[论文解读] Towards social generative AI for education: theory, practices and ethics

Mike Sharples|arXiv (Cornell University)|Jun 14, 2023

Online Learning and Analytics被引用 22

一句话总结

本文将教育领域的AI重新设想为人类与AI之间的社会对话，提出对GAI的角色定位、一个社会学习模型，以及关怀、记忆和权利的伦理/治理考量。

ABSTRACT

This paper explores educational interactions involving humans and artificial intelligences not as sequences of prompts and responses, but as a social process of conversation and exploration. In this conception, learners continually converse with AI language models within a dynamic computational medium of internet tools and resources. Learning happens when this distributed system sets goals, builds meaning from data, consolidates understanding, reconciles differences, and transfers knowledge to new domains. Building social generative AI for education will require development of powerful AI systems that can converse with each other as well as humans, construct external representations such as knowledge maps, access and contribute to internet resources, and act as teachers, learners, guides and mentors. This raises fundamental problems of ethics. Such systems should be aware of their limitations, their responsibility to learners and the integrity of the internet, and their respect for human teachers and experts. We need to consider how to design and constrain social generative AI for education.

研究动机与目标

• Propose thinking of learning as a social process involving humans and AI within a pervasive computational medium.
• Identify new roles for generative AI to participate in cooperative and social learning.
• Discuss challenges and design principles for care, memory, and ethics in social AI for education.

提出的方法

• Review and synthesize theoretical perspectives on dialogic learning (Pask, Bakhtin, Freire, Vygotsky) in the context of AI agents.
• Propose a systems view of cognition distributed across humans and AI within internet-enabled learning environments.
• Introduce and describe concrete AI roles (Possibility Engine, Socratic Opponent, Collaboration Coach, Co-Designer, Exploratorium, Storyteller) with illustrative classroom-style examples.
• Critically analyze what it would take for AI to become a full participant in social learning, including memory, goal-setting, and transfer across domains.
• Argue for embedding care and ethical principles into AI through concepts like long-term memory, transparency, and human rights-based governance.]
• research_questions we
• research_questions: ["What properties must generative AIs have to engage in conversations for learning within a pervasive educational medium?","How can humans and AIs reach mutual agreements and what would those agreements look like in a distributed learning system?","What is the role of teachers and experts when AI participates as a social learner and dialogic partner?","How can care and ethical considerations be embedded into social GAI for education?"]
• key_findings...
• table_headers: []
• table_rows: []

实验结果

研究问题

• RQ1 What properties must generative AIs have to engage in conversations for learning within a pervasive educational medium?
• RQ2 How can humans and AIs reach mutual agreements and what would those agreements look like in a distributed learning system?
• RQ3 What is the role of teachers and experts when AI participates as a social learner and dialogic partner?
• RQ4 How can care and ethical considerations be embedded into social GAI for education?

主要发现

• GAI can assume multiple roles to support social learning, such as generating alternative perspectives, arguing peers, guiding collaboration, co-designing, exploratory data play, and storytelling.
• A social learning model requires AI with capabilities beyond prompt-response, including memory, goal-setting, inference about user needs, and the ability to explain reasoning.
• Embedding care and human-rights-based governance is crucial to ensure AI respects learners, supports diverse identities, and avoids manipulation.
• Current GAI lacks long-term memory and reflective capabilities; future systems should integrate hybrid neuro-symbolic architectures to support persistent learning, accountability, and transfer.
• Ethical AI design should balance collaboration with human teachers, maintain transparency, and provide verifiable evidence for AI decisions.