摘要
作为对联合国教科文组织发布的《生成式人工智能教育与研究应用指南》进行系列解读的第一篇,文章着重对生成式人工智能技术原理进行追本溯源的辨析并考证其教育适用性。生成式人工智能是可跨人类思维表征符号加工和生成新内容的人工智能技术,但其并不理解语义和真实世界的技术局限会限制其变革教育的潜力。垄断性基础模型已引发数字基础设施升级的安全忧患并存在投射西方价值观和语言文化偏见的风险,故研发自主可控的基础模型和更具适用性的教育基础模型是生成式人工智能教育应用的安全底线。生成式人工智能主要以成人应用互联网产生的数据作为训练用数据,其基于聊天的内容服务方式不适合未成年人,应设置独立聊天的年龄下限。从教育内容提供的视域考证,基于广而杂数据集训练并采取付费内容服务模式的人工智能生成内容作为主要的教育内容获取来源,存在既不符合技术设计初衷也不符合教学目的的悖论。而从教学育人过程的视域考证,生成式人工智能输出内容可用自动化的“内容预制菜”明喻其惰化学习主体通过解构内容实现知识建构和技能获得的反智风险。课程内容制作、双基学习、特殊学习需求、高阶思维和探究实践,是设计主体适应和学科适用的生成式人工智能教学应用的主要情境与脉络。
The paper is the first edition of a series of interpretative articles on the Guidance for Generative Artificial Intelligence(GenAI) in education and research released by UNESCO,focusing on the tracking of the technique principle behind GenAI and the examination of its educational applicability.GenAI is a category of AI technology that can produce new content across symbolic representations used by human thinking,but its technological limitation of not being able to understand semantics of text will reduce its potential in transforming education.As foundation models,GenAI has triggered the upgrading of monopolized digital infrastructure and generated threats on digital security and it risks to project the linguistic and cultural bias held by western people whose data were used to train GPT models.The development of self-automatic GenAI models,especially EdGPT,is threshold for the secure and trustable use of GenAI in education.Based on the datasets generated by adults' application of the Internet used to train GPT models,the chatbots-based service is not appropriate for independent use by children,and national regulatory agencies should set up the age restriction for children's independent conversations with GenAI chatbots.From the perspective of provision of education content,the use of the service of generative content priced by tokens as main sources of education content is a paradox that is not supported by the original design of GenAI and the main aims of education.From the processes of learning and values fostering,an allegory of “automatic prefabricated content dishes” helps reveal that the unconscious use of outputs of GenAI risks of enfeeblement through reducing human agency in deconstructing content in order to construct knowledge and develop skills.The co-creation of inclusive curricular content,foundational learning,learners with special needs,and higher-order thinking as well as inquiry practices are the main contextual criteria to design human-agent and pedagogy-proper interactive use of GenAI in teaching and learning.
作者
苗逢春
MIAO Feng-chun(National Engineering Laboratory for Cyberlearning and Intelligent Technology,Faculty of Education,Beijing Normal University,Beijing,China 100875;Unit for Technology and AI in Education,Education Sector,Headquarters of UNESCO,Paris,France 75007)
出处
《现代教育技术》
CSSCI
2023年第11期5-18,共14页
Modern Educational Technology
