University of Hong Kong · AI Literacy & Controlled Deployment

University of Hong Kong's AI Literacy & Controlled Deployment program operates as an institutional deployment focused on controlled access to LLMs rather than experimental or pilot prototyping. It is an established policy-level initiative aiming to embed AI literacy systematically alongside restricted AI tool use within courses, emphasizing risk reduction concerning academic integrity violations. Agentivity is in the lower range (1–2/6) reflecting limited autonomous AI agency and strong administrative governance.

HKU — программа AI Literacy + controlled deployment как ответ на массовое adoption GenAI в азиатском контексте. Тип [[D-governance-training-ecosystem]].

Prior to AI integration, academic risks such as plagiarism and misuse of AI-generated content were prominent concerns. Traditional educational processes lacked mechanisms to ensure transparent and ethical AI use, leading to potential academic misconduct. The absence of AI literacy programs left students unprepared to responsibly engage with AI tools, and faculties had insufficient normative frameworks to manage AI deployment within curricula.

The program promises to improve AI literacy among students and faculty while controlling risks associated with AI misuse. By mandating AI-literacy modules and requiring students to disclose AI use, it aims to foster transparent, ethical AI integration. The controlled deployment model is designed to maintain academic integrity and support educators in defining acceptable AI use scenarios, rather than promoting autonomous AI decision-making.

The AI architecture is centered around multiple LLMs functioning as assistive tools limited to analysis and content generation within tightly controlled scenarios. AI tools do not operate autonomously but serve as assistants embedded within the educational workflow. Institutional IT infrastructure governs access and maintains compliance with privacy regulations, deploying AI services in a sandboxed environment. This setup reflects MOD (modular orchestration) and HYBR (hybrid control) facets, balancing AI assistance with human governance.

Roles are clearly delineated: university administration enforces policy and governance structures; faculty members define permissible AI usage parameters and integrate AI tools into course content; students engage with AI tools under mandated transparency rules by declaring AI use; AI itself acts as a non-autonomous assistant facilitating analysis and generation. This explicit role separation supports controlled deployment and ensures accountability within the educational ecosystem.

AI functions as a low-agentivity assistant (level 1–2/6), primarily supporting analysis and generative assistance without decision-making autonomy. It is integrated as a complementary educational tool rather than an independent agent. This role facilitates learning while mitigating risks associated with autonomous AI actions, aligning with the institution’s goals for controlled deployment and AI literacy.

Governance frameworks include mandatory AI-literacy education modules, institutional policies mandating disclosure of AI use in assignments, and a compliance regime to monitor these disclosures. The loop involves the administration setting policies, faculty enforcing acceptable use standards, and students participating under transparency obligations. This explicit governance maintains academic integrity and aligns with legal and ethical institutional requirements.

By 2024–2025, University of Hong Kong reported institutional scale deployment with mandatory AI literacy module completion for students and structured AI integration into courses. Although specific quantitative data on impact metrics (like plagiarism rates or literacy assessment scores) were not provided, the case is referenced as an institutional-level adoption with systematic governance mechanisms. Agentivity ratings consistently cluster at 1–2/6, indicative of controlled, low-autonomy AI usage.

The program exemplifies a deliberate institutional choice to limit AI autonomous agency despite evolutionary trends pushing towards agentic or multi-agent AI architectures. The conscious suppression of AI agentive depth highlights the tension between aspirations for AI-enhanced education and concerns over academic integrity, showing resistance to full autonomous AI participation. This stance functions as a countersignal within the broader meta-discourse on AI in education.

This case illustrates the conflict outlined in the tension between 'autonomization vs manageability' [[H4-autonomy-vs-control]], where the institution prioritizes governance and risk mitigation over increased AI agency. It aligns with scholarly observations of dominant institutional strategies favoring hybrid or modular orchestration models (MOD, HYBR) with agentivity levels capped at 1–2/6 [[A-governed-access]]. The program exemplifies layered orchestration including an administrative governance loop, human-faculty curatorship, and controlled AI assistant roles, corresponding to known educational agentic frameworks [[LIN]].