【英文】美国教育部教育技术办公室报告：人工智能与教学的未来（71页）

Issues related to racial equity and unfair bias were at the heart of every listening session we held.  In particular, we heard a conversation that was increasingly attuned to issues of data quality and  the consequences of using poor or inappropriate data in AI systems for education. Datasets are  used to develop AI, and when they are non-representative or contain undesired associations or  patterns, resulting AI models may act unfairly in how they detect patterns or automate decisions.  Systematic, unwanted unfairness in how a computer detects patterns or automates decisions is  called “algorithmic bias.” Algorithmic bias could diminish equity at scale with unintended  discrimination. As this document discussed in the Formative Assessment section, this is not a new  conversation. For decades, constituents have rightly probed whether assessments are unbiased  and fair.

Just as with assessments, whether an AI model exhibits algorithmic bias or is judged to  be fair and trustworthy is critical as local school leaders make adoption decisions about using AI  to achieve their equity goals. We highlight the concept of “algorithmic discrimination” in the Blueprint. Bias is intrinsic to  how AI algorithms are developed using historical data, and it can be difficult to anticipate all  impacts of biased data and algorithms during system design. The Department holds that biases  in AI algorithms must be addressed when they introduce or sustain unjust discriminatory  practices in education. For example, in postsecondary education, algorithms that make  enrollment decisions, identify students for early intervention, or flag possible student cheating  on exams must be interrogated for evidence of unfair discriminatory bias—and not only when  systems are designed, but also later, as systems become widely used.