Let Real People Grade the Tests
Just because something is possible does not mean it is a good idea. But companies developing artificial intelligence programs have made a great deal possible, and too many government entities looking to jump on the bandwagon are failing to ask whether the ideas are good.
Approximately 75 percent of state testing for Ohio students is now graded entirely by computer. For those little fill-in-the-bubble tests, no problem. But state school board members were told recently computers grade the entire test — including the long answers and essays — using artificial intelligence from American Institutes for Research.
“The motivation is to be as effective and efficient and accurate in grading all these things,” said State Superintendent Paolo DeMaria. “The research is really compelling. Not only is it being used in this setting, but even college professors are having AI grade their essays.”
While it should be no surprise that some college professors have jumped on the latest trend to save themselves time and effort, it should be a concern to Buckeye State parents that their childrens’ test scores depend on a computer program understanding and accepting their answers.
This brings teaching to the test to a whole new level. The change was brought to the attention of a wider audience only after several school districts noticed their fall 2017 third-grade reading test scores included a higher-than-expected number of answers that received zero points. Teachers have received little guidance on coaching students to please the computer.
Ohio Department of Education officials are hanging their hats on the fact that, when scores for approximately 1,000 questions were appealed by school districts, humans who rescored the questions changed the score for only one. (That might make quite a difference for one child. Thank goodness the district appealed.) They also claim using the programs will save money … someday … in the long run.
Children prone to unconventional and creative thinking might have been able to count on a human scorer to give credit to an answer that did not fall within parameters set by a computer program. There is too much at stake to leave the assessment of such young minds up to an artificial intelligence program.