NCPA - National Center for Policy Analysis

A Machine To Grade Essays

January 27, 1999

Beginning early in February, applicants to business schools around the country will have the two essays they are required to write graded by a human and an electronic robot called the E-rater. The machine was developed by the Graduate Management Admissions Council, which owns and sponsors the test. It is administered by the Educational Testing Service.

Until now, two experts have read each essay, assigning a grade ranging from a low of 0 to a high of 6. If the readers differed by more than one point, the essay went to a third expert who assigned a final score. Now, the E-rater will serve as the second reader.

Here is what the machine will look for in the typed essays:

  • Transitional phrases such as "for example" and "therefore" are recognized as indicators of a structured argument.
  • The use of synonyms, which indicate the writer has a strong vocabulary.
  • Good sentence structure based on accurate use of pronouns, verbs, etc.
  • The E-rater can catch errors in spelling and grammar that a human grader might miss.

The developers claim the E-rater will also reward logical arguments -- latching on to assumptions and alternative explanations.

To prevent the computer from dominating the procedure, if the electronic score differs from the expert's score by more than one point, the essay will be judged by a second expert before going to a final referee if the readers still disagree.

Although some critics point out that the machine cannot judge the creativity of writing, other experts predict the machine will someday have a wide impact on educational testing.

Source: William H. Honan, ""High Tech Comes to the Classroom: Machines That Grade Essays," New York Times, January 27, 1999.


Browse more articles on Education Issues