Scholarlink

Background <p>The growing societal presence of artificial intelligence (AI) necessitates that students develop not only conceptual understanding (CU) of AI mechanisms but also epistemic understanding (EU) of how AI knowledge is generated, constructed, warranted, and constrained. Responding to this need, this study proposes Enact-Examine-Extract (E<sup>3</sup>), an epistemic tool designed to scaffold upper elementary students’ CU and EU and evaluate its pedagogical effectiveness. Grounded in experiential and epistemological learning theories, E<sup>3</sup> engages students in three iterative phases: (i) enacting core AI mechanisms by simulating AI’s internal operating mechanisms, such as encoding features, applying decision rules, and updating classifications, to appropriate how intelligent systems process information, (ii) examining model behavior through direct interaction with real AI platforms to verify whether the enacted logic manifests in actual system outputs, and (iii) extracting essential principles of AI by reconciling where simulated logic and model behavior converge or diverge, thereby revealing the warrants, limits, and uncertainty of AI knowledge. A ten-week classroom intervention with 41 Grade 5 students in Hong Kong investigated how E<sup>3</sup> supported students’ conceptual and epistemic growth across five AI-related topics.</p> Results <p>Qualitative thematic analysis revealed progression in students’ CU, with many learners moving from multi-structural toward relational and, in some cases, extended-abstract reasoning across the instructional sequence. Five strands of EU were identified: the nature, purpose, source, knowledge-building, justification and evaluation of AI knowledge. Across activities, students’ EU did not develop in a linear, topic-bound manner; rather, epistemic dimensions were activated and refined interactively through engagement with different tasks. Over time, students increasingly articulated calibrated, pattern-based reasoning that recognized AI knowledge as probabilistic, data-dependent, iteratively constructed, and shaped by contextual and stakeholder considerations.</p> Conclusions <p>Positioning E<sup>3</sup> as an epistemic tool yielded substantial gains in both CU and EU of AI, supporting students’ informed, critical, and responsible engagement with AI systems. The findings contribute to ongoing epistemic discourse in AI education and highlight the need for K-12 AI curricula to couple concept building with sustained epistemic inquiry and transparency practices. The E<sup>3</sup> design and the proposed five-dimensional EU framework offer actionable principles for AI education, with relevance for curriculum design and classroom practice across diverse AI topics and educational contexts.</p>

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

Enact-Examine-Extract (E3): an epistemic tool to scaffold upper elementary students’ conceptual and epistemic understanding of artificial intelligence

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月16日

查看全文
  • Miao Yue,
  • Morris Siu-Yung Jong

摘要

Background

The growing societal presence of artificial intelligence (AI) necessitates that students develop not only conceptual understanding (CU) of AI mechanisms but also epistemic understanding (EU) of how AI knowledge is generated, constructed, warranted, and constrained. Responding to this need, this study proposes Enact-Examine-Extract (E3), an epistemic tool designed to scaffold upper elementary students’ CU and EU and evaluate its pedagogical effectiveness. Grounded in experiential and epistemological learning theories, E3 engages students in three iterative phases: (i) enacting core AI mechanisms by simulating AI’s internal operating mechanisms, such as encoding features, applying decision rules, and updating classifications, to appropriate how intelligent systems process information, (ii) examining model behavior through direct interaction with real AI platforms to verify whether the enacted logic manifests in actual system outputs, and (iii) extracting essential principles of AI by reconciling where simulated logic and model behavior converge or diverge, thereby revealing the warrants, limits, and uncertainty of AI knowledge. A ten-week classroom intervention with 41 Grade 5 students in Hong Kong investigated how E3 supported students’ conceptual and epistemic growth across five AI-related topics.

Results

Qualitative thematic analysis revealed progression in students’ CU, with many learners moving from multi-structural toward relational and, in some cases, extended-abstract reasoning across the instructional sequence. Five strands of EU were identified: the nature, purpose, source, knowledge-building, justification and evaluation of AI knowledge. Across activities, students’ EU did not develop in a linear, topic-bound manner; rather, epistemic dimensions were activated and refined interactively through engagement with different tasks. Over time, students increasingly articulated calibrated, pattern-based reasoning that recognized AI knowledge as probabilistic, data-dependent, iteratively constructed, and shaped by contextual and stakeholder considerations.

Conclusions

Positioning E3 as an epistemic tool yielded substantial gains in both CU and EU of AI, supporting students’ informed, critical, and responsible engagement with AI systems. The findings contribute to ongoing epistemic discourse in AI education and highlight the need for K-12 AI curricula to couple concept building with sustained epistemic inquiry and transparency practices. The E3 design and the proposed five-dimensional EU framework offer actionable principles for AI education, with relevance for curriculum design and classroom practice across diverse AI topics and educational contexts.