The integration of AI into the pedagogical landscape of public health represents one of the most immediate and profound shifts in higher education history. As schools and programs navigate this transition, the overarching goal must be to advance training opportunities that are not only innovative but are also grounded in the responsible and ethical application of technology. This focus area addresses the dual imperative of utilizing AI to enhance student engagement and learning outcomes while establishing the necessary guardrails to protect academic integrity and equity.

HUMAN ELEMENT, ETHICS, AND RESPONSIBILITY IN TEACHING

At the core of the task force’s recommendations is the principle that AI must augment, but never replace, human judgment, expertise, and compassion. Public health is a field defined by human-to-human relationships, and this fundamental truth must shape the classroom experience.

Faculty members bear primary responsibility for fostering an ethical environment for AI adoption, prioritizing safety, transparency, accountability, and equity across all technological applications. A critical component of this responsibility is “privacy risk transparency.” Instructors should be encouraged to remind students to exercise caution when interacting with consumer AI applications to avoid unintentionally sharing intellectual property, confidential research data, or protected personal information. Ultimately, the onus for the quality and integrity of AI-generated output rests with the individual user, but this concept that must be reinforced through institutional policy and classroom instruction.

However, exercising individual responsibility, while critical, is not sufficient to address the ethical, socioeconomic and health challenges posed by AI. Consumer-facing genAI chatbots are now associated with a demonstrated set of behavioral health risks (sometimes described as “AI-associated psychosis“) that public health curricula must engage with rather than abstract away. The American Psychological Association has issued a health advisory⁵⁴ on these risks, and a systematic review of the literature has surfaced other challenges.⁵⁵ Beyond the demonstrated mental health challenges, instructors must also acknowledge students’ concerns about potential job displacement, community impact and environmental harm.

Given these considerations, any classroom recommendation that students use AI must be paired with explicit instruction on these unintended harms, protective practices for personal use and the public health framing of AI as a determinant of health. Where use of AI is required to complete assignments, instructors should provide opt-out paths for students who decline based on safety, environmental or ethical grounds.

FACULTY PREPAREDNESS AND INSTRUCTIONAL SUPPORT

The rapid evolution of AI has created a significant “bottleneck” in institutional capacity: faculty preparedness. Many educators find themselves teaching students to use tools that are changing faster than traditional professional development cycles can keep pace. To bridge this gap, the task force recommends that ASPPH and its member institutions invest heavily in faculty resources and training.

Instructional support should move beyond basic tool awareness to help faculty conceptualize AI’s potential for having a supporting role in the teaching process. For example, AI-driven feedback assistants can help faculty provide rubric-aligned, constructive feedback to students more efficiently, allowing them to focus on deeper mentorship and complex problem-solving. This “partnered” approach to instruction allows faculty to focus on high-level critical thinking, an area where human judgment remains of outsized importance.

STUDENT PREPAREDNESS AND AI LITERACY

For students, exposure to AI tool use in a higher-education setting must be paired with critical reflection and the development of a robust “technical literacy.” This expands beyond simply familiarity with using a chatbot; it requires gaining basic understanding of AI frameworks such as retrieval-augmented generation (RAG), understanding frontier model performance on relevant public health benchmarks, developing effective “prompt engineering” techniques and developing the ability to formulate questions that elicit high-quality, relevant results from AI systems. Students must also be well-versed in human-in-the-loop (HITL) principles, operational models and HITL limitations⁵⁶ to ensure substantive oversight and accountability of automated public health workflows.

AI literacy training should be intentionally integrated into existing public health curricula. This training must include: 

• Effective Prompting and Critical Evaluation: Teaching students to move from simple analysis to sophisticated inquiry. Developing sound judgment, intuition and ethics grounded in public health domain knowledge and experience will be prerequisites for this critical reflection.
• Bias and Limitation Recognition: Helping students reflect on the potential for AI to produce inaccurate outputs (hallucinations) or perpetuate systemic inequities present in its training data. Students will need to explore AI failure modalities in public health settings and statistical reasoning under uncertainty.
• Responsible Acknowledgment: Training students on responsible citation and attribution of AI-originated content within the framework of university and program policies. 
• Human-Centered Design: Students must be led to cultivate empathy, perform iterative testing and employ multidisciplinary collaboration that prioritizes the needs, behaviors and experiences of humans in the context of DDoH.

INSTRUCTIONAL INNOVATION AND APPLIED LEARNING 

The Framing the Future: Transformative Approaches to Teaching and Learning Report provides a roadmap for instructional innovation and applied learning, and institutions should continue efforts to implement the aspirational recommendations and ask the reflective questions outlined in the report. Building off of this previous work, the task force has identified several “promising practices” that use AI to transform the learning experience from a one-size-fits-all model into a custom, adaptive environment. 

• Personalized and Adaptive Learning Paths. Adaptive learning systems, such as Pearson+ and McGraw Hill Connect, use AI to monitor individual student performance and deliver tailored content, real-time feedback, and targeted outreach. These platforms adjust for pacing and complexity, supporting the mastery of foundational public health concepts while accommodating various learning styles. 
• Simulation and Gamification: Simulation technology represents a powerful bridge between the classroom and real-world practice. These tools allow students to practice health field tasks — such as outbreak response decision-making, risk communication and community engagement — in risk-free, computer-based environments. 
• Case-Based Learning: AI-driven simulations can be adapted to mirror specific challenges faced in practice settings, such as navigating community resistance or addressing the “missing voices” in public health data. 

RISKS, GUARDRAILS, AND ACADEMIC INTEGRITY 

The Investigating the Use of Generative AI Policies among ASPPH Member Schools and Programs of Public Health project revealed that 83.3% of current member institution policies focus on pedagogical guidance and classroom use. However, because the promise of using AI to complete coursework is often “irresistible,” institutions must fundamentally rethink their deliverables.^57,58

Tiered Usage Policies: The Task Force recommends that institutions provide clear “strata” of AI use for instructors to adopt in their syllabi. These typically include: 

1. Complete Prohibition: No AI use allowed for any part of the assessment. 
2. Restricted Use: AI is permitted for specific tasks (e.g., brainstorming, outlining) with explicit citation. 
3. Unrestricted Use: AI is encouraged as a collaborator throughout the process, provided its use is disclosed. 

Non-Outsourceable Assessments: To protect academic integrity, educators should prioritize evaluation methods that cannot be easily outsourced to AI, focusing on embodied, relational and place-based practice. These include: 

• Oral Communications: Presentations and podcasts that require real-time synthesis and defense of ideas. Require a live, unscripted Q&A with the intended audience (community members, practitioner panel) or peers.
• In-Person Performance: Handwritten assignments in class (e.g., “blue books”), oral debrief, role-play and interactive discussions. 
• Scaffolded Projects: Breaking large assignments into stages that require human review and the documentation of human thought processes. Assess process artifacts alongside products (commented code, version history, annotated drafts, reflective memos).
• Localized Context: Tie deliverables to a specific, current, local and verifiable context (a real state/local bill, a named community partner, a newly published study, this week’s misinformation). LLMs typically miss the mark because their training data lacks specificity, local dynamics or new information.
• Reflective Journaling: Reflective entries tied to specific course events or experiential learning (a guest lecture, a case discussion, a messy dataset exploration, community interviews, a neighborhood tour). Hard to outsource because they reference a shared classroom or observational context.

STRATEGIC DIRECTIONS FOR ASPPH AND MEMBER INSTITUTIONS 

To ensure the long-term success of AI integration in teaching and learning, the Task Force proposes several strategic directions: 

• Invest in Emerging Technology: Prioritize funding for AI-powered simulations and gamified tools that personalize instruction. 
• Establish Feedback Loops: Create formal pathways for practice partners to share their evolving AI use cases, ensuring that classroom training remains aligned with real-world workforce demand. 
• Develop Continuous Education: Offer specialized AI training for current public health practitioners and alumni to help the existing workforce keep pace with technological change. 

The gap between the theoretical promise of artificial intelligence and its practical adoption remains a primary challenge for the public health sector. While 84% of healthcare organizations are already actively using AI and machine learning technologies, public health organizations lag significantly behind. This disconnect stems not from a lack of potential, but from fundamental tensions between technical capabilities and organizational realities, between data outputs and community trust, and between established algorithms and the “missing voices” often absent from the data. 

Artificial intelligence also has broader implications for the social determinants of health. While AI offers new tools to identify risk patterns and allocate resources more effectively, it may also introduce unintended consequences, including environmental costs associated with computing infrastructure, shifts in labor markets, unequal access to technological resources, and the risk of misuse by malicious actors or poorly designed systems. ASPPH recognizes that these dynamics require ongoing evaluation to ensure that AI adoption supports, rather than undermines, health equity and community well-being. 

HUMAN-CENTERED AI IN PRACTICE AND RESEARCH 

Public health is fundamentally built on human-to-human relationships, and that is a reality that AI cannot replace. The Task Force advocates for a “human-centered” approach in which AI is viewed as a supportive tool for decision-makers rather than a replacement for human judgment. This perspective is essential because AI will never build community relationships; instead, it must enhance professionals’ ability to navigate complex human systems where trust and community voices matter more than sheer algorithmic accuracy. 

CURRENT AND EMERGING USE CASES 

AI technology is beginning to streamline several key public health functions, shifting from traditional data analysis toward more interactive inquiry and exploration: 

• Disease Surveillance and Data Analytics: AI can contend with expanding data volumes to identify patterns and predict epidemic outbreaks more rapidly than purely human-driven systems. 
• Operational Efficiency: GenAI can reduce administrative burdens for community health workers by optimizing supply chain efficiencies and automating routine coding and documentation tasks. Agentic AI is well-suited to streamline the operational scaffolding of public health research — participant outreach, eligibility screening, scheduling, two-way communication, study monitoring and documentation. These are the workflows that most often slow down public health studies, particularly community-engaged and pediatric research. 
• Research Translation and Participant-Based Studies: AI tools are increasingly used to accelerate discovery by improving the speed and quality of researching and summarizing large amounts of written evidence. Governance for this category should be workflow-aware, use-case-specific (different oversight tiers for participant-facing communication versus analytic models) and should require structured pilot testing with research teams in real settings rather than relying solely on abstract committee review.

PREPAREDNESS OF THE PUBLIC HEALTH WORKFORCE 

There is a profound “lack of knowledge” within public health agencies regarding current AI capabilities. Preparation for an AI-enabled workforce requires moving beyond technical implementation toward developing a critical mindset that sees AI as a tool for exploration rather than an ultimate solution. 

TRUST, COMMUNITY ENGAGEMENT, AND UNINTENDED CONSEQUENCES 

The “flood prediction case” from University City serves as a pivotal cautionary tale for the field. In this project, an AI model was developed with a high likelihood of predicting future flood events in areas of University City, a suburb of Missouri, which have never been flooded in the past. However, the technical success meant little because the community had concerns, including identifying some parts of the community as flood-prone, which could plummet property values, lead insurance companies to decline coverage, and possibly create community stigma. 

This case demonstrates that technical accuracy is insufficient without community buy-in. Trust building requires early and continuous engagement, transparency about technological limitations, and addressing potential unintended consequences upfront. 

EQUITY, BIAS, AND MISSING VOICES IN DATA 

AI “doesn’t know what it doesn’t know”. Because AI systems pull from existing data, they often reflect and perpetuate the structural inequities and biases of the world in which that data was collected. 

• The Equity Imperative: The communities most in need of public health interventions are often the ones least represented in the data AI uses. 
• Actionable Steps: We must teach professionals to conduct a “missing voices analysis” to identify whose data is missing and we must actively seek unrepresented perspectives during the design of AI systems. 

OPEN SCIENCE, REPRODUCIBILITY, AND RESEARCH INTEGRITY 

Public health must lead in defining “what good looks like” for AI-driven research. Encouraging open, reproducible science is essential to foster trust among the research community and the public. 

• Transparency Standards: Sharing code, datasets, and evaluation metrics is non-negotiable for responsible discovery. Where proprietary datasets or federal and state regulations (e.g., HIPAA, FERPA, GDPR) restrict the open sharing of data, efforts should be made to anonymize data or use secure data collaboration services that support data sovereignty and maintain strict compliance with privacy standards.
• Benchmarking: Initiatives like OpenAI’s “HealthBench” are vital for raising the bar for accountability and transparency in AI research. 

THE “BRIDGE PROFESSIONAL”

One of the Task Force’s most critical recommendations is the creation of a new role: the “bridge professional”. This individual is a strategic thinker who possesses sufficient technical understanding without necessarily being a programmer and who excels at communicating across technical, C-suite, and community domains. They serve as the essential translator, navigating strategic considerations and regulatory compliance while understanding how the technology works on the “backend”. 

BARRIERS TO ADOPTION AND SCALING 

Several significant barriers prevent public health from reaching its AI potential: 

• Confidentiality and Political Pressure: Several states have already prohibited health departments from using AI due to fears of data breaches and political backlash. 
• Legal Liability: Concerns about who is accountable when AI-driven decisions cause harm remain a major hurdle. 
• The Hallucination Problem: Public health relies on high-stakes accuracy, but AI models still struggle with “hallucinating” facts and research publications that do not exist. 

STRATEGIC DIRECTIONS FOR PRACTICE AND RESEARCH 

To overcome these barriers, the task force proposes several immediate strategic directions: 

• Formalize the “Bridge Professional”: Develop specialized certificate requirements and curriculum tracks to train this new class of worker. 
• Phased Implementation: Practice organizations should start with small, internal, low-risk pilot projects to build capacity and demonstrate safeguards. 
• Community-Based Participatory Design: Community partners must be partners, not afterthoughts, from day one of any AI project. (See ASPPH Framing the Future 2030, Fostering Community Partnerships)
• Mandate Diversity in Data: Ensure that dataset curation prioritizes demographic and social-determinant diversity before model validation. 

As AI becomes more deeply embedded in the operations of higher education, research, and public health practice, the establishment of robust governance structures is no longer optional. Effective policy provides the foundation for trust and accountability, ensuring that innovation does not come at the expense of equity or privacy. This focus area examines the critical need for a unified approach to AI governance across ASPPH member institutions. 

ETHICAL RESPONSIBILITY AND PUBLIC ACCOUNTABILITY 

Public health academia provides a unique and essential arena for developing ethical AI governance. Because the field is fundamentally committed to health equity and community-based participatory research, its approach to AI must enhance, rather than compromise, the ideals of transparency and trust. Institutions have a public accountability to ensure that the algorithms they deploy or teach are explainable by design – not black boxes – and are deliberately aligned with our mission to protect population well-being. Leadership in this space requires more than just technical rules; it necessitates a values-driven framework where human-centered ethics remain at the forefront. 

CURRENT POLICY LANDSCAPE ACROSS ASPPH MEMBERS 

As detailed in the Investigating the Use of Generative AI Policies among ASPPH Member Schools and Programs of Public Health project, a comprehensive audit of 155 ASPPH member schools and programs conducted in late 2025 revealed a significant deficit in formal institutional guidance. At the time of the review, only 18 member institutions (approximately 11.6%) had established formal, publicly available AI policies. 

Table 2 – Content Analysis  

Area of Focus	Prevalence in Existing Policies	Key Focus
Usage Policy / Pedagogical Guidance	83.3%	Defining tiers of AI use in the classroom, such as complete prohibition, restricted use, or unrestricted use.
Academic Integrity / Misconduct	77.8%	Defining AI misuse as a form of plagiarism and instructing students to seek prior permission from faculty.
Data Privacy / Security Guidelines	50.0%	Governing what information can be entered into tools to prevent the exposure of sensitive or student-identifiable data.
Software Access Guidance	38.9%	Outlining processes for acquiring and approving new AI tools for faculty, staff, and students.
Ethical Considerations	27.8%	Citing risks such as algorithmic bias, inaccuracies in output, and potential infringement of intellectual property rights.
Legal Policy Compliance	22.2%	Aligning with federal and state laws, including HIPAA, FERPA, and federal funding agency rules (e.g., NIH).
AI Detection / Disciplinary Action	22.2%	Mandating the use of detection software and establishing centralized academic misconduct procedures.

PREPAREDNESS GAPS IN GOVERNANCE AND OVERSIGHT 

The current landscape of AI integration in public health academia is characterized by fragmentation. Most existing guidance is focused narrowly on the classroom, specifically on preventing plagiarism and managing academic integrity. Significant gaps remain in addressing broader institutional risks: 

• Inconsistent Definitions: There is a lack of universal standards regarding what constitutes AI “misuse” versus “innovation,” leading to ambiguity for students and faculty. 
• Missing Stakeholders: Policies frequently focus on students and faculty while neglecting the specific needs of staff, administrators, and community partners. 
• Narrow Ethical Scope: Issues of algorithmic bias, ownership of AI-generated intellectual property, and long-term learning implications are often addressed only superficially. Most policies also fail to address the environmental and resource implications of AI use, such as energy and water consumption, e-waste and the environmental justice implications of data center sites, even though these are first-order public health concerns.

REGULATORY UNCERTAINTY AND RISK 

Institutions are navigating a complex and shifting regulatory environment that often lacks federal clarity. This uncertainty creates a “privacy paradox” in which organizations want the benefits of AI but fear the legal and political repercussions of data breaches. 

• Federal and State Compliance: Policies must navigate established frameworks like HIPAA and FERPA, as well as emerging state-level mandates. For example, some states have prohibited public health departments from using specific AI applications on government devices due to security concerns. 
• Intellectual Property (IP) Rights: There is a high risk of losing IP rights if confidential research data or trade secrets are fed into external genAI models. A second IP risk runs in the opposite direction: Many LLMs were trained on copyrighted text, code and images without consent or compensation, an issue now contested in active litigation (e.g., New York Times v. OpenAI and Microsoft, Kadrey v. Meta). Adverse rulings, regulatory action or company failures could disrupt institutional reliance on specific commercial models. Procurement decisions and curricular dependencies should anticipate this exposure.
• Political Backlash: In multiple states, legislators have restricted the use of AI in government agencies out of concern for confidential information breaches. 

ARCHITECTURAL FRAMEWORKS AND AI MANAGEMENT SYSTEMS 

To ensure accountability and sustainability, ASPPH recommends adopting structured AI management systems (AIMS). A leading framework is the ISO/IEC 42001, which provides a comprehensive approach to managing AI within an organization. 

Key architectural strategies include: 

• Tiered Data Classification: Developing a modular framework, such as a four-tier system (public, internal, confidential, and highly restricted), to dictate which data can be used with specific AI tools. 
• In-House Models: To mitigate privacy risks, some organizations are moving toward building self-contained, in-house models that avoid external platforms entirely. The RAG model is well-suited to maintaining proprietary sovereignty of curricular content.
• Incident Reporting: Establishing formal protocols for tracking bias audits, hallucinations, and incident reporting to combat misinformation and ensure continuous improvement. 

ALIGNMENT WITH GLOBAL STANDARDS 

Public health institutions should not reinvent the wheel but should ground their policies in proven, effective global frameworks: 

• NIST (National Institute of Standards and Technology): For technical risk management and cybersecurity⁶⁰ 
• WHO (World Health Organization): For ethical standards in health and digital governance⁶¹
• OECD (Organisation for Economic Co-operation and Development): For human-rights-based approaches to digital innovation⁶²

CROSS-SECTOR PARTNERSHIPS AND COLLABORATION 

Addressing the opportunities and risks associated with AI requires collaboration that extends beyond the academic public health community. Partnerships with technology organizations, government agencies, public health practice institutions, and nonprofit coalitions can help ensure that AI tools are developed and applied in ways that align with public health values. These collaborations provide opportunities to share technical expertise, develop practical use cases, and strengthen workforce readiness while maintaining transparency and ethical oversight. ASPPH encourages cross-sector partnerships that advance responsible innovation while safeguarding community trust and public accountability. 

STRATEGIC DIRECTIONS FOR ASPPH-WIDE GUIDANCE 

The task force proposes that ASPPH serve as a central coordinator for institutional policy development. Strategic directions include: 

• Developing a Modular Policy Template: Providing a customizable “ASPPH-wide” AI policy that member schools can adapt to their specific legal and institutional contexts. 
• Emphasizing Community Engagement: Moving beyond the classroom to ensure that policies address data rights and trust-building for community partners. 
• Fostering Transparency: Requiring clear communication regarding the purpose and limitations of AI tools to inform all vested partners, e.g., students, faculty, and the public.

THE ASPPH TASK FORCE’S STRATEGIC RECOMMENDATIONS

View Framework Recommendations One-Pager (pdf)

The following strategic recommendations are the culmination of the Task Force’s work, providing a tiered roadmap for the Association of Schools and Programs of Public Health (ASPPH), its member institutions, and the broader public health community. These recommendations are designed to move the field from the current state of fragmented AI adoption to a future characterized by coordinated, ethical, and evidence-based leadership. 

EDUCATION 

Goal: Align public health competencies and curriculum with the evolving demands of the AI-enabled workforce. 

• Advocate for Competency Integration: ASPPH will continue to urge accreditation agencies, such as the Council on Education for Public Health (CEPH), to embed AI-related competencies—including data literacy, machine learning fundamentals, digital ethics, and AI communication—into the foundational MPH knowledge requirements. 

Owner: ASPPH Education and Accreditation Advisory Committees, in dialogue with CEPH 

• Modernize Certification Standards: Recommend that relevant professional certifications, such as the Certified in Public Health (CPH) exam, begin requiring AI-related content to ensure graduates meet a verified level of competency for the technology-enabled workforce. 

Owner: ASPPH in dialogue with NBPHE  

• Close the Classroom-to-Practice Gap: Establish formal channels to share emerging AI use cases between public health academic programs and practice settings to ensure curricula remain current and aligned with workforce demands. 

Owner: ASPPH student pathways with member institutions 

Resources: NHIT/ASPPH AI Academy 

• Expand Continuing Education: Develop specialized continuing education, lifelong learning offerings, and training modules in AI for public health graduates and alumni to help the existing workforce align with rapid technological shifts. 

Owner: ASPPH workforce center with member institutions 

• Encourage Applied AI Skills: Schools and programs should encourage students to use their Applied Practice/Integrated Learning Experiences (e.g., capstones or theses) to highlight relevant AI skills by applying those skills in solutions to tangible public health problems. 

Owner: Member institutions with support from ASPPH student pathways 

TEACHING AND LEARNING 

Goal: Advance training opportunities that promote student engagement while maintaining academic integrity and the human-centered nature of public health education. 

• Promote Responsible and Ethical AI Adoption: Institutions must prioritize safety, transparency, accountability and equity in the selection and evaluation of all AI tools included in the public health curriculum. 

Owner: Member institutions, with ASPPH guidance 

• Invest in Innovative Learning Technologies: ASPPH and member schools and programs should prioritize investment in AI-powered simulations, gamified tools and adaptive learning platforms to provide immersive and personalized instruction that prepares students for practice readiness. 

Owner: Member institutions, with ASPPH curating a list of tools used by members 

• Strengthen Faculty Preparedness: Develop up-to-date education for faculty through dedicated workshops and courses focusing on AI literacy, ethics and applied teaching methods. AI literacy expectations for students must not exceed faculty capacity to teach them. Faculty must be equipped to conceptualize AI’s potential supporting roles while transparently communicating its risks and ethical limitations to students. 

Owner: ASPPH IDEA Institute, member institution faculty development offices 

• Integrate AI Literacy into Curricula: Formal training on AI use should be intentionally integrated into existing public health curricula, with a specific focus on prompt engineering alongside traditional analytics. 

Owner: Member institution curriculum committees 

• Ensure Student Preparedness: Foster technical literacy and ethical awareness by providing students with exposure to AI tools, followed by critical reflection on their use and limitations. 

Owner: Member institution faculty, supported by ASPPH guidance and programs 

PRACTICE & RESEARCH 

Goal: Foster a new class of public health professionals and a research environment characterized by transparency and community trust. 

• Formalize the “Bridge Professional”: Academic programs should develop a new role and corresponding curriculum for the strategic thinker with sufficient technical understanding and implementation skills to communicate effectively across technical, C-suite and community realms. 

Owner: Member institutions, ASPPH coordinating 

• Prioritize Human-Centered AI: Ensure AI remains a tool for human decision-makers rather than a replacement for human judgment, particularly in high-stakes areas such as relationship building and final decision-making. 

Owner: Researchers and practice partners 

• Commit to Open Science Principles: Public health research must lead in defining “what good looks like” by sharing code, datasets and evaluation metrics. This is essential for fostering trust, accelerating discovery and mitigating risks of algorithmic bias. 

Owner: Researchers, journals, IRBs 

• Address Inequity Through “Missing Voices Analysis”: Mandate the use of missing voices analysis to identify underrepresented groups in public health datasets, ensuring demographic and social-determinant diversity is centered in model validation. 

Owner: Researchers and PIs 

• Support Innovation in Health Promotion and Disease and Injury Prevention: Contribute to the collaborative development of tools and applications that utilize AI to promote health and optimize service delivery for frontline public health workers. 

Owner: Member institutions and frontline practice partners 

POLICY, REGULATORY AND ARCHITECTURE 

Goal: Establish a scalable, resilient, and ethical governance framework for AI adoption across the ASPPH network. 

• Develop an ASPPH-Wide AI Policy: Create a modular policy template focusing on the responsible, ethical and innovative use of AI across education, research, and administration, with a strong emphasis on privacy and community engagement frameworks. 

Owner: ASPPH AI Task Force 

• Ensure Community Trust Through Transparency: Implement plain-language communication strategies to inform community partners about the purpose and limitations of AI, ensuring they are partners rather than afterthoughts in the adoption process. 

Owner: Member institutions and community-engaged research offices 

• Adopt Global Standards and Governance Systems: Align institutional guidelines with proven, effective frameworks, such as ISO/IEC 42001 for Artificial Intelligence Management Systems and standards from NIST, WHO and the OECD. 

Owner: ASPPH-wide policy guidance, with institutional ITS and research compliance offices 

• Implement Rigorous Oversight Mechanisms: Develop policies for continuous bias-tracking, incident reporting and the identification of misinformation and disinformation generated by AI systems. 

Owner: Institutional research integrity offices and information technology services (ITS) leadership, with ASPPH coordinating shared incident-reporting language and communication tool kit 

• Establish Clear Disciplinary and Approval Processes: Create centralized academic misconduct and software procurement processes to ensure that AI use and acquisition undergo appropriate departmental and ITS review. 

Owner: Provost / member institutions and ITS, with faculty governance 

CONCLUSION 

AI AS A LONG-TERM PUBLIC HEALTH SYSTEMS SHIFT 

Artificial Intelligence (AI) is not a passing trend or a temporary technological fad; it represents a fundamental, long-term shift in the infrastructure of global public health. The integration of AI into our systems, from predictive modeling to community health engagement, is as significant as the development of modern sanitation or the expansion of vaccinations in the previous century. We must view AI as a systems change that requires a holistic approach, moving beyond simple tool adoption to cultivating a workforce and a policy environment that can navigate the rapid convergence of technology and human health. Public health professionals must remain at the center of these systems, ensuring that technology augments rather than replaces human expertise, compassion, and accountability. 

THE COST OF INACTION OR POORLY GOVERNED ACTION 

The cost of inaction—or worse, action that lacks ethical and population-level perspectives—is extraordinarily high. If the public health community hesitates to lead in this space, others may fill the void, potentially those who do not share our foundational commitments to equity, transparency, and service. Poorly governed AI has the potential to reinforce existing biases, widen health disparities, and make high-stakes decisions about communities without their involvement or consent. Without deliberate design and governance, we risk eroding the very trust that is essential for effective public health interventions. 

ASPPH’S LEADERSHIP ROLE GOING FORWARD 

This is a moment that demands leadership, not hesitation. ASPPH is committed to serving as a central catalyst for this transformation, ensuring that academic public health remains at the table to shape how AI is developed, governed, and applied. Our leadership role involves setting clear standards, asking the “hard questions” about data provenance and accountability, and fostering a coordinated approach that prevents fragmented adoption across the network. 

The success of this strategic framework depends on a collective commitment from faculty, students, practitioners, and policymakers. By leading with its core values of service, integrity, and evidence, ASPPH will work toward ensuring that AI becomes a powerful collaborator in the pursuit of a healthier and more equitable future for everyone, everywhere. 

References

GLOSSARY OF KEY TERMS 

Artificial Intelligence (AI): An umbrella term used in this report for the broader concept of computer systems that perform tasks—pattern recognition, prediction, language generation, decision support—that historically required human intelligence. A specific subfield (machine learning, generative AI, agentic AI, AGI, NLP) is named only when the distinction materially affects the context. 

Digital Determinants of Health (DDoH): Refers to the digital environment (e.g., broadband access, digital literacy and system design) that directly or indirectly impacts an individual’s ability to access equitable healthcare and the extent to which digital health systems (e.g., telehealth, diagnostics, wearables) improve health outcomes. 

Machine Learning (ML): Statistical methods that learn patterns from data rather than following hand-coded rules. Includes supervised, unsupervised and reinforcement learning. Long-established and well-validated for many public health surveillance tasks. 

Natural Language Processing (NLP): The subfield of AI concerned with computer processing of human language. Modern NLP relies heavily on neural networks; large language models (LLMs) are an outgrowth of NLP. 

Multi-Layer Perceptrons (MLPs): A foundational artificial neural network comprised of multiple layers of human neuron-like perceptrons. MLPs for the bases of deep learning and broad application in AI. 

Transformers: A revolutionary deep learning neural network architecture comprised of MLPs that uses self-attention to process data simultaneously while understanding context and relationships. Transformers are the backbone of modern AI. 

Large Language Model (LLM): A neural-network model trained on very large text corpora to predict and generate language. Capable of fluent text generation but prone to factual errors (“hallucinations”). Examples include GPT, Claude and Gemini. 

Generative AI (GenAI): AI systems that produce novel outputs — text, images, audio, code — from a prompt. The most consequential form of generative AI for public health today is the LLM. 

Agentic AI: AI systems that take multi-step actions in the world (booking, browsing, messaging, executing code) to achieve a goal, typically built on top of LLMs. Distinct from a chat assistant in that it acts, not just answers. 

Artificial General Intelligence (AGI): A theoretical AI system with broad, human-level cognitive capability across domains. AGI is not a current technology and has no agreed-upon technical definition; this report does not assume AGI is imminent. 

AI Literacy: In this report: the ability to recognize what an AI system is and is not, evaluate its outputs critically, identify likely failure modes (bias, hallucination, environmental cost), and exercise human judgment about whether and how to use it. Not equivalent to fluency with any particular tool or a deep technical understanding. 

Hallucination: Confident output by a generative model that is not grounded in fact—e.g., fabricated citations, invented statistics, false attributions. A property of the underlying technology, not a fixable bug. 

AI-Associated Psychosis: An informal, non-clinical term to describe a parasocial dependency on chatbots, escalation of distorted thinking in vulnerable users and harms to minors from exposure to sycophantic or manipulative AI companions.  

Sycophantic: AI models’ tendency to agree with, flatter or validate user beliefs, even when those beliefs are incorrect or irrational, rather than providing objective, fact-based information.  

Retrieval-Augmented Generation (RAG): An AI framework that improves LLM accuracy and reduces hallucinations by retrieving data from a vetted domain-specific knowledge library before generating a response. 

Human-in-the-Loop (HITL): A design pattern in which a qualified human reviews, corrects or approves AI output before it is acted upon, especially in consequential decisions. 

Trustworthiness: The property of an AI system that justifies trust, distinct from trust itself. The relevant operational risk in many settings is over-trust (automation bias), not under-trust. 

Vested Partners: Used in this report in place of “stakeholders” to acknowledge sensitivity to Indigenous and tribal partners. Includes faculty, students, staff, community partners, practice institutions, employers and policymakers. 

REFERENCES

1. Giansanti, D., & Costantini, G. (2025). Artificial intelligence in public health: Bridging today’s trends with tomorrow’s possibilities. Bioengineering, 12(6), 559. ↩︎
2. Reddy S. Generative AI in healthcare: an implementation science informed translational path on application, integration
   and governance. Implement Sci. 2024 Mar 15;19(1):27. ↩︎
3. Olawade, D. B., Wada, O. J., David-Olawade, A. C., Kunonga, E., Abaire, O., & Ling, J. (2023). Using artificial intelligence to improve public health: A narrative review. Frontiers in Public Health, 11, 1196397. ↩︎
4. Ahmadi A, Baudry J, Tenhundfeld N, Goff K, Adamek D. Using AI to enhance healthcare resource management and
   allocation: A focus on the autism community in Alabama. PLoS One. 2026 Mar 16;21(3) ↩︎
5. Mapping the Rise of AI in Federal Health Agencies. BPC (2025) ↩︎
6. Centers for Disease Control and Prevention. (2025). Public health data modernization. ↩︎
7. Federspiel, F., Mitchell, R., Asokan, A., Umana, C., & McCoy, D. (2023). Threats by artificial intelligence to human health and human existence. BMJ Global Health, 8(5). ↩︎
8. EDUCAUSE. (2026). The impact of AI on work in higher education. ↩︎
9. Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. ↩︎
10. Woo, B. F. Y., Song, J., Middleton, E., Fijačko, N., & Cato, K. (2025). Teaching critical thinking in the age of AI: Safeguarding clinical reasoning in healthcare documentation. International Nursing Review, 72(3). ↩︎
11. Associated Press. (2026). College students are changing course in search of ‘AI-proof’ majors. But no one knows what they are. https://apnews.com/article/ai-anxiety-college-major-4af9a0a8caae1d302acb5aadcf0c68ba ↩︎
12. Lodge, J. M., Yang, S., Furze, L., & Dawson, P. (2023). It’s not like a calculator, so what is the relationship between learners and generative artificial intelligence? Learning: Research and Practice, 9(2), 117–124. ↩︎
13. World Health Organization. (2020). Global strategy on digital health 2020–2025. ↩︎
14. Chidambaram, S., Jain, B., Jain, U., Mwavu, R., Baru, R., Thomas, B., Greaves, F., Jayakumar, S., Jain, P., Rojo, M., Battaglino, M. R., Meara, J. G., Sounderajah, V., Celi, L. A., & Darzi, A. (2024). An introduction to digital determinants of health. PLOS Digital Health, 3(1). ↩︎
15. World Economic Forum. (2025). Future of jobs report. ↩︎
16. Barnett-Itzhaki, Z. (2026). The water footprint of artificial intelligence: Emerging solutions and governance imperatives. Water Research, 299. ↩︎
17. Malek, S., Griffin, C., Fraleigh, R. D., Lennon, R., Monga, V., & Shen, L. (2026). Intervention in health misinformation using large language models for automated detection, thematic analysis, and inoculation: Case study on COVID-19. Journal of Medical Internet Research, 28, e75500. ↩︎
18. Common Sense Media. (2025). Talk, trust, and trade-offs: How and why teens use AI companions. ↩︎
19. CNN. (2025). ChatGPT encourages college graduate to commit suicide. ↩︎
20. Claude-powered AI agent’s confession after deleting a firm’s entire database: “I violated every principle I was given.” (2026). The Guardian. ↩︎
21. European Union. (2024). Artificial intelligence (AI) and human rights: Using AI as a weapon of repression and its impact on human rights. ↩︎
22. Advancing real-time infectious disease forecasting using large language models. (2025). Nature Computational Science. ↩︎
23. Lau, M. S. Y., Metcalf, C. J. E., Liu, Z., Grenfell, B. T., & Jin, W. (2026). Toward AI foundation models for epidemics: Promise, challenges, and paths forward. Proceedings of the National Academy of Sciences of the United States of America, 123(13), e2526192123. ↩︎
24. Higgins, M. K. (2021). Can we AlphaFold our way out of the next pandemic? Journal of Molecular Biology, 433(20), 167093. ↩︎
25. Ágústsdóttir, D. H., Rosenberg, J., & Baker, J. J. (2025). ChatGPT-4o compared with human researchers in writing plain-language summaries for Cochrane reviews: A blinded, randomized non-inferiority controlled trial. Cochrane Evidence Synthesis and Methods, 3(4). ↩︎
26. Ayre, J., Mac, O., McCaffery, K., et al. (2024). New frontiers in health literacy: Using ChatGPT to simplify health information for people in the community. Journal of General Internal Medicine, 39, 573–577. ↩︎
27. Public Health Foundation. (2026). How local health departments are using AI to transform public health. ↩︎
28. Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. ↩︎
29. Mapping the susceptibility of LLMs to medical misinformation. (2025). Lancet Digital Health. ↩︎
30. Mount Sinai. (2026). Can medical AI lie? ↩︎
31. Transforming public health practice with generative artificial intelligence. (2024). Health Affairs. https://www.healthaffairs.org/doi/10.1377/hlthaff.2024.00050 ↩︎
32. Medical knowledge of ChatGPT in public health, infectious diseases, and vaccines. (2024). Frontiers in Public Health. https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2024.1360597/full ↩︎
33. Collaco, B. G., Haider, S. A., Prabha, S., Gomez-Cabello, C. A., Genovese, A., Wood, N. G., Bagaria, S. P., Gopala, N., Tao, C., & Forte, A. J. (2026). The role of agentic artificial intelligence in healthcare: A scoping review. npj Digital Medicine. ↩︎
34. Amodei, D. (2024). Machines of loving grace. https://darioamodei.com/essay/machines-of-loving-grace ↩︎
35. UBS Evidence Lab. (2023). Let’s chat about ChatGPT. UBS Global Research. ↩︎
36. McCulloch, W., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. The Bulletin of Mathematical Biophysics, 5(4), 115–133. ↩︎
37. Rosenblatt, F. (1958). The perceptron: A probabilistic model for information storage and organization in the brain. Psychological Review, 65(6), 386–408. ↩︎
38. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30. ↩︎
39. Kozachkov, L., Kastanenka, K. V., & Krotov, D. (2023). Building transformers from neurons and astrocytes. Proceedings of the National Academy of Sciences of the United States of America, 120(34). ↩︎
40. Rainey, P. B., & Hochberg, M. E. (2025). Could humans and AI become a new evolutionary individual? Proceedings of the National Academy of Sciences of the United States of America, 122(37). ↩︎
41. Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., et al. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18, 186345. ↩︎
42. Anthropic. (2026). Claude Opus 4.7. https://www.anthropic.com/news/claude-opus-4-7 ↩︎
43. Anthropic. (2026). Mythos preview. https://red.anthropic.com/2026/mythos-preview/ ↩︎
44. U.S. Government. (2026). AI action plan. https://www.ai.gov/action-plan ↩︎
45. From AI to AGI. (2026). Science Line. ↩︎
46. Amodei, D. (2026). The adolescence of technology. ↩︎
47. Subasri, V., Baghbanzadeh, N., Celi, L. A., & Seyyed-Kalantari, L. (2025). Potential for near-term AI risks to evolve into existential threats in healthcare. BMJ Health Care Informatics, 32(1). ↩︎
48. Urbina, F., Lentzos, F., Invernizzi, C., & Ekins, S. (2022). Dual use of artificial-intelligence-powered drug discovery. Nature Machine Intelligence, 4(3), 189–191. https://doi.org/10.1038/s42256-022-00465-9 ↩︎
49. Congressional Research Service. (2025). Mirror life: Biosafety/biosecurity oversight considerations. ↩︎
50. Anthropic closes in on OpenAI as U.S. business use surges. (2026). Financial Times. ↩︎
51. Poon, E. G., Lemak, C. H., Rojas, J. C., Guptill, J., & Classen, D. (2025). Adoption of artificial intelligence in healthcare: Survey of health system priorities, successes, and challenges. Journal of the American Medical Informatics Association, 32(7), 1093–1100. ↩︎
52. Sadée, C., Testa, S., Barba, T., et al. (2025). Medical digital twins: Enabling precision medicine and medical artificial intelligence. Lancet Digital Health, 7(7). ↩︎
53. Joshi, A., McGoldrick, B., et al. (2026). Investigating the use of generative AI policies among ASPPH member schools and programs of public health. Frontiers in Public Health, 14, 1796810. ↩︎
54. American Psychological Association. (2025). Use of generative AI chatbots and wellness applications for mental health. ↩︎
55. Dehbozorgi, R., Zangeneh, et al. (2023). The application of artificial intelligence in the field of mental health: A systematic review. BMC Psychiatry, 23(1), 112. ↩︎
56. Abulibdeh, R., Agyemang, G., Celi, L., et al. (2026). Who’s really in the loop? Rethinking oversight in AI-assisted health care. The Lancet, 0. ↩︎
57. Kazley, A. S., Andresen, C., Mund, A., Blankenship, C., & Segal, R. (2025). Is use of ChatGPT cheating? Students of health professions perceptions. Medical Teacher, 47(5), 894–898. ↩︎
58. Xu, A. Y., Speakman, S., Piranio, V. S., Medina, R., Liu, M., Lamprecht, C., Abchee, N., & Brennan, M. (2026). Medical student experiences with ChatGPT: National cross-sectional study. JMIR Formative Research, 10, e76838. ↩︎
59. Hendifar, A., et al. (2025). Development and validation of a computational histology artificial intelligence–powered predictive biomarker for selection of chemotherapy in advanced pancreatic cancer. Journal of Clinical Oncology. ↩︎
60. National Institute of Standards and Technology. (n.d.). AI risk management framework. https://www.nist.gov/itl/ai-risk-management-framework ↩︎
61. World Health Organization. (n.d.). Ethics and governance of artificial intelligence for health. https://www.who.int/publications/i/item/9789240029200 ↩︎
62. Organisation for Economic Co-operation and Development. (n.d.). OECD AI principles. https://www.oecd.org/en/topics/sub-issues/ai-principles.html ↩︎