Scholarlink

We present PromptHive, a next-generation open-source content production platform that empowers domain experts through a collaborative interface for AI-driven prompt engineering. PromptHive ensures subject matter experts (SMEs) remain central to AI-assisted content creation, providing a structured process to curate, refine, and validate instructional materials at scale. Designed to guide large language model (LLM) outputs with precision, PromptHive supports rapid experimentation, iterative refinement, and collaborative authoring, all while aligning with educational workflows. Users can customize generation parameters–including model selection (GPT-3.5, GPT-4, GPT-4o) and temperature–and experiment with both lesson-level and textbook-level prompting to fine-tune outputs for instructional effectiveness. PromptHive integrates with structured content formats (e.g., JSON), making it easily adoptable across educational ecosystems. We validate PromptHive through usability and cognitive load studies with subject matter experts, and demonstrate its effectiveness via a learning gain study with 358 students–showing performance on par with human-curated hints. This makes PromptHive a powerful tool for advancing scalable, high-quality, and human-centered educational content development.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

PromptHive: Demonstrating Collaborative, Human-Centered OER Creation with LLMs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zachary A. Pardos,
  • Shreya Bhandari,
  • Ioannis Anastasopoulos

摘要

We present PromptHive, a next-generation open-source content production platform that empowers domain experts through a collaborative interface for AI-driven prompt engineering. PromptHive ensures subject matter experts (SMEs) remain central to AI-assisted content creation, providing a structured process to curate, refine, and validate instructional materials at scale. Designed to guide large language model (LLM) outputs with precision, PromptHive supports rapid experimentation, iterative refinement, and collaborative authoring, all while aligning with educational workflows. Users can customize generation parameters–including model selection (GPT-3.5, GPT-4, GPT-4o) and temperature–and experiment with both lesson-level and textbook-level prompting to fine-tune outputs for instructional effectiveness. PromptHive integrates with structured content formats (e.g., JSON), making it easily adoptable across educational ecosystems. We validate PromptHive through usability and cognitive load studies with subject matter experts, and demonstrate its effectiveness via a learning gain study with 358 students–showing performance on par with human-curated hints. This makes PromptHive a powerful tool for advancing scalable, high-quality, and human-centered educational content development.