责任对齐的认知卸载梯度——以能动性为中心的生成式AI与远程开放教育融合框架
[土耳其]哈伦·塞尔皮尔
（阿纳多卢大学 课程与教学系，埃斯基谢希尔 26470，土耳其）
肖俊洪 译
【摘 要】生成式AI正在快速重塑学生的学习方式，这对远程开放教育的影响尤为突出。远程教学主要以数字工具为媒介，且学生的远程学习具有高度自主性，教师临场参与非常有限，因此，对于远程开放教育而言，“生成式AI是促进还是妨碍了远程学习？”成为一个关键性问题。本文认为这个问题的答案不是取决于学生在多大程度上把学习卸载给AI，而是如何设计这种卸载。当前研究主要着眼于AI的使用程度，即从把AI作为工具到作为学习伙伴使用的渐变过程，这样一来却忽视了随着AI在学习中承担了更加主动的角色，学生的思维过程归属也随之发生了变化。鉴于此，本文首先介绍了认知卸载梯度框架，以区分不同性质的卸载形式：从安全卸载（AI增强思维能力）到危险卸载（AI代替思维）。文章在这个基础上进一步提出责任归属这个独立的维度，把认知卸载梯度扩展为责任对齐的认知卸载梯度框架。这个框架有两个核心成分：责任漂移（学习者负起责任的层级从高阶的判断能力向遵守程序性规定递减）和分层级责任对齐（对照要求学习者必须承担的责任层级卸载相应的学习任务）。本文从分布式认知和元认知的角度出发，基于近年来远程开放教育的研究成果，阐述如何把责任对齐的认知卸载梯度框架应用于远程学习环境，并简述其对课程设计、学习支持服务和学术诚信的启示。责任对齐的认知卸载梯度框架重新审视了生成式AI与远程开放教育的融合，认为这种融合是一个如何发挥能动性的问题，而不是以效率作为衡量标准的问题。
【关键词】远程开放教育；生成式AI；认知卸载；学习者能动性和责任；自我调节学习
Cognitive Offloading Ladder with Responsibility (COL-R): A Framework for Agency-Centered Generative AI Integration in Open and Distance Education
Harun Serpil
(Department of Curriculum and Instruction, Anadolu University, 26470 Eskişehir, Turkey)
Abstract: Generative AI (GenAI) is rapidly reshaping how students learn, and its consequences are most immediate in open and distance education (ODE), where instruction is mediated through digital tools and learners study with high autonomy and limited instructor presence. A central question for the field is whether GenAI facilitates or hinders learning. This paper argues that the answer depends not on how much learners offload toAI, but on how that offloading is structured. Prevailing models conceptualizeAI use along a single continuum from tool to partner, capturing the degree of involvement while overlooking how responsibility for thinking shifts as AI assumes a more active role. To address this gap, the paper presents the Cognitive Offloading Ladder (COL), which distinguishes qualitatively different forms of offloading across a safe zone, in which AI amplifies thinking, and a danger zone, in which it substitutes for it. It then extends COL into COL-R (Cognitive Offloading Ladder with Responsibility) by introducing responsibility ownership as a second, independent dimension. Central to the framework are responsibility drift, the downward migration of accountability from higher-order judgment toward procedural compliance, and layered responsibility alignment, the deliberate matching of offloaded tasks to the forms of responsibility that must remain with the learner. Drawing on distributed cognition, metacognition, and recent ODE research, the paper illustrates how COL-R operates in distance settings and outlines implications for course design, learner support, and academic integrity. COL-R reframes GenAI integration as a problem of agency architecture rather than efficiency.
Keywords: open and distance education; generative AI; cognitive offloading; learner agency and responsibility; self-regulated learning
下载： 
责任对齐的认知卸载梯度.pdf
