人工智能教育大模型赋能综合素质评价:理念、模型与展望

Generative Education Special Transformer Empowering Comprehensive Quality Evaluation:Concept,Model,and Prospect

综合素质评价是教育评价改革的“深水区”,也是实现高质量教育的“发展区”。在人工智能快速发展的背景下,综合素质评价需要遵循四大原则,即在“以育人为导向”的目标引领下,通过设计“教—学—评”一体化的评价框架,采用综合评价与个性化评价相结合的方式,运用数智化的动态评价形式,构建完备的综合素质评价体系。然而,实践中教育大数据技术发展滞后往往导致综合评价难以“平衡”,评价系统的智能推荐能力不足使得个性化评价难以“求异”,这些困境限制了评价育人功能的发挥。为此,文章提出以“第四代评价”为理论基础,结合“OSEMN”大数据分析框架和“1+N分布式智能体系统”结构,构建“教—学—评”一体化的人工智能教育大模型,赋能综合素质评价。未来教育教学中,教师应在预训练与微调阶段优化模型的评价能力,通过多种途径促进人机和谐共处,推动大模型与综合素质评价深度融合。

Comprehensive quality evaluation is both the'deep water zone'of educational evaluation reform and the'development zone'for achieving high-quality education.With the rapid advancements in artificial intelligence,comprehensive quality evaluation must adhere to four guiding principles:orienting evaluation goals by focusing on educating students,designing a'teaching-learningevaluation'integrated evaluation framework,combining comprehensive evaluation with personalized evaluation,and utilizing intelligent and dynamic evaluation methods to establish a complete system for comprehensive quality evaluation.However,the slow advancement of big educational data technology has impeded this goal.At the same time,the limited intelligent recommendation capabilities of existing evaluation systems have obstructed the differentiation for personalized assessments.This limitation ultimately restricts the effectiveness of evaluations in supporting student growth.This paper proposes leveraging a generative education special transformer to address these challenges and enhance comprehensive quality evaluation.Based on the fourth-generation evaluation theory,the model adopts the OSEMN data analysis framework and incorporates a“1+N distributed agent system”structure,building a generative education special transformer framework to integrate“teaching-learningassessment.”To further explore the potential applications of the large model,future efforts should focus on optimizing the model’s evaluation capabilities during the pre-training and fine-tuning stages while promoting human-machine collaboration through various approaches.This will facilitate the deep integration of the AI model with comprehensive quality evaluation.