人工智能与未来教育评价

Artificial Intelligence and the Future of Education Assessment

人工智能算法如何才能准确把握人与人之间交流时传递的深层次思想,是人工智能教育评价系统面临的最大挑战之一。这需要相关工作者对人工智能、动态对话以及统计分析有着更加深入的了解。在医学应用中,人工智能决策判断的有效性主要由“敏感性”和“特异性”两个统计指标进行衡量。虽然这些统计数据有助于了解总体情况,但却忽略了一个事实,即无论如何增加机器学习的训练数据,都无法消除判断的不确定性。因此,在教育应用中,必须考虑人与机器如何协同工作,以提高未来教育评价的效率,并为对话教育创造更多空间。文章着重分析机器学习、贝叶斯统计方法对改变基于人工智能的教育评价的可能性,强调未来教育评价最基本的问题是明晰人与机器所擅长的领域各不相同,人工智能在教学过程中提供自动、高效和准确的反馈,可以帮助学生实现自主学习和自我评价;而对于机器无法确定的决策判断,则需要教师的参与和干预。据此提出,将当前人工智能在医学诊断等领域的成功应用,拓展到教育评价中,是未来教育改革的必然趋势,将带来人(教师和学生)与机器之间的密切合作。其中,信任是这个过程中最重要的因素,要增强人们对人工智能教育评价的信任,就需对机器学习过程进行更全面的检测,并用更丰富的信息来判断特定结果的准确度。而准确度可能是未来教育评价技术中最为重要一个部分,其能够引发新的学校教育实践,并更有效地利用教师专业知识,同时也能促进自主学习、师生对话和互动。

One of the big challenges in using Artificial Intelligence(AI)for education assessment system is how to capture and represent the deep thoughts developed through communication between people engaged in learning.This requires a deep understanding of artificial intelligence,statistical analysis,and rich dynamic dialogue data.The effectiveness of AI judgement-particularly in medical applications-is typically measured as percentages of“sensitivity”and“specificity”.While these statistics are helpful in understanding the overall picture,they ignore the fact that machines have a high degree of certainty about decisions far from the threshold,but less certainty about decisions near the threshold.Furthermore,no amount of training data for machine learning can eliminate the uncertainty of judgment.Therefore,in educational applications,it is necessary to consider how humans and machines work together to improve the efficiency of future education assessments and create more space for dialogue in education.This paper focuses on the possibility of machine learning and Bayesian statistical methods to change the education assessment by the development of AI.It emphasizes that one of the key issues of future education assessment is to clarify that people and machines are good at different problem areas.AI can provide automatic,efficient,and accurate feedback to help students achieve self-directed learning and self-assessment,while the decision-making that cannot be determined by machines,teachers’participation and intervention are required.Based on this,it proposed that the currently successful application of artificial intelligence in medical diagnosis and other fields to be extended to assessment in education.It is inevitable that future education reform will need to bring close cooperation between people(teachers and students)and machines.Among them,trust is the most important factor in this process.To enhance trust in AI education assessment,it is necessary to conduct more comprehensive inspections for the machine learning process and use more abundant information to assess the accuracy of specific results.One of the most important aspects of future education assessment technology is that it can lead to innovative teaching and learning practices,make more effective use of teachers’expertise,and also promote self-regulated learning and the dialogue and interaction among teachers and students.