Machine Learning Perspectives

Machine learning perspectives are common enough in the public dialogue that they don’t bare much repeating, but on a high-level, they center on:

• Bias and fairness
• Hallucinations / false information
• Transparency
• Various distopian visions, especially centered on automation displacing fundamentally human endevours

These are grounded in the recent surge in LLMs and generative AI. However, more broadly, machine learning models tend to operate as black boxes, and to make complex inferences without a clear basis of what leads to those inferences. Early AI algorithms might, for example, recognize photos of airplanes simply by looking for blue sky. In education, this sort of correlational inference can, for example, fail to identify overperforming students from underperforming groups and vice-versa (broadly defined; for example, looking at surface features like zip code or college pedigree), and otherwise, lead to bad decisions in less common cases.

Think about how to structure such a project through an AI ethics lens, as well as how these principles apply specifically to education.

A quick web search will bring up unlimited articles like these:

• Wikipedia article
• UNESCO Recommendations
• Nature article