Number 2 pencils and fill-in-the-bubble are out, writes Matteo Wong in The Atlantic. Standardized tests are going digital. In a few years, artificial intelligence could make it possible to replace multiple-choice questions with "more creative and interactive questions," while still providing rapid, low-cost scoring.
"Artificial intelligence is now as good as a human at evaluating a standard five-paragraph essay and giving feedback on its logic and persuasion," writes Perpetual Baffour, research director of the Learning Agency Lab.
In tests, the best-performing algorithms achieved a 75 percent accuracy rate, comparable to the human readers. "These AI models can identify and evaluate the lead, position statement, supporting claims and evidence as well as a human," Baffour writes. "They also were able to evaluate how well a student organized an essay and developed arguments."
Many teachers don't assign much writing because it takes so long to grade students' work, she writes. AI can cut grading time in half. AI could encourage teachers to assign more writing, giving students more opportunities to develop their skills.
AI isn't quite as good as teachers in assessing middle-school writing, conclude Hillary Greene Nolan and Mai Chou Vang of Digital Promise. But AI is getting close to being an accurate grader and a useful writing coach.
In their study, AI and teachers agreed on the best and worst essays. In the middle, the teachers were tougher graders. In part, that's because teachers could "focus on the essay as a whole — the flow, the voice, whether it was just a summary or built an argument, whether the evidence suited the argument or whether it all made sense as a whole." AI is trained to read sentence by sentence.
AI sees use of transition words as evidence of good organization, Nolan and Vang write. Teachers can evaluate whether the transitions make sense.
AI also can be fooled by sophisticated vocabulary words that don't add meaning.
Like Baffour, Nolan and Vang see great potential for AI to relieve teachers' grading burdens and give students lots of writing practice once it's trained to "see essays more holistically the way teachers do."
Thomas Arnett, a senior research fellow for the Clayton Christensen Institute uses AI to shape his ideas into a first draft, which he can rewrite. It's a great time saver, he writes.
If AI can grade an essay, it can teach the student who will use it to develop their skill. The best students won't need to waste hours in a classroom for maybe an hour of actual instruction. The teachers can babysit the unmotivated and hope they inspire one or two to do more.