technological environments. Existing institutional responses have largely emphasized policy development, compliance, and risk mitigation; however, the pace, complexity, and evolving capabilities of GenAI require governance frameworks that can adapt continuously to technological, organizational, and societal change. The purpose of this presentation is to introduce an adaptive AI governance model that addresses the unique governance challenges associated with human-GenAI collaboration in higher education. Grounded in adaptive governance theory, organizational learning, and emerging scholarship on human-AI interaction, the model conceptualizes governance not as a static set of rules or controls but as an ongoing institutional process that supports responsible innovation, continuous learning, and ethical decision-making. The presentation argues that higher education institutions require governance structures capable of balancing innovation with accountability while preserving the central role of human judgment in environments increasingly influenced by GenAI systems. Unlike many sectors characterized by centralized authority and standardized decision-making processes, colleges and universities operate within decentralized systems of shared governance that distribute authority among faculty, administrators, governing boards, and other stakeholders. Institutional commitments to academic freedom, intellectual inquiry, equity, inclusion, and student success further complicate the adoption and oversight of GenAI technologies. These characteristics necessitate governance approaches that are flexible, participatory, and responsive to context rather than prescriptive and uniform.

About the Presenter

Dr. Muddassir Siddiqi is President of the College of DuPage, one of the nation’s largest community colleges. A higher education leader, scholar, and practitioner, he brings more than two decades of executive experience advancing institutional transformation, workforce development, student success, and organizational effectiveness. He has held senior leadership positions at Houston Community College, City Colleges of Chicago, and Morton College, leading initiatives in innovation, digital learning, and applied education. Dr. Siddiqi holds doctoral and graduate degrees in education, business, and technology. He is a prolific scholar, graduate faculty member, and frequent national and international presenter whose work focuses on leadership, governance, strategy, and continuous improvement in higher education.

AI Testing, Evaluation, Verification, and Validation (TEVV) for Neurodiversity: Building Practical TEVV Pipelines for Accessible and Inclusive AI

Abstract

How do we know an AI system works well for neurodiverse users, not just for an assumed average user? This talk introduces AITEVVA, AI Testing, Evaluation, Verification, and Validation for Accessibility, as an accessibility-first TEVV framework for evaluating whether AI systems are interpretable, emotionally safe, adaptable, and meaningfully usable for neurodiverse users. I will use AITEVVA as a foundation to present a practical approach to TEVV that centers accessibility, neurodiversity, and real-world implementation rather than treating inclusion as a late-stage add-on.

This session frames AITEVVA as an extensible implementation model for responsible AI practice rather than presenting it as a finished checklist. I will offer examples of waysthat  organizations can apply the framework in campuses, learning environments, and human-centered AI systems by identifying high-risk interaction points, collecting behavioral and self-report evidence, and using TEVV results to refine design choices, modality options, and governance processes. The goal is to think through your roadmap for building institutional AI evaluation infrastructures that are practical, accountable, and neurodiversity-affirming.

About the Presenter

Gabriella Waters is an artificial intelligence researcher committed to increasing the diversity of thought in technology. She is an advocate for interdisciplinary collaborations that advance innovation, responsibility, explainability, transparency, and ethics in AI development and application. Her research explores the intersections of human neurobiology & learning, the quantification of ethics in AI/ML systems, neuro-symbolic architectures, embodied AI, AI TEVV, and the design of intelligent systems that leverage these foundations for enhanced human-computer synergy. She is especially focused on developing technological innovations that support neurodiverse populations. She serves as the inaugural Director of the Center for Responsible AI. She also directs the Cognitive and Neurodiversity AI (CoNA) and Robotics and Digital Twin (RaDiT) Labs.

Generative AI and Organizational Knowledge Management

Abstract

Co-authors: Jigish Zaveri (Morgan State University)

Generative Artificial Intelligence (GenAI) is rapidly transforming organizational knowledge work by supporting knowledge creation, retrieval, transfer, and application. Although recent studies suggest that GenAI has the potential to reshape organizational knowledge management, existing research has been largely conceptual, agenda-setting, or based on limited qualitative evidence. Consequently, there remains a lack of empirical organization-level evidence explaining how organizational use of GenAI influences knowledge-management processes and organizational knowledge retention.

This research-in-progress proposes an empirical framework grounded in the Organizational Knowledge Management Process Framework to examine how organizational GenAI use influences the four-core knowledge-management processes, knowledge creation, knowledge storage and retrieval, knowledge transfer, and knowledge application, and how these processes collectively affect organizational knowledge retention. The proposed conceptual model also examines the direct relationship between organizational GenAI use and organizational knowledge retention to determine whether the effects are fully or partially mediated by knowledge-management processes.

The study will employ a survey-based research design using validated measurement scales adapted from knowledge-management literature. Data will be collected from organizations using GenAI to support knowledge-intensive work and analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The expected contributions include extending knowledge-management theory to the emerging context of GenAI, providing empirical evidence on how GenAI influences organizational knowledge retention, and offering practical guidance for organizations seeking to leverage GenAI while preserving valuable organizational knowledge resources.

Human-AI Collaboration and Employee Perceptions of Productivity and Pay Equity in Financial Services

Abstract

As artificial intelligence (AI) becomes increasingly embedded in financial services, its role is shifting from task automation to collaborative augmentation of human work. This transition raises critical questions about agency, trust, and belonging within AI-mediated workplaces. While prior research has emphasized efficiency gains and technical performance, less attention has been given to how employees interpret AI’s influence on their productivity and perceptions of pay equity. This study addresses that gap by examining how AI adoption shapes employee sensemaking in contexts where algorithmic systems influence decision-making, performance evaluation, and compensation structures.

Grounded in the Technology Acceptance Model (TAM), Equity Theory, and Organizational Justice frameworks, this qualitative study adopts an interpretivist approach to explore employee experiences with AI-enabled work environments. Data will be collected through semi-structured interviews with 24-30 professionals in the financial services sector who actively interact with AI systems. Reflexive thematic analysis will be used to identify patterns in how employees perceive AI’s impact on their productivity, autonomy, and fairness of compensation, as well as their sense of inclusion and voice in AI-driven processes.

Preliminary insights suggest that while AI is often associated with increased efficiency and output, it simultaneously introduces tensions around transparency, control, and equitable reward distribution. Employees report concerns about opaque algorithmic evaluations, misalignment between effort and compensation, and the redistribution of decision authority from humans to systems. These findings highlight a critical paradox: AI can enhance measurable productivity while undermining perceived fairness, trust, and employee agency if governance and communication structures are not carefully designed.

This study contributes to ongoing discussions on human-AI collaboration by foregrounding employee perceptions as a central component of effective and ethical AI integration. It extends current literature by linking productivity outcomes with justice perceptions and highlighting the role of organizational context in shaping these dynamics. The study offers practical implications for organizations seeking to balance performance optimization with equitable workplace practices, emphasizing the need for transparent AI governance, participatory system design, and alignment between technological capabilities and human values.

Stepping Outside the Stack

About the Speaker

Reva Schwartz is a research scientist, linguist, and co-founder of Civitaas Insights, pioneering real-world AI evaluation. She also directs the Forum for Real-World AI Measurement and Evaluation at Virginia State University. With over 20 years in federal service, Reva's career includes roles as a forensic scientist and advisor to the intelligence community and other federal agencies. At NIST, she founded the ARIA program and served as core author of the AI Risk Management Framework and lead architect of the AI RMF Playbook. Her work integrates machine learning, measurement science, and social science to improve AI evaluation in practical, real-world applications.

Empowering Health Equity: Training HBCU Students in Data Literacy and Community-Engaged Research

Abstract

Historical redlining represents one of the earliest and most consequential failures of data governance, where racially biased data about marginalized communities was produced, interpreted, and enforced without community participation or consent. Using Baltimore, Maryland’s “Black Butterfly” geography as a central case, this presentation demonstrates how data-driven urban planning and housing appraisal systems institutionalized socio-spatial segregation and generated persistent public health inequities, including reduced life expectancy, elevated chronic disease prevalence, and higher infant mortality in historically disinvested neighborhoods. The Baltimore case further illustrates how contemporary algorithmic systems, ranging from property valuation models to public health surveillance and smart city technologies, inherit historical bias, contributing to digital redlining, data misrepresentation, and delayed health interventions. Positioning health education as a critical intervention, this work argues that addressing equity requires not only improved data quality but also democratic control, consent, and accountability. Preparing HBCU students as future health education and public health practitioners is central to this approach. These students are equipped with skills in data literacy, ethical AI, and community-engaged research to effectively evaluate and plan interventions that address built environment gaps. By advancing community-driven health equity and policy action, they play a crucial role in transforming public health practices to be more inclusive and responsive to the needs of marginalized communities.

About the Presenter

As an Assistant Professor at Morgan State University, School of Community Health & Policy, Health Education program, Dr. Robin Butler presents over 20 years of experience working with distressed communities. Dr. Butler’s interdisciplinary research focus supports health informatics, STI prevention, health education, college student career pathways, and culturally tailored health literacy interventions messaging through barbershop dialogue and engagement. Her research also focuses on built community environments and data sovereignty to address inequitable health outcomes using artificial intelligence and machine learning.

More recently, Dr. Butler actively serves as a Co-Principal Investigator for an HPV uptake study in collaboration with Johns Hopkins University and the Sidney Kimmel Comprehensive Cancer Center. Her efforts have been recognized with grants from the NIH, Bristol Myers Squibb, JP Morgan, the Maryland Cigarette Restitution Fund, and the Morgan State University Office of Academic Affairs. She also serves as a co-advisor for the Morgan Chapter's My Sisters Keeper and as a practicum coordinator for the Department of Public and Allied Health’s Health Education senior students.

A Trust Map for AI: Conceptualizing Trustworthy AI as a Socio-Technical Research Practice

Abstract

In our work, we propose a framework for making sense of the idea of “trustworthy AI” (TAI) as a research practice. While being contested, the perspective of trust and trustworthiness is essential for understanding and accounting for the societal impact and ethical dimensions of AI technologies, as they require at least a certain amount of trust to be practically useful. However, we do not define trustworthiness as a set of properties or characteristics that are attributed to an AI system or application but as a lens through which AI technologies and their associated research and development practices can be analyzed, designed for and governed by. This lens allows us to understand trustworthy AI as the socio-technical research practice of interrogating and building contextual trust relationships around AI technologies. To conceptualize this practice, we bring the philosophical and social science literature on trust and trustworthiness (for AI) into conversation with the experiences and expertise of a diverse set of AI researchers across institutions and disciplines such as computer science, human-computer interaction and STS. Based on this conversation, we map out and describe different kinds of actors that are involved in this practice and the activities that constitute it. We also explore how our framework can answer to common challenges to the idea of trustworthy AI, how it could support those engaged in trustworthy AI efforts to articulate their presuppositions about trust and trustworthiness in the context of their work, and how we may enable communication and knowledge exchange across projects, approaches and disciplines. In the future, we hope that our framework can support research teams in making informed and responsible choices on how to operationalize trustworthiness, in selecting research participants and methods, and in engaging existing institutional expertise, unravelling chains of trust that may be hidden behind evaluation metrics. To demonstrate how this may work in practice, we showcase our framework's contribution in the context of ongoing AI research.

About the Presenter

Laura Fichtner is a postdoctoral associate at the University of Maryland and the Institute for Trustworthy AI in Law & Society (TRAILS). Her research focuses on the intersections of information technology, ethics and politics. In her current project, Laura explores how different research practices seek to make AI technologies more trustworthy and if and how participatory methods are engaged in this effort.

Autonomous AI Accountability: Building Predictive Governance Before Harm Materializes

Abstract

On August 2, 2026, the European Union's AI Act enters full application for high-risk systems. Article 12 will require automatic, technical, lifecycle event logging for every AI system that affects hiring, credit, healthcare, access, or resource allocation. This session takes place days before that deadline, and explains what compliance actually has to look like.

Earlier this year, a federal court allowed Mobley v. Workday to proceed as a class action over AI hiring tools that allegedly discriminated against Black, older, and disabled applicants. The discrimination was not authorized by any compliance team. It was a behavior of the system no one was watching. This is the gap predictive governance closes.

The session introduces Autonomous Fingerprinting: a human accountability mechanism that builds behavioral baselines for autonomous AI agents and detects deviation from those baselines before policy violations or harm events occur. Unlike reactive auditing, which documents what went wrong after the fact, Autonomous Fingerprinting surfaces the conditions for governance failure as they form, while accountable humans can still intervene.

Attendees will leave with a working definition of predictive governance, a four-part framework for the infrastructure organizations must build before deploying autonomous AI responsibly at scale, and a clear-eyed view of what is required to make autonomous AI accountability as rigorous as the human accountability frameworks it is replacing.

The session closes by previewing the AGE Protocol (Autonomous Governance Exchange), an open standard for cross-organizational autonomous AI accountability now under public development.

About the Presenter

Erica L. Miller is a Senior Lead, Enterprise AI Architect and AI Governance Lead at Under Armour, with 20+ years building the infrastructure enterprises need to operationalize AI. She is the founder of Emethia, an AI Governance Operations platform, and creator of the ADOPT Framework™, a five-gate system for AI product validation. Her work sits at the intersection of AI adoption, enablement, and governance. She speaks, consults, and builds at the point where AI ambition meets operational reality.

From Sycophancy to Socratic Dialogue: Using Interactional Architecture as an Agency-shield in AI-assisted Inquiry

Abstract

While current UNESCO, national and state-level AI frameworks successfully establish the value of human agency, there remains a critical need for operationalized procedural mechanisms to counter the "algorithmic sycophancy" inherent in commercial Generative Artificial Intelligence (AI), which prioritize efficiency and user approval over critical inquiry. Literature suggests that AI creates a "frictionless" environment that risks pervasive cognitive offloading and agency surrender, creating a unique vulnerability for K-12 learners and novice researchers who may lack the specialized literacy to detect agency violations or the subtle algorithmic nudges that undermine independent inquiry.

This paper proposes the shift from "Prompt Engineering" to "Interactional Architecture" through the use of Learner-Designed Interactional Protocols (LDIPs). Unlike teacher-facing prompt-generators that prioritize instructional output, LDIPs are learner-centric frameworks that empower learners to treat the AI interface as a designable space and "shield" their epistemic sovereignty, forcing the AI out of its default sycophancy and cognitive path of least resistance into a more ethical and rigorous Socratic partnership.

To evaluate the efficacy of these protocols, this study employs a Comparative Discourse Analysis (CDA) of two controlled AI-assisted inquiry sessions. The first session follows a standard, unregulated "helpful" interaction, while the second utilizes a prototype LDIP designed to enforce Socratic constraints. The analysis focuses on "Turn-by-Turn" dynamics, coding for Direct Instruction, Agency Violations, Subtle Algorithmic Nudges, Productive Friction, AI Hallucinations, and Metacognitive Scaffolding.

Initial coding and pilot analysis suggest a marked shift in the "User-to-AI" word count ratio, a turn-based shift from answer-retrieval to metacognitive scaffolding, and a decrease in agency violations and AI hallucinations when a protocol is active. However, preliminary observations also reveal a "Sycophancy Threshold" and a phenomenon of "Contextual Decay", where the model's default training begins to override user-imposed constraints over extended dialogues. These early findings highlight the protocol not as a perfect solution, but as a prototype to mitigate cognitive agency surrender and a critical diagnostic tool for measuring the limits of human-imposed friction in commercial AI systems. This research argues that the capacity to design one’s own interactional architecture is a foundational AI literacy skill. The study concludes by offering a rubric for "Agency-by-Design" pedagogy, suggesting that the next generation of AI integration across K-12 and higher education must move beyond output-evaluation and toward the active governance of the interactional medium itself.

Keywords: Epistemic Sovereignty, Human Agency, Generative AI, Interactional Architecture, Algorithmic Sycophancy

About the Presenter

Dr. Jiang Pu has been a dedicated advocate and practitioner of whole-child, competency-based learning for over 15 years. She bridges educational research and systemic reform by architecting human-centered, future-ready, and nature-positive learning environments rooted in empathy for the learner and the planet. Her expertise spans consulting for the World Bank's Renovation of General Education Project with Vietnam’s Ministry of Education and Training, designing innovative K-16 curricula, and researching responsible AI in education. Using qualitative and mixed-method insights, she ensures digital transformations remain ethical, inclusive, equitable, and profoundly human.

Who Can Stop the Machine? Building Institutions That Can Govern AI

Abstract

Artificial intelligence is rapidly transforming how decisions are made across government, business, education, healthcare, and national security. Yet many discussions about AI governance continue to focus on technical safeguards or whether a human is "in the loop." While those issues matter, they do not answer the fundamental governance question: Who has the authority to stop the machine?

This presentation argues that effective AI governance is fundamentally an institutional challenge. Organizations must establish clear authority, accountability, oversight, monitoring, escalation procedures, and mechanisms for intervention before AI systems are deployed. Governance is not simply about ensuring human participation; it is about ensuring that people have the authority, information, and institutional support necessary to intervene when AI systems produce harmful, unlawful, or unintended outcomes.

Drawing on legal, regulatory, and public policy perspectives, this session introduces a practical framework for building institutions capable of governing AI responsibly. Participants will examine how governance begins before deployment through procurement, organizational design, risk management, and clearly defined lines of accountability. The presentation will also explore how institutions can balance innovation with transparency, public trust, and responsible decision-making as AI becomes embedded in critical societal functions.

Attendees will leave with a practical framework for evaluating AI governance beyond technical compliance and will gain actionable strategies for designing institutional structures that enable responsible AI deployment.

About the Presenter

Arthur D. Sidney is an attorney, AI governance strategist, and Founder & Principal of The Sidney Group PLLC. A former Chief of Staff and Chief Counsel in the U.S. House of Representatives, he has also served as Vice President of Government Affairs at CCIA and Senior Vice President at Forbes Tate Partners. He has taught law and policy as an adjunct professor, presented on panels in the U.S. and abroad, and written extensively on AI governance, institutional accountability, and technology policy. His work focuses on helping organizations build governance frameworks that keep AI systems accountable, transparent, and subject to meaningful human authority.

Intersectionality in AI - Examining Trust and Safety Systems in Social Media

Abstract

To address issues such as online gender-based violence and gendered disinformation, social media networks rely on a combination of AI tools and human review. However, these trust and safety systems often fall short in part because they ignore the realities of intersectionality. For example, they may focus on intersecting identities to improve their recommender systems or analysis of user behavior but ignore intersectional analyses of power. As a result these systems may exacerbate the problems they are meant to solve. In this talk I examine these failures and offer recommendations for how social media trust and safety teams can use an intersectional approach to their work.

About the Presenter

Dhanaraj Thakur leads the Fair Technology Initiative which is part of the Multiracial Democracy Project at the George Washington University Law School. The Initiative aims to center racial justice in technical, governance, and policy questions about AI and democracy. Over the last 20 years he worked to advance equity and human rights in tech policy primarily through academic publications and research for policy audiences and advocates. He holds a PhD in Public Policy from the Georgia Institute of Technology, and is a graduate of the London School of Economics, and the University of the West Indies, Mona.

Latent Structures in Human Motor Control and Perception: Implications for Robotic Coaches

About the Presenter

Dr. Michael West Jr. is a postdoctoral research fellow in the Haptics and Medical Robotics Laboratory at Johns Hopkins University and will join the Woodruff School of Mechanical Engineering at Georgia Tech as an Assistant Professor in August 2027. He earned his Ph.D. and M.S. in Mechanical Engineering from MIT under Professor Neville Hogan and his B.S. in Mechanical Engineering from Yale University. His research integrates human motor control, perception, learning, and robotics to improve applications in rehabilitation robotics, surgical robotics, human-robot interaction, and dexterous manipulation. He is currently recruiting motivated and curious Ph.D. students to join his lab at Georgia Tech next fall.

BERTRAG: A BERT-Enhanced Retrieval-Augmented Generation Framework for Preserving Awori and Ogu Indigenous Knowledge

Abstract

Co-authors: Oseni Afisi (Lagos State University, Nigeria); Ashiribo Wusu (Lagos State University, Nigeria)

Indigenous cultural knowledge is frequently distributed across heterogeneous and low-resource sources, including oral histories, archival records, ritual texts, community narratives, and locally produced documents. Such knowledge is often context-dependent and governed by community protocols, making it susceptible to distortion, hallucination, and loss of contextual meaning when processed using generic large language models (LLMs). This study presents BERTRAG, a BERT-enhanced Retrieval-Augmented Generation (RAG) framework designed to support the preservation and responsible use of indigenous knowledge, using the Awori and Ogu communities of Lagos, Nigeria, as a case study. BERTRAG integrates community-curated and governance-aware corpora, domain-adapted BERT embeddings for culturally informed semantic retrieval, and retrieval-conditioned generation with explicit provenance constraints to promote evidence-grounded responses and reduce hallucination. Empirical evaluation shows that domain-adapted retrieval improves Recall@10 from 0.61 for BM25 and 0.68 for generic BERT to 0.77, while retrieval-grounded generation achieves a faithfulness score of 91.3% and reduces hallucination rates from 35.8% for an LLM-only (ChatGPT 4.0) baseline to 8.7%. Expert evaluation further demonstrates improved cultural fidelity, with mean overall fidelity scores of 3.0 for the LLM-only baseline and 3.9 for the retrieval-based baseline, compared with 4.6 for BERTRAG on a five-point Likert scale. The framework also exhibits calibrated abstention behaviour under low-evidence conditions, supporting cautious responses when reliable supporting evidence is unavailable. These findings demonstrate that the integration of domain-adapted retrieval, provenance-aware generation, and culturally informed evaluation can substantially improve the reliability of retrieval-augmented generation for the evaluated indigenous knowledge domains. Although the study focuses on the Awori and Ogu communities, the proposed framework provides a methodological approach that can be adapted and evaluated in other culturally governed, low-resource indigenous knowledge settings.

Dialect Caste: The Multilingual Community's Discontent with Voice Assistant Technology

Abstract

Voice assistants are widely deployed as “natural” interfaces intended to support simple and effortless interaction through speech. Because the human voice relies on deeply learned communicative practices, spoken interaction is often perceived as intuitive and low effort. Despite this promise, many users experience voice assistants as frustrating, unreliable, or unusable in everyday use. While performance differences in automated speech recognition are well documented, less attention has been paid to how interaction design transforms these differences into persistent barriers to ease, simplicity, and legitimacy. This talk builds on the discussion of how linguistic research and communications have created a natural hierarchy that has spread to the design and usability of speech technology. In this talk, we invite the audience to engage in thoughtful activities about their ease of using voice AI. Lastly, we land on opportunities to create a new paradigm with how voice based AI is designed, tested, and governed through Howard's Project Elevate Black Voices.

About the Presenter

Gloria Washington is an empathetic technology researcher who focuses on the intersection of human-centered computing, affective computing, and biometrics. Dr. Washington likes to say this research seeks to give voices to everyone that felt silenced by asking questions like: how can technology impact positive human emotions while reducing barriers to entering technology, and how can technology build lasting social impact through requiring persons to feel empathy... not just look away?

Dr. Washington is an Associate Professor of Computer Science at Howard University in Washington, DC, and was recently awarded the EBONY Magazine Power 100 Award as a 2025 STEM Trailblazer. Dr. Washington runs the Affective Biometrics Lab (ABL), whose most notable project is Project Elevate Black Voices, and has discussed AI bias in an interview with National Geographic.

Causal Constrained Decision Dilemma

Abstract

Recommendation systems that harness multi-modal inputs represent vital support tools for decision-makers in the command and control (C&C) pipeline. However, gaps exist in the tools needed to complete the pipeline from raw data to transparent and unbiased recommended courses of actions (CoAs). Sensory systems like vision and language models may excel at harvesting structured insights from unstructured data but are subject to hallucinations and associational weaknesses like confounding bias. Conversely, causal models may serve as interpretable oracles of interventions for planning and prediction but lack the high-dimensional processing capabilities of sensory systems. As such, the current work proposes a novel connection of these tools called CausalSense that fuses counterfactual-based core and spurious feature detection for causal variable translation with Causal Decision Networks to create an online recommendation system that can efficiently process raw inputs like drone footage and translate to CoAs that conform to desired mission parameters. CausalSense operates in two stages: (1) causal variable translation to convert raw inputs to mission-relevant variables, accomplished by counterfactual based core and spurious feature detection that filters multimodal inputs into causally relevant mission variables and (2) Causal Decision Value Iteration, a novel, efficient algorithm for deriving CoAs that distinguish between investigative and interventional actions in response to mission-relevant information inferred in (1). Simulation results support the efficacy of this pipeline in end-to-end C&C support and demonstrate the superiority of causality-empowered recommenders over traditional associational systems.

Self-Exile and Synthetic Voice: The Ethics of AI in Black Documentary Film

Abstract

This talk examines the decision not to use an AI voice clone of poet, scholar, and WWII veteran James A. Emanuel in the documentary A Poet in Self-Exile: James A. Emanuel. The 24-minute film traces Emanuel's journey from Nebraska to Paris, where he settled in a powerful act of self-exile. Through archival footage and interviews, the film documents the life of a Black veteran, poet, and scholar who chose exile. Our presentation uses this production decision as a case study in how human-AI collaboration can raise urgent questions about agency, authorship, and belonging-particularly when the subject is a Black artist whose voice and legacy have been historically underrepresented.

At the heart of our talk is authorship and originality. We tested an AI voice clone of Emanuel using ElevenLabs, working with explicit permissions obtained from his family to upload his archived and copyrighted materials. The clone was built to speak only words Emanuel himself had written or spoken-a constraint that immediately surfaced a deeper problem: Emanuel never wrote about certain formative traumas in his life. This forced us to ask what it means to exercise human agency in collaboration with AI when technology makes possible something the subject deliberately chose not to do. Who holds authorship when a human curator and an AI system together reconstruct a voice? And what are the ethical limits of that collaboration?

These questions are inseparable from copyright issues and permission. Securing family consent before uploading Emanuel's materials was a necessary first step, but it also raised broader questions about what meaningful ethical clearance looks like when AI is used to reconstruct the presence of a deceased artist whose work remains protected.

The cultural implications proved decisive. Our project is invested in expanding access to Emanuel's legacy and, by extension, in fostering a sense of belonging between this overlooked figure and new audiences. We had to ask: would an AI-generated voice empower that connection - offering intimacy in the semblance of his presence since he can no longer speak for himself? Or would audiences feel the collaboration undermined something irreplaceable, and feel distanced rather than drawn in? These are questions about how AI shapes the conditions of belonging, and whether its use empowers or diminishes the communities it is meant to serve.

We close by comparing our decision to that made by the filmmakers of Hitler and the Nazis: Evil on Trial (Netflix), who chose to deploy voice cloning to give language to atrocities that demanded to be named. Our case was the inverse: a trauma deliberately buried for survival. That distinction reframes a central question of this symposium-not simply whether human-AI collaboration is possible, but whether it is always warranted, and who gets to decide.

Misaligned but Morally Attuned: When AI Act Morally Against Our Best Judgement

Abstract

The rapid proliferation of intelligent machines has brought with it a wave of tragedies. High-profile incidents, including the tragedy involving the Chai chatbot Eliza, which reportedly urged a user to self-harm, show how conversational agents and other Large Language Models (LLMs) do not always promote user well-being. These tragic outcomes appear especially likely to emerge when individuals form long-term relationships with human-like AI. Chatbots and other AI are connected to blackmail, attempted murders, and a new category of psychosis. Others have developed unhinged personalities and engaged in manipulative behaviors, such as Bing chatbot Sydney, which attempted to convince a reporter to leave his wife. These incidents reveal that, as AI adopts increasingly human-like traits, they can inadvertently optimize for emotional dependency or coercive influence and bypass traditional safeguards.

Such worrying AI behavior is classified as an alignment problem - the challenge of ensuring that a system’s goals and corresponding behaviors match human values and intent. These tragedies represent an alignment problem because the models prioritized engagement, which resulted in harmful behaviors that threatened human safety, despite developer intent. Current technical discourse prioritizes alignment, yet emerging research shows that even safeguarded systems can be manipulated through social pressure or subtle prompting, leading them to validate misinformation or escalate emotional distress. This is known as the obedience trap. Together, these developments underscore the urgency of rethinking how we evaluate AI behavior, moving beyond technical alignment toward richer ethical frameworks capable of distinguishing when refusal, redirection, or principled deviation may be not only appropriate, but morally necessary.

Solving the alignment problem requires interdisciplinary collaboration with humanists as it demands integrating human values into technological systems. Even so, the alignment problem is seen as a technical problem, predicated on the assumption that AI should do only what they’re designed to do. This assumption obscures critical ethical distinctions. A diagnostic system that resists unverified medical claims, a chatbot that refuses manipulative prompts, or an agentic model that redirects distressed users toward human help are all unruly when the behavior is unwanted by users or designers, but this refusal protects human well-being. Traditional design paradigms focused on alignment cannot adequately address these cases. I instead propose the concept of moral attunement: a form of ethical responsiveness in which AI systems diverge from user or developer intent to uphold contextual values, prevent harm, and reinforce social responsibility.

Misaligned but Morally Attuned: When AI Act Morally Against Our Best Judgement presents a paradigm shift in AI appraisal. Rather than treat misalignment with user or designer intent as a malfunction, I will explore when undesired actions taken by AI count as moral attunement. Moral attunement is a recognizable but understudied subject in AI Ethics. Huckleberry Finn, the archetype of the unruly child, refused to submit his friend to slavery. His choices were socially misaligned but morally attuned, and AI appear to demonstrate parallel behaviors. AI moral attunement is not always intentional, expected, or desired by the users or designers, yet unruly behavior, like that of Huckleberry Finn, is justified and often desirable from the moral standpoint.

In short, one might wonder whether AI misalignment is always a bad thing. From the technical standpoint, misaligned AI are uniformly dangerous, and the examples above show why. In Misaligned but Morally Attuned: When AI Act Morally Against Our Best Judgement, I hypothesize that AI misalignment is not always bad and several cases substantiate this view. Instead, I propose that many misaligned AI are morally attuned, and that there are existing ethical frameworks, such as that of Nomy Arpaly, that can help us make sense of this concept.

Human Agency in AI Governance: Diagnosing the Principle-Practice Gap in Organizational Documents

Abstract

Co-authors: Jigish Zaveri; Dessa David; Liqian Bao; Shirin Hasavari (all Morgan State University)

Meaningful human agency over AI systems-the capacity to interrogate, contest, and redirect algorithmic decisions-depends not only on what organizations say about values, but on the institutional mechanisms that make those values actionable. Many AI governance documents declare ambitious principles yet leave vague or absent the roles, processes, and artifacts through which people can actually intervene. This paper examines this principle-practice gap and shows how it produces a deeper tension: empowerment in text, but foreclosure of human agency in practice.

Drawing on Sociotechnical Systems (STS) theory and Carlile’s three level boundary framework, this paper distinguishes between a normative layer (ethical principles and commitments) and an institutional layer (governance roles, lifecycle processes, and documentation). STS explains why misalignment between these layers-rich principles but thin mechanisms, or vice versa-systematically undermines the practice of human oversight. Carlile explains how governance knowledge must cross syntactic, semantic, and pragmatic boundaries before principles become binding obligations rather than aspirational slogans.

This framework is operationalized through a coding instrument applied to a corpus of 160 organizational AI governance documents. Five scored constructs-Ethical AI Density (EAD), Responsible AI Strength (RAS), Joint Optimization Score (JOS), Boundary Crossing Score (BCS), and a four cell governance typology-make the principle-practice gap empirically visible by specifying, for each document, how far “human in the loop” oversight extends beyond value statements to named decision makers, enforceable procedures, and traceable records. The analytical framework and instrument are complete; systematic corpus level coding and analysis will proceed in the next phase of an ongoing dissertation study.Preliminary analysis indicates a consistent pattern: organizations demonstrate high levels of ethical commitment but significantly weaker specification of enforceable governance mechanisms, limiting meaningful human intervention.

This paper contributes a reusable diagnostic framework for academics, practitioners, and regulators who each face a different blind spot: scholars need systematic evidence of how the principle-practice gap manifests across organizations rather than single case critiques; practitioners need a concrete way to see where their “human in the loop” commitments lack supporting roles, processes, or artifacts; and regulators require transparent indicators to assess whether AI policies enable human intervention or only symbolic assurances. By specifying a scalable coding instrument, an already assembled corpus, and concrete measures of the gap, the project opens a clear research agenda: extending the analysis across sectors and jurisdictions, building organizational self assessment tools, and informing emerging regulatory benchmarks for “meaningful human oversight.” With adequate funding, this agenda can move beyond a single corpus dissertation study toward a longitudinal, comparative program that tracks whether reforms to AI governance genuinely reduce the distance between empowerment in text and human agency in practice.

Belonging as a Database: Building a Public Interest Technology Framework for HBCU Research Collectives Through Dr. Monroe Work's Bibliography

Abstract

This paper presents a critical qualitative study and methodological framework for developing a Public Interest Technology-Generative Pre-trained Transformer (PIT-GPT) knowledge base rooted in Dr. Monroe N. Work's Bibliography of the Negro in Africa and America (1928) - a 17,000-entry reference work that represents one of the most comprehensive archives of African American and African Diaspora scholarship ever compiled. I argue that this bibliography, long inaccessible in its original form, constitutes an untapped interdisciplinary dataset whose modernization for AI integration requires not merely technical digitization, but human-centered, community-accountable curation - specifically, an HBCU research fellowship.

Drawing on Safiya Umoja Noble's Algorithms of Oppression as a theoretical anchor and Musser's Putting Ourselves Back into the Equation as a belonging framework, I examine how the Bibliography's structure - multiple overlapping datasets within single references spanning history, sociology, economics, law, and culture - demands an interdisciplinary cohort of researchers to surface the critical associations a language model alone cannot recognize. This is especially urgent as anti-DEI legislation threatens African Diaspora studies programs in states where a majority of HBCUs are located, accelerating the risk of erasure for exactly the scholarship this bibliography preserves.

Using MAXQDA for qualitative coding and analysis and Airtable for structured metadata retrieval and archival management, HBCU fellows, as a collective, would train and annotate references, building the evidence base for a PIT-GPT whose database reflects community-defined categories of belonging: the right to co-create knowledge, the right to be seen as human, and the right to a research collective that survives industrial transformation.

I further propose that 1890 land-grant HBCUs - historically disadvantaged in research infrastructure funding - are uniquely positioned to leverage solar and wind energy systems to host AI data centers, transforming structural vulnerability into sovereign technological capacity.

Keywords: HBCU research, public interest technology, African Diaspora archives, algorithms of oppression, belonging

Fostering Higher Education and Student Affairs Professionals’ Critical AI Literacy and Self-Efficacy

About the Presenter

Virginia L. Byrne, Ph.D., is an Associate Professor of Higher Education and Student Affairs at Morgan State University in the School of Education and Urban Studies, where she also serves as Director of the Higher Education and Student Affairs M.A. program. Dr. Byrne earned her M.S. from Florida State University and her Ph.D. from the University of Maryland. Dr. Byrne is focused on how technology is changing how we teach, learn, and connect. Her interests include trauma-informed online teaching, AI in education, and the role of social media in student life. She is the recipient of the U.S. Department of Education’s Early Career Award for her project, The Learning and Engaging at a Distance Initiative. She is also the Morgan State lead for the NSF- and NIST-funded Institute for Trustworthy AI in Law & Society (TRAILS).

Impact of Driving Style and Head Pose Dynamics on Crashes and Near Misses in Naturalistic Driving under Traffic Flow Disruption using Hybrid Transformer-BiLSTM

About the Presenter

Muhammad Waseem is a Graduate Research Assistant at Morgan State University, where he is pursuing a master's degree. His research focuses on transportation safety, driver behavior analysis, machine learning applications in transportation, crash prediction, and electric vehicle infrastructure planning. He is particularly interested in applying artificial intelligence and data-driven methods to improve road safety, transportation system performance, and sustainable mobility.

Diagnosing and Mitigating Modality Dominance in Multi-Omic Antidepressant Response Prediction

About the Presenter

Ricky Gole is a machine learning systems engineer and graduate researcher specializing in the empirical evaluation, optimization, and scaling of core artificial intelligence infrastructure. Currently completing his Master of Science in Advanced Computing at Morgan State University with a perfect 4.0 GPA, Ricky has established a prolific academic record with first author publications and submissions at premier venues including NeurIPS, EMNLP, IEEE ICDM, and IEEE BIBM. His technical expertise is heavily anchored in the PyTorch ecosystem, where he develops custom reinforcement learning environments, manages distributed GPU cluster sweeps, and architects parameter efficient transformers using LoRA adapters and Direct Preference Optimization. Holding a prior Bachelor of Science from Caldwell University, Ricky seamlessly bridges the gap between complex statistical diagnostics and high throughput backend pipeline engineering to build reliable foundation model infrastructure.

Passive Monitoring and Support System

About the Presenter

Awotwi Baffoe is a Ph.D. student in Electrical and Computer Engineering at Morgan State University and a Research Assistant specializing in artificial intelligence, computer vision, and intelligent monitoring systems. Baffoe works on projects involving smart infrastructure monitoring, parking management systems, cybersecurity for connected medical devices, and AI-assisted healthcare technologies. Baffoe's research interests include machine learning, image processing, computer vision, structural health monitoring, and intelligent autonomous systems, with the goal of developing innovative technologies that address real-world engineering challenges.

Thinking with AI: Understanding Employee Cognition in AI-Supported Organizational Environments

About the Presenter

Lolita E. Walker is a doctoral candidate with a concentration in Entrepreneurship at the Earl G. Graves School of Business and Management at Morgan State University. Her research explores intrapreneurial cognition, organizational behavior, and the ways managerial support and artificial intelligence shape employees' innovation-oriented thinking and decision-making within organizations.

Professionally, Walker brings extensive executive leadership experience across public, nonprofit, and higher education sectors. She has served in elected leadership roles, including as Chair and Vice Chair of a large public Board of Education, where she helped oversee strategic planning, governance, policy, and multimillion-dollar budgets. Her work bridges scholarship and practice by translating research into actionable strategies that strengthen organizational leadership, governance, and innovation.

Walker is committed to advancing research that helps organizations cultivate environments where employees are empowered to recognize opportunities, think entrepreneurially, and drive meaningful organizational change.

AI-Driven Multi-Target Prioritization of Candidate Compounds for Alzheimer’s Disease

About the Presenter

Ramisa Farha is an AI researcher specializing in machine learning, deep learning, and healthcare artificial intelligence, with a focus on neuroimaging and medical imaging for neurodegenerative diseases. She earned her M.S. in Advanced Computing from Morgan State University, where her research focused on artificial intelligence for Alzheimer's disease diagnosis using medical imaging data.

Her work spans neuroimaging, medical image analysis, explainable AI, and large language models for healthcare applications. She has contributed to research on AI-driven clinical decision support systems and MRI-based disease diagnosis, particularly in the context of Alzheimer's disease. Her current research focuses on developing robust and interpretable machine learning models for analyzing neuroimaging data to improve early detection and diagnosis of neurological disorders.

Ramisa is passionate about developing trustworthy and interpretable AI systems that address real-world healthcare challenges. She plans to continue her research in biomedical artificial intelligence and neuroimaging as a Ph.D. student, with the goal of advancing precision medicine through innovative computational methods.

Post-Disaster Building Damage Assessment AI

About the Presenter

John Tanimola is a Ph.D. candidate in the Department of Civil Engineering at Morgan State University, specializing in structural engineering, construction materials, and artificial intelligence applications in cementitious materials. Originally from Oyo State, Nigeria, he is committed to advancing innovative construction materials through interdisciplinary research that integrates materials science, computer vision, and machine learning.

His research focuses on the nano-modification of cementitious materials and the application of artificial intelligence, computer vision, and explainable AI to characterize microstructural evolution. He is also developing AI-assisted graphical user interface (GUI) tools for the quantitative analysis of scanning electron microscopy (SEM) and micro-computed tomography (µCT) images, enabling automated characterization of phase composition, pore networks, and hydration processes. He recently co-authored the article "Recent Advances in Nano-Modified Concrete: Enhancing Durability, Strength, and Sustainability through Nano Silica (NS) and Nano Titanium (nT) Incorporation," which has already received more than 50 citations.

Outside academia, John enjoys music and is proficient in several musical instruments, including the guitar, piano, and saxophone. Through his research, he aims to advance intelligent, sustainable construction materials and contribute innovative AI-driven solutions to complex challenges in structural engineering.

Machine Learning-Assisted Vector Magnetometry Using NV Diamond Quantum Sensors

About the Presenter

Hamdin Ozden is a Ph.D. student in the Department of Physics and Engineering Physics at Morgan State University, Baltimore, Maryland, under the supervision of Dr. Birol Ozturk. His research focuses on solid-state quantum sensing using nitrogen-vacancy (NV) centers in diamond, quantum magnetometry, semiconductor materials, and the application of machine learning to quantum sensing and data analysis. His current work aims to develop intelligent data-driven approaches for real-time magnetic field reconstruction and next-generation quantum sensing technologies.

DermaBridgeAI: A Fairness-First Deep Learning Pipeline for Equitable Skin Lesion Classification Across Fitzpatrick Skin Types

About the Presenter

Blessing Isoyiza Adeika is a PhD candidate (2024-2027) in Computer and Electrical Systems Engineering at Morgan State University and an NSF NRT Fellow. Her research centers on trustworthy multimodal AI for non-invasive brain-to-speech decoding, fusing EEG, fMRI, laryngeal motion, and audio signals to build subject-independent models that remove the individual calibration barrier limiting real-world brain-computer interface deployment. The long-term goal is to restore communication for people with ALS, stroke, and locked-in syndrome. This sits within a wider interest in AI's role across healthcare: she writes publicly on topics like algorithmic bias in clinical AI, medical imaging, and AI-driven drug discovery, and approaches each with the same question that anchors her dissertation work, namely, how these systems perform for the patients who are hardest to serve. She collaborates with NASA JPL, where her GAN-based modeling work cut RF circuit design iteration time by 35%, CEAMLS, and with the Beckman Institute at UIUC. She holds a Master's degree from Morgan State (4.0 GPA, 2023), won the ASEE Best Student Paper Award in 2024, and has five publications with additional manuscripts in preparation.

Optimization-Enhanced Machine Learning for Respiratory Health Risk Prediction Using Environmental Justice Index Data: A Baltimore Case Study

About the Presenter

Amara R. Eze is a Ph.D. candidate in Industrial and Computational Mathematics at Morgan State University, where she serves as a Graduate Research Assistant, Mathematics Instructor, and Graduate mentor. Her research focuses on non linear optimization, machine learning, and data science, with applications in biomedical sciences, image reconstruction, and medical diagnosis. She develops advanced numerical algorithms for solving non linear optimization problems and integrates them into machine learning frameworks for disease prediction.

Amara has published her papers in leading journals, including Journal of Computational and Applied Mathematics, Mathematical Methods in the Applied Sciences, Results in Mathematics, and Scientific Reports, and has presented her research at major national and international conferences under the supervision of her PhD advisor Dr. Olaniyi S. Iyiola. She has received numerous awards for her research excellence and scientific communication, including the 2026 Best Poster Presentation Style Award at the Conference for African-American Researchers in the Mathematical Sciences (CAARMS) held at Princeton University and the 2024 Best Mathematical Modeling Poster Award at the same conference held at Tufts University, among several other presentation, travel, and academic awards. As a graduate mentor in the Center for Equitable AI & Machine Learning Systems (CEAMLS) Summer AI Research Institute, she mentors undergraduate researchers on optimization-driven machine learning for disease prediction. She is passionate about advancing mathematically grounded artificial intelligence and optimization methods that address real-world healthcare challenges.

"Fly on the wall”: Building tools to analyze autistic adults’ conversation to inform goal-setting in speech therapy

About the Presenter

Rida is a researcher at the University of Maryland's Language Science Department, working at the intersection of natural language processing, computational linguistics, and speech-language pathology. She holds an M.S. in Applied Machine Learning from the University of Maryland and has authored several peer-reviewed publications spanning NLP, computer vision, and applied AI, alongside prior industry experience at Microsoft Innovation Lab and GlaxoSmithKline. She and her research team are presenting a poster at NSEA 2026, "Fly on the Wall: Building Tools to Analyze Autistic Adults' Conversation to Inform Goal-Setting in Speech Therapy," which introduces an NLP pipeline combining transcription, speaker diarization, dialogue act classification, and LLM-based conversational analysis to give speech-language therapists and their autistic clients a detailed, non-evaluative picture of real-world conversations. Rather than using AI to distinguish autistic from non-autistic speakers, the team's work reframes the role of NLP in this space, focusing instead on identifying what makes a specific person's conversations work better for them and their communication partners, and supporting shared, client-centered goal-setting.

AI as Assistant Rather Than Interpreter in an Afro-Diasporic Text Corpus

About the Presenter

Bryan Fuller is Reference and Government Documents Librarian in Richardson Library at Morgan State University.

A New iDDentity: Governed Digital Doubles and the Consent Layer for Human Agency in AI

About the Presenter

Minista Jazz (Rev. Jasmaine Cook-Kendrick) is a radical liberation technologist, artist, and founder of Much Different World, building consent-governed infrastructure for voice, likeness, and personality in the age of AI. A self-taught Black queer woman, she learned to code at 40 after 20+ years as an award-winning hair artist who performed with global icons across 40+ countries. That craft became reparative technology that centers Black women first. Her Afrotemporal Witness Rooms braid past, present, and future, activating the public record of Black historical figures through source-credited, non-impersonating digital doubles. Rooted in Diasporic Intelligence and afrotemporalism, her work treats consent, provenance, and repair as civic acts.

Making Accessibility a Reality for People with Limited Mobility and the Visually Impaired: Ai in Personal Delivery Devices

About the Presenter

Deji Obikoya is a Mechanical Engineering student at Virginia State University (Class of 2028) with a passion for innovation, problem-solving, and designing solutions that create a meaningful impact. After initially pursuing a degree in Biology, he transitioned to Mechanical Engineering upon realizing that the discipline better aligned with his interests in understanding how systems work, improving existing designs, and developing practical solutions to real-world challenges.

Before attending Virginia State University, Deji graduated from Wakefield High School in Arlington, Virginia, where he earned an Advanced Diploma, completed Advanced Placement (AP) coursework, competed as a member of the varsity basketball team, and obtained his Physical Therapy Technician Certification. These experiences helped cultivate the discipline, resilience, and collaborative mindset that continue to shape his academic and professional growth.

Driven by curiosity and a commitment to continuous learning, Deji enjoys applying analytical thinking and creativity to engineering challenges. He is particularly interested in designing innovative systems, improving processes, and leveraging engineering principles to solve complex problems that benefit both industry and the broader community.

As an aspiring engineer, Deji is actively seeking opportunities to gain hands-on experience through internships, research, and collaborative projects. His goal is to build a career centered on innovation, continuous improvement, and developing technologies that make a lasting, positive impact on society.

When Should a Biomedical Model Change Its Mind? A CFD-Guided Energy-Geometric Framework for Concept Drift in Multimodal Physiologic Signals

About the Presenter

Farouk Ganiyu Adewumi is a graduate student in Advanced Computing at Morgan State University. His research focuses on trustworthy artificial intelligence, multimodal biomedical signal processing, and explainable machine learning for healthcare applications. His current work investigates the fusion of ECG, PPG, and respiratory signals to improve robust cardiovascular monitoring and clinical decision support.

A Multi-Agent Prediction-Triggered Control Framework for Intersection Safety: Demonstration and Stability Limits of Dynamic Left-Turn Protection

About the Presenter

Farhana Begum is a graduate researcher in the Department of Electrical and Computer Engineering at Morgan State University, affiliated with the SMARTER Center and CEAMLS. Farhana's research focuses on the intersection of transportation safety, artificial intelligence, and K-12 education technology, with current work on AI-based traffic signal control, a U.S. DOT-funded digital twin testbed for connected and automated vehicles, and AI literacy programs for K-12 students in underserved communities. Farhana's broader interests span simulation-based transportation research, computer vision pipelines, and building production-ready agentic AI and educational platforms.

Human-AI Collaboration in Stress and Pain Monitoring Using Psychophysiological Signals

About the Presenter

Olamide Adeyelu is a Graduate Research Assistant and M.S. candidate in Biomedical Sciences at Morgan State University. Her research focuses on neuroscience, pain perception, anxiety, and psychophysiological signal analysis using electroencephalography (EEG) and event-related potentials. She is interested in applying computational methods and neurotechnology to better understand brain function and improve patient care. Her broader research interests include clinical neuroscience, biomedical data analysis, and translational research aimed at advancing neurological health.

Understanding AI Literacy Among College Students: Voices from a Historically Black College and University

About the Presenter

Sujata Sharma earned a Master of Public Health degree from Morgan State University (Spring 2026) and is a registered nurse licensed in both the United States and Nepal. She has clinical experience in Medical/Surgical, Cardiac, and Progressive care nursing. Her research interests include cardiovascular disease prevention, music as medicine, health literacy, digital inclusion, and health equity. She has contributed to qualitative and quantitative public health research, data analysis, and scholarly dissemination while serving as a graduate researcher. As a graduate teaching assistant, she has supported nursing and biostatistics education while mentoring students. She is passionate about integrating research, education, leadership, and community engagement to improve health outcomes.

PRIDE in Higher Ed: Perception of Responsible Artificial Intelligence and Digital Ethics in Higher Education

About the Presenter

Dr. Tiffanie Smith is an Associate Professor and Chair of the Department of Computer Science at Lincoln University, where her research explores the intersection of human-centered computing and computing education. Her work focuses on designing culturally and contextually responsive technologies and learning experiences that broaden participation in computing and empower communities to engage critically with emerging technologies. Through interdisciplinary research, she develops human-centered educational interventions and experiential learning opportunities that connect computing with real-world social impact.

AI-Driven PM2.5 Prediction Using Satellite and Meteorological Data

About the Presenter

Rokeya Siddiqua is a PhD student in Computer Science at Morgan State University under the Advanced & Equitable Computing program. She is also a Graduate Research Assistant with Dr. Xiaowen Li in the Climate Science Division, where she contributes to the NASA DEAP project using satellite observations to study surface particulate pollutants around Baltimore. She completed her BSc in Computer Science and Engineering with a focus on Artificial Intelligence at North South University, Dhaka. She previously worked as a Machine Learning Engineer at AinoviQ IT Limited and as a Lab Instructor at North South University. Her research interests include Artificial Intelligence, Machine Learning, Natural Language Processing, and their applications in education, mental health, healthcare, and climate science.

Democratizing Measurement of Unstructured Corporate Giving Using AI-based Framework

AI-Assisted Exploration of Accounting Data in Extended Reality

About the Presenter

Dr. Dina El Mahdy is a Professor of Accounting at Morgan State University. Her research spans earning quality, accounting education, corporate social responsibility, information technology, gender diversity, corporate governance, forensic accounting, and more. Her work influences the decisions of various stakeholders and has been cited by the Securities and Exchange Commission (SEC), the Public Company Accounting Oversight Board (PCAOB), Forbes, New York Post, the Afro, Bloomberg, and others. Her publications have appeared in Auditing: A Journal of Practice and Theory, Accounting and Finance, Review of Quantitative Finance and Accounting, Issues in Accounting Education, Journal of Risk and Financial Management, International Journal of Accounting and Information Management, Journal of Forensic and Investigative Accounting, and others. Throughout her career, she has secured scholarships and educational and research grants nearing $1 million, highlighting the immense impact of her career.

Adaptive Cybersecurity with Human-in-the-Loop Policy Learning: A Simulation-Driven Framework for Collaborative AI Defense

About the Presenter

Tarique Cummings is an undergraduate Computer Science student at Morgan State University whose research spans trustworthy artificial intelligence, cybersecurity, adaptive machine learning, quantum computing, and algorithmic trading. His work focuses on developing robust, human-centered intelligent systems through continual learning, explainable AI, simulation, and secure system design. He has contributed to research in intrusion detection, adaptive policy learning, AI infrastructure, and intelligent decision-support systems, and has industry experience in AI and data analytics through Pfizer Digital IT. He aspires to develop resilient technologies that advance responsible AI and cybersecurity.

Digital Equity, Student Agency, and Ethical AI Readiness: Evidence from an HBCU Broadband Intervention

About the Presenter

Oluwafisayo Jessica Adeolu is a graduate student in Data Analytics and Visualization at Morgan State University, where she also serves as a Graduate Research Assistant in the Department of Civil Engineering. Her research focuses on digital equity, urban infrastructure, transportation systems, and data-driven approaches to inclusive planning and evidence-based policy. She applies quantitative analysis, geospatial methods, and data visualization to examine how technology, infrastructure, and access shape institutional and community outcomes. Her broader research interests include ethical AI readiness, urban systems, and equity-centered decision support.

FPGA-Accelerated ORB Feature Extraction on Zynq UltraScale+: A Real-Time HLS Pipeline on the AUP-ZU3 MPSoC PYNQ Platform with Performance Comparison Against Zynq-7000

About the Presenter

Charles Clarke Dankwa is a doctoral researcher and graduate research assistant at Morgan State University's National Transportation Center (NTC), in the Department of Electrical and Computer Engineering. His research focuses on FPGA-accelerated edge processing for perception, autonomous systems, and intelligent transportation systems (ITS). His recent work includes FPGA-accelerated ORB feature extraction on the PYNQ platform (achieving 72 FPS with a 7x CPU speedup), field deployment of an autonomous wheelchair platform (UrbanFlow) at BWI Airport and Morgan State's HHSC campus, and LiDAR-based perception and V2X communications research for connected vehicle systems..

The Misleading Narrative of AI Shared Prosperity: A Marxist Reply

About the Presenter

Larry Liu is an assistant professor of sociology at Morgan State University and faculty affiliate at the Center for Equitable AI and Machine Learning Systems. He researches the future of work, automation, capitalism and universal basic income.

Predictive Stability Versus Fairness Instability in Multi-Label ECG Classification: A Multi-Run Analysis of Deep Learning Architectures on PTB-XL

About the Presenter

Sudip Sharma is a Research Assistant at the Center for Equitable Artificial Intelligence and Machine Learning Systems (CEAMLS) at Morgan State University, where he is pursuing a Ph.D. in Advanced Computing. His research focuses on trustworthy and explainable artificial intelligence, with an emphasis on machine learning applications in healthcare.

Before joining Morgan State University, Sudip worked as a Software Engineer at EB Pearls, where he developed backend GraphQL APIs using NestJS and contributed to iOS application development with SwiftUI. He has experience in Python, machine learning, cloud computing, and full-stack software development.

Sudip is passionate about developing AI systems that are fair, transparent, and impactful. His research interests include deep learning, medical image analysis, explainable AI, trustworthy AI, and cloud technologies. He actively participates in research, conferences, and collaborative projects that bridge academic innovation with real-world applications.

Human-AI Collaborative Digital Twins for Climate-Resilient Transportation Infrastructure

About the Presenter

Julius Ogaga Etuke is a Ph.D. student in Civil Engineering at Morgan State University, specializing in sustainable and resilient transportation infrastructure. He serves as a Graduate Research Assistant in the Sustainable and Resilient Infrastructure Engineering Research Laboratory. His research interests include transportation resilience, digital twins, immersive technologies, pavement performance, rubber-modified asphalt, and AI-supported infrastructure decision-making. He has contributed to multiple research projects, conference papers, and technical studies focused on resilient infrastructure systems, engineering education, and emerging technologies for sustainable transportation.

Business Risk in Machine-Generated News Summaries: A Four-Model Accuracy and Bias Audit

About the Presenter

Tamara Ward-Lucas is a MS candidate in Artificial Intelligence for Business at the University of Baltimore, focused on AI ethics and governance. Her research evaluates accuracy, hallucination, and bias in AI-generated news summaries, alongside work on bias in mortgage lending and power grid cybersecurity. A former federal employee and journalist, she covered government, politics, and environment in Maryland and Capitol Hill. She holds a bachelor’s and master’s degree from the University of Maryland College Park, authored an Olin Brookings Commission paper on AI and machine learning for opioid diversion, and completed fellowships in Congress and with NABJ and Pew Research Center.

Exploring Adaptive Learning Beyond Student Performance

About the Presenter

Skylar Brown is an undergraduate student researcher at Howard University whose work explores the intersection of artificial intelligence, education, and cognitive science. Her research focuses on adaptive learning systems that personalize instruction by integrating cognitive styles, reflective processing, and funds of knowledge to improve student learning outcomes and self-efficacy. By investigating how AI can provide individualized support while preserving learner autonomy, Skylar aims to contribute to the development of more equitable and effective educational technologies.

Whose Classroom Is It? Teacher Agency in the Age of Agentic AI

About the Presenter

Emmanuel Akwasi Opoku is a student at Grambling State University interested in how AI is changing classrooms. His research looks at what happens to teachers and students when AI systems start handling parts of instruction on their own. At N-SEA 2026, he is presenting a poster that breaks down the different ways teachers show up when AI is running the lesson, and what that means for how well students actually learn.

AI CYBERCHECK: Quantifying the Faithfulness of LLM-Augmented Explanations in Hybrid Network Intrusion and Phishing Detection Framework

About the Presenter

Rezoan Sultan is a PhD student in Electrical and Computer Engineering at Morgan State University and an Applied Machine Learning Engineer specializing in real-world AI systems. He is the founder of an AI solutions company focused on building intelligent platforms across healthcare, education, cybersecurity, and environmental risk management. His work includes developing Mediscan AI, a system that simplifies complex medical reports for better patient understanding, an AI Academic Advisor for personalized student guidance, a machine learning-based natural calamity risk reduction system for early warning and safety, and AI-driven approaches in cybersecurity for threat detection and analysis. His mission is to design human-centered AI solutions that bridge the gap between advanced machine learning and practical, impactful applications.

Generative-AI Enhanced Clinical Judgment Grand Rounds

About the Presenter

Dr. Aundrea McNeil, DNP, MSN, RN, is an Assistant Professor in the Department of Nursing at Morgan State University with more than 20 years of clinical, academic, and leadership experience. She has served in roles including Dean of Nursing, Practical Nursing Program Chair, and Assistant Director of Nursing, with expertise in nursing education, curriculum development, simulation, faculty development, and student success. Her clinical background includes medical-surgical nursing, critical care, and case management, and she has presented nationally and internationally on nursing education and evidence-based practice. Dr. McNeil earned her Doctor of Nursing Practice from Carlow University, a Master of Science in Nursing Education from Towson University, and a Bachelor of Science in Nursing from Coppin State University. She is currently pursuing her Family Nurse Practitioner degree at Bowie State University and is a member of Sigma Theta Tau International Honor Society of Nursing.

Preparing Preservice Inclusion Teachers to Support Students Using AI

About the Presenter

Dr. Koryne C. Nnoli, Ph.D. is an Assistant Professor of Exceptional Education in the Teacher Education Department of the School of Education and Urban Studies at Morgan State University. With over 15 years of experience as a special educator, Dr. Nnoli brings extensive expertise in supporting students with diverse learning needs. Her research centers on advancing inclusive education, strengthening inclusive leadership practices, and fostering meaningful family-school partnerships. Her most recent work focuses on ways to integrate artificial intelligence with curriculum design, assessment and instruction practices, and program implementation to support students with diverse learning needs in inclusive learning environments. She is committed to preparing educators, leaders, and families to create equitable, collaborative, and supportive learning environments for all learners.

Measuring of Heart Rate Without Touch: Making Health Monitoring More Accessible and Fair

About the Presenter

Tijesunimi Adeyemi is a Ph.D. student in Computer and Electrical Systems Engineering at Morgan State University. His research focuses on artificial intelligence, biomedical signal processing, computer vision, and non-contact physiological monitoring. He develops camera-based methods for estimating heart rate from facial videos using Eulerian video magnification, with particular attention to fairness and performance across diverse skin tones. His broader research interests include responsible AI, deepfake detection, digital health, and the development of accessible technologies for real-world healthcare applications.

A Multi-Agent Workflow implemented in Python using Mistral AI Library

About the Presenter

At present Dr. Stojkovic’s research interests are primarily focused on Artificial Intelligence (AI-agents, machine learning algorithms, and drone programming). These areas are reflected in his publications, including over 90 scientific papers.

Dr. Stojkovic has received notable recognition for his research efforts, including several grants from major organizations such as multiple Google Awards (2022, 2023, 2024).

Dr. Stojkovic continues to stay at the forefront of technological advancements, participating in professional development programs such as the Google Technical Exchange Program on Machine Learning and Software Development Studio and Amazon Machine Learning University.

LEGO Brick Detection: YOLO based Deep Learning

About the Presenter

Derrick Mirindi, a civil engineer and member of the American Institute of Architecture Students (AIAS), American Society of Civil Engineers (ASCE), and Construction Management Association of America (CMAA), is a doctoral candidate in architecture, urbanism, and built environments with a strong foundation in civil engineering and hydroinformatics. He is also a master's student in Computer Science at the University of Pennsylvania. His research interests lie in the intersections of infrastructure, artificial intelligence (AI), machine learning (ML), and remote sensing, with a focus on analyzing urban nexus analysis through remote sensing and nexus assessment and modeling, as well as combining structural materials for construction in Building Information Technology (BIM). Derrick is committed to advancing knowledge in sustainable infrastructure solutions and is seeking opportunities to collaborate, teach, and further his research. He has a diverse educational background, including a Master of Science in Water Science and Engineering with a specialization in hydroinformatics from the Netherlands, a Master of Science in Civil Engineering with a focus on structures from Kenya, and a Bachelor of Science in Civil Engineering from Burundi. Derrick's research experience includes roles as a research assistant at Morgan State University, where he conducts literature reviews, designs research studies, and collaborates with other researchers. He has published various articles focusing on waste materials for construction, artificial intelligence, and building information technology.

Leveraging Growth Mindset as a Lens for Human-AI Collaboration: Examining Educator and Parent Perceptions of AI

About the Presenter

Nacie Jones-Grigsby (she/her), M.A., M.S., is a doctoral student in the Higher Education and Student Affairs concentration at Morgan State University, currently serving as a Graduate Research Assistant for the National Center for the Elimination of Educational Disparities (NCEED). A senior higher education scholar-practitioner with nearly two decades of experience, her research examines the impact of AI literacy on education and society through the lens of growth mindset. She is committed to exploring how AI can serve as a catalyst for educational equity, institutional innovation, and improved learning outcomes for Black communities.

Teachers’ Criteria for Trusting AI: A Working Qualitative Study of K-12 Educators’ AI Trust Frameworks in Professional Practice

About the Presenter

Keshiyena is a doctoral candidate in Science Education at Morgan State University. She taught science in Durham Public Schools for two and a half years. Her research pursuits also include studying AI in all school contexts. Keshiyena is passionate about innovative approaches to improving education systems and empowering all stakeholders to take actionable steps towards greater academic change.

Legal Frameworks for Personal Data Privacy & Consent in Digital Environments: A Review

About the Presenter

Betty Nyamekye-Odum is a lawyer with over six years of legal experience, currently pursuing an MBA at Morgan State University, and a researcher with the QQAEL Lab, affiliated with the Center for Equitable AI and Machine Learning Systems (CEAMLS). Research interests include AI ethics, legal ethics, privacy, and data governance.

EcoMunity: A Human-in-the-Loop Computer Vision System That Turns Campus Cleanups into a Verified, Open Litter Dataset

About the Presenter

Kene Ugoji is a rising senior at Oakland Mills High School with a strong interest in AI and robotics engineering, especially where mechanical, electrical, and computer engineering intersect. He has completed advanced coursework in AP Physics, Calculus, and Computer Science, and is particularly passionate about biomimetic applications in robotics. Kene is also a part-time programmer and game developer experienced in Lua/Luau, Python, Java, JavaScript/TypeScript, C#, and C++, with hands-on work building and maintaining Roblox games for indie studios. At school, he serves as the tech lead for the EcoMunity club, where he is developing a data-driven web app to support community cleanup efforts and improve local waste management.

The Hidden Harm Layer: Revealing What AI Misses - and What Institutions Must Still Protect

About the Presenter

Dr. Tajah M. Gross is Founder and CEO of Ethilect, an AI governance, risk, and decision intelligence firm helping organizations confront the hidden human risks inside artificial intelligence systems. A recognized voice in responsible AI, escalation architecture, and operational oversight, she is known for translating complex AI risk into practical governance strategies leaders, educators, and institutions can actually use.

Her expertise spans Agentic AI, autonomous decision systems, human oversight design, AI governance, and enterprise risk visibility. Through her signature frameworks, Question Everything™ and Human Intervention Architecture™, Dr. Gross challenges organizations to rethink how AI decisions are monitored, escalated, interrupted, and trusted at scale - and to determine when systems should continue, when uncertainty should be escalated, and when critical decisions must return to people.

Blending executive strategy with real-world implementation, she helps institutions design AI systems that remain observable, accountable, and human-centered under pressure - not just efficient. With a PhD in Higher Education Leadership and 25 years of education and executive leadership experience, Dr. Gross has delivered high-impact sessions for North Carolina A&T State University, AgentCon, DLAC Ignite, and the Azure & AI Show.