Scholarlink

<p>As artificial intelligence reshapes educational environments, it carries powerful epistemic assumptions—that knowledge is discrete, quantifiable, and best optimized through speed and scale. This paper examines a consequential semantic slippage at the heart of this transformation: the term "deep learning," which originated in Marton and Säljö's (1976) landmark distinction between surface-level and deep-level processing and was subsequently expanded by educational theorists to encompass sustained reflective inquiry, ethical attunement, and self-cultivation, now overwhelmingly signifies layered neural networks optimized for classification and prediction. Situating this shift within a broader pattern whereby engineering disciplines appropriate and narrow humanistic concepts through metaphor and simulation (Simon, 1996), I argue that the displacement of humanistic by computational "deep learning" in educational discourse is not a case of benign polysemy but a form of epistemic substitution with material consequences for what gets funded, measured, and valued in education. Drawing on Bergson's philosophy of duration and Barad's onto-epistemology, I contend that this substitution risks foreclosing dimensions of knowing—affective,embodied, relational, and temporally extended—that are irreducible to algorithmic logic and constitutive of the knower's being. I engage with recent scholarship on formative epistemic injustice (Smith, this volume), epistemic consolidation (Tanchuk, this volume), and the threat AI poses to intellectual character development (Pritchard, this volume) to show that the stakes of this displacement are at once personal and structural, ontological and political.Rather than rejecting AI, I advocate a both/and approach that cultivates AI literacy while preserving pedagogical practices—including contemplative pedagogy, slow reading, and ritualized aesthetic engagement—that resist algorithmic capture. I conclude by arguing for an educational ethos that reclaims "deep learning" for the formation of persons and that resists reducing education to informational throughput or behavioral optimization.</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

Rethinking Deep Learning in Education: Toward Humanist Onto-Epistemologies in an Age of AI

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Rebecca Taylor

摘要

As artificial intelligence reshapes educational environments, it carries powerful epistemic assumptions—that knowledge is discrete, quantifiable, and best optimized through speed and scale. This paper examines a consequential semantic slippage at the heart of this transformation: the term "deep learning," which originated in Marton and Säljö's (1976) landmark distinction between surface-level and deep-level processing and was subsequently expanded by educational theorists to encompass sustained reflective inquiry, ethical attunement, and self-cultivation, now overwhelmingly signifies layered neural networks optimized for classification and prediction. Situating this shift within a broader pattern whereby engineering disciplines appropriate and narrow humanistic concepts through metaphor and simulation (Simon, 1996), I argue that the displacement of humanistic by computational "deep learning" in educational discourse is not a case of benign polysemy but a form of epistemic substitution with material consequences for what gets funded, measured, and valued in education. Drawing on Bergson's philosophy of duration and Barad's onto-epistemology, I contend that this substitution risks foreclosing dimensions of knowing—affective,embodied, relational, and temporally extended—that are irreducible to algorithmic logic and constitutive of the knower's being. I engage with recent scholarship on formative epistemic injustice (Smith, this volume), epistemic consolidation (Tanchuk, this volume), and the threat AI poses to intellectual character development (Pritchard, this volume) to show that the stakes of this displacement are at once personal and structural, ontological and political.Rather than rejecting AI, I advocate a both/and approach that cultivates AI literacy while preserving pedagogical practices—including contemplative pedagogy, slow reading, and ritualized aesthetic engagement—that resist algorithmic capture. I conclude by arguing for an educational ethos that reclaims "deep learning" for the formation of persons and that resists reducing education to informational throughput or behavioral optimization.