Multi-Stakeholder Alignment in LLM-Powered Collaborative AI Systems: A Multi-Agent Framework for Intelligent Tutoring
2025/10/27 by Uchoa, Alexandre P, Oliveira, Carlo E T, Motta, Claudia L R +1
#FOS: Computer and information sciences #Human-Computer Interaction (cs.HC) #Multiagent Systems (cs.MA)
paper · doi:10.48550/arxiv.2510.23245
Abstract
The integration of Large Language Models into Intelligent Tutoring Systems pre-sents significant challenges in aligning with diverse and often conflicting values from students, parents, teachers, and institutions. Existing architectures lack for-mal mechanisms for negotiating these multi-stakeholder tensions, creating risks in accountability and bias. This paper introduces the Advisory Governance Layer (AGL), a non-intrusive, multi-agent framework designed to enable distributed stakeholder participation in AI governance. The AGL employs specialized agents representing stakeholder groups to evaluate pedagogical actions against their spe-cific policies in a privacy-preserving manner, anticipating future advances in per-sonal assistant technology that will enhance stakeholder value expression. Through a novel policy taxonomy and conflict-resolution protocols, the frame-work provides structured, auditable governance advice to the ITS without altering its core pedagogical decision-making. This work contributes a reference architec-ture and technical specifications for aligning educational AI with multi-stakeholder values, bridging the gap between high-level ethical principles and practical implementation.
Citations
- Selecting and Merging: Towards Adaptable and Scalable Named Entity Recognition with Large Language Models
- A Survey of LLM-Driven AI Agent Communication: Protocols, Security Risks, and Defense Countermeasures
- Explainable Rule Application via Structured Prompting: A Neural-Symbolic Approach
- Multi-Agent Language Models: Advancing Cooperation, Coordination, and Adaptation
- Reasoning with RAGged events: RAG-Enhanced Event Knowledge Base Construction and reasoning with proof-assistants
- Open Challenges in Multi-Agent Security: Towards Secure Systems of Interacting AI Agents
- Quantitative Auditing of AI Fairness with Differentially Private Synthetic Data
- GLaMoR: Consistency Checking of OWL Ontologies using Graph Language Models
- Enforcement Agents: Enhancing Accountability and Resilience in Multi-Agent AI Frameworks
- AgentSpec: Customizable Runtime Enforcement for Safe and Reliable LLM Agents
- Speaking the Language of Teamwork: LLM-Guided Credit Assignment in Multi-Agent Reinforcement Learning
- Position: Towards a Responsible LLM-empowered Multi-Agent Systems
- "Would You Want an AI Tutor?" Understanding Stakeholder Perceptions of LLM-based Systems in the Classroom
- Few-shot Policy (de)composition in Conversational Question Answering
- Multi-Agent Collaboration Mechanisms: A Survey of LLMs
- zkLLM: Zero Knowledge Proofs for Large Language Models
- A proposal for a revised meta-architecture of intelligent tutoring systems to foster explainability and transparency for educators
- AI and personalized learning: bridging the gap with modern educational goals
- From Guidelines to Governance: A Study of AI Policies in Education
- Trust in AI: Progress, Challenges, and Future Directions
- Multi-stakeholder Perspective on Responsible Artificial Intelligence and Acceptability in Education
- Large Language Model based Multi-Agents: A Survey of Progress and Challenges
- Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- Exploring Large Language Model based Intelligent Agents: Definitions, Methods, and Prospects
- Human-AI Collaboration in Real-World Complex Environment with Reinforcement Learning
- AI Control: Improving Safety Despite Intentional Subversion
- The Unlocking Spell on Base LLMs: Rethinking Alignment via In-Context Learning
- Understanding Your Agent: Leveraging Large Language Models for Behavior Explanation
- Large Language Models can Strategically Deceive their Users when Put Under Pressure
- Empirical Study of Zero-Shot NER with ChatGPT
- Exploring Collaboration Mechanisms for LLM Agents: A Social Psychology View
- Avalon's Game of Thoughts: Battle Against Deception through Recursive Contemplation
- How to Catch an AI Liar: Lie Detection in Black-Box LLMs by Asking Unrelated Questions
- The Rise and Potential of Large Language Model Based Agents: A Survey
- Are Emergent Abilities in Large Language Models just In-Context Learning?
- AI Deception: A Survey of Examples, Risks, and Potential Solutions
- MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework
- SentimentGPT: Exploiting GPT for Advanced Sentiment Analysis and its Departure from Current Machine Learning
- DP-Auditorium: a Large Scale Library for Auditing Differential Privacy
- Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate
- Training Socially Aligned Language Models on Simulated Social Interactions
- Improving Factuality and Reasoning in Language Models through Multiagent Debate
- Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinement
- A Survey of Federated Evaluation in Federated Learning
- A Comprehensive AI Policy Education Framework for University Teaching and Learning
- Tool Learning with Foundation Models
- Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark
- On the Creativity of Large Language Models
- Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback
- On the Planning Abilities of Large Language Models (A Critical Investigation with a Proposed Benchmark)
- Toolformer: Language Models Can Teach Themselves to Use Tools
- LLM-Planner: Few-Shot Grounded Planning for Embodied Agents with Large Language Models
- Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?
- Emergent Abilities of Large Language Models
- A Transparency Index Framework for AI in Education
- WebGPT: Browser-assisted question-answering with human feedback
- Language Models are Few-Shot Learners
- The Surprising Effectiveness of Test-Time Training for Few-Shot Learning
- BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models
- MetaAgents: Large Language Model Based Agents for Decision-Making on Teaming
Related