Can Artificial Intelligence Grade Essays Effectively?

Science-fiction is filled with terrifying "what-if" accounts of the rise of artificial intelligence, but one alarming scenario is already playing out in schools across America.

According to Vice Media's tech-focused website Motherboard, at least 21 U.S. states currently rely on automated systems to score student essays on certain standardized tests.

Only three of those states also require a human to score every test; the rest randomly select tests for human confirmation. The systems are programmed to determine grades by patterns, not by actually understanding what they are "reading."

If an essay fits a pattern found in previous successful essays, it gets a good grade. Computational linguistics Prof. Emily Bender of the University of Washington says that unless something changes, the flaw will grow, and it will lead to certain linguistic and demographic groups being favored over others.

"The problem is that bias is another kind of pattern, and so these machine learning systems are also going to pick it up. And not only will these machine learning programs pick up bias in the training sets, they’ll amplify it.”

In a test, Motherboard submitted nonsense essays written with sophisticated vocabulary. They passed.

The truth about artificial intelligence: