跳转至

文章背景与核心概要

本文提出了一种新颖的提示工程框架，旨在跨越不同的学科领域，在通用型AI助教（如 Jill Watson）中实现可扩展的微观级个性化。该框架无需进行成本高昂的模型重新训练，而是利用两个核心组件动态调节大语言模型（LLMs）与基于RAG（检索增强生成）的系统：1. 六个特定于学习者的维度（自我评估、抽象偏好、冗长偏好、感知取向、信息处理风格以及理解水平，共可生成 96 种独特的配置文件）；2. 认知复杂度估计（通过布鲁姆分类法实时分析学生的查询）。

通过自然语言处理（NLP）指标以及包含五名参与者的真人研究评估，该研究证明了基于提示的调节能够有效改变智能体的行为，针对不同的个体学习者产生可感知、可测量的响应风格和结构差异。这为利用提示实现LLM驱动型教育代理的自适应行为提供了初步证据。

A Prompt-Engineering Approach to Develop Scalable, Flexible, and Real-Time Hybrid Micro-Level Personalization in a General Purpose AI Teaching Assistant

A Prompt-Engineering Approach to Develop Scalable, Flexible, and Real-Time Hybrid Micro-Level Personalization in a General Purpose AI Teaching Assistant

Summary

Summary

本文提出了一种新颖的提示工程框架，旨在跨多个学科扩展通用型AI助教（例如 Jill Watson）中的微观个性化。该框架无需进行昂贵的模型重新训练，而是利用两个主要组件动态调节大语言模型（LLMs）和基于RAG的系统： 1. 六个特定于学习者的维度： 自我评估、抽象偏好、冗长偏好、感知取向、信息处理风格和理解水平（共生成 96 种独特的配置文件）。 2. 认知复杂度估计： 通过布鲁姆分类法在实时交互中分析学生的查询。

通过自然语言处理（NLP）指标和五名参与者的真人研究进行评估，本研究证明基于提示的调节能有效修改智能体行为，针对个别学习者产生可感知、可测量的响应风格和结构差异。

This paper presents a novel prompt-engineering framework designed to scale micro-level personalization in general-purpose AI teaching assistants (such as Jill Watson) across various academic disciplines. Without requiring costly model retraining, the framework dynamically conditions Large Language Models (LLMs) and RAG-based systems using two main components: 1. Six Learner-Specific Dimensions: Self-assessment, abstraction preference, verbosity preference, perceptual orientation, information processing style, and level of understanding (generating 96 unique profiles). 2. Cognitive Complexity Estimation: Analyzing student queries in real-time via Bloom's Taxonomy.

Evaluated through both Natural Language Processing (NLP) metrics and human studies with five participants, the study proves that prompt-based conditioning effectively modifies agent behavior, creating perceptible, measurable differences in response style and structure tailored to individual learners.

Document Metadata

Document Metadata

属性 详情
arXiv 标识符 arXiv:2609.03402 [cs.AI]
一级学科 计算机科学 > 人工智能 (cs.AI)
作者 Saptarshi Basu, Sandeep Kakar, Ashok Goel
提交日期 2026年9月3日
论文备注 7页，9幅图，IAAI27 会议
许可证 知识共享署名 4.0 国际许可协议 license icon
Attribute Details
arXiv Identifier arXiv:2609.03402 [cs.AI]
Primary Subject Computer Science > Artificial Intelligence (cs.AI)
Authors Saptarshi Basu, Sandeep Kakar, Ashok Goel
Submission Date September 3, 2026
Comments 7 pages, 9 figures, IAAI27 conference
License Creative Commons Attribution 4.0 International license icon

Abstract

Abstract

由大语言模型（LLMs）驱动的人工智能（AI）助教能够提供可扩展的教育支持，但往往缺乏个性化。本研究提出了一种基于提示工程的框架，用于跨学科和课程个性化定制通用型 LLM/RAG 驱动的 AI 助教（如 Jill Watson）。该框架使用六个学习者特定维度来调整回答：自我评估、抽象偏好、冗长偏好、感知取向、信息处理风格和理解水平，从而产生 96 种不同的学习者画像。此外，利用布鲁姆分类法分析学生查询，以估计交互层面的认知复杂度。学习者属性和认知评估被编码在结构化提示中，在无需重新训练模型的情况下调节 LLM。该框架通过 NLP 指标实验和包含 5 名参与者的真人研究进行了评估。结果表明，在不同的个性化条件下，响应风格和结构存在可感知的差异，统计分析识别出了与可测量响应变化相关的学习者属性。这些发现为基于提示的个性化可以支持 LLM 驱动的教育代理的自适应行为提供了初步证据。

Artificial intelligence (AI) teaching assistants powered by large language models (LLMs) offer scalable educational support but often provide limited personalization. This study presents a prompt-engineering-based framework for personalizing general-purpose LLM/RAG-based AI teaching assistants such as Jill Watson across academic disciplines and courses. The framework adapts responses using six learner-specific dimensions: self-assessment, abstraction preference, verbosity preference, perceptual orientation, information processing style, and level of understanding, yielding 96 distinct learner profiles. Student queries are additionally analyzed using Bloom's Taxonomy to estimate cognitive complexity at the interaction level. Learner attributes and cognitive assessments are encoded in structured prompts that condition the LLM without requiring model retraining. The framework is evaluated through experiments using NLP metrics and a human study with five participants. Results show perceived differences in response style and structure across personalization conditions, with statistical analyses identifying learner attributes associated with measurable response changes. These findings provide preliminary evidence that prompt-based personalization can support adaptive behavior in LLM-powered educational agents.

Full-Text & Reference Links

Full-Text & Reference Links