随着人工智能技术渗透至高等教育评估环节,澳大利亚各大学正面临如何平衡效率、公平性与数据安全的复杂局面。尽管商业机构自1998年起便开始开发AI评分系统,且Pearson报告称其专门训练的系统已为包括论文、简答和方程在内的3900万份回复评分 1,但学术界对完全依赖机器判卷仍持谨慎态度。2024年8月发表的一项研究显示,ChatGPT在标记生物科学本科论文时,平均给出的分数高于人类,且无法可靠地解释评分标准 1。这表明AI在准确性和提供针对性反馈方面仍不及人类教师,过度依赖可能掩盖潜在的教学问题并损害师生关系 1。
为应对这些挑战,不同高校制定了差异化的监管政策以规避法律和伦理风险。迪肯大学禁止教师将包含学生作业的大学数据上传到公开的AI中,以避免违反澳大利亚隐私法规和数据离岸存储风险 1。墨尔本大学允许教师使用批准的AI支持评估,但禁止完全依赖AI确定成绩,要求人类教师对最终成绩负责 1。此外,部分院校开始反思传统的数字化评分体系,新南威尔士大学在一些科目中试点取消数字分数,改为基于定义结果的胜任能力评估 1。
The integration of artificial intelligence into university assessment processes has sparked debate over the role of human educators, as commercial AI scoring systems have been under development since at least 1998 1. Pearson reports that its specifically trained systems have already graded 39 million responses, including essays, short answers, and equations 1. Despite this scale, current research indicates that AI still lags behind humans in accuracy, often providing inflated scores and lacking targeted feedback 1. A study published in August 2024 found that ChatGPT awarded higher average grades than human markers for undergraduate biology essays and failed to reliably explain its grading criteria 1.
Australian universities are adopting varied policies to manage these risks while maintaining academic integrity. Deakin University prohibits teachers from uploading student data containing university assignments to public AI platforms, citing concerns about violating Australian privacy laws and the risks associated with offshore data storage 1. The University of Melbourne permits the use of approved AI tools to support assessment but explicitly bans complete reliance on AI for determining final grades, requiring human instructors to retain responsibility 1. To address deeper structural issues beyond technical substitution, the University of New South Wales is piloting the removal of numerical grades in some subjects, shifting toward competency-based assessments defined by specific outcomes 1.
评论
还没有评论,欢迎留下第一条。