Bad news for those who want to apply completely automated, industrial processes to writing assessment. A recent UK experiment with computerized marking progrmas has given them a failing grade. The Chartered Institute of Educational Assessors (http://www.ciea.org.uk/) recently put the computer marker through its paces by feeding it samples of essays written by the likes of Hemmingway and Churchill. The machine did not take kindly to the test and gave poor marks to some of the best writers of the 20th century. Not that this news is surprising, for computers are still fairly clumsy at natural language processing – just ask your grammar checker in MS Word. To imagine that highschool essays like those required in the British A-levels and CGSEs do not fall into the category of natural language is foolish. Foolish because in many of the rubrics designed to help teachers assess student writing at the highschool level – six plus traits, for instance – to demonstrate mastery, the student must break out of the mold of rigid formula. They must show spark by using metaphor, descriptive language, unconventional sentence structure, and all of the other things that set great writing apart from merely competent writing. How can a parser understand whether a metaphor works, or falls flat? How can a machine see the difference between stylish concision and lack of vocabulary?
Comarker skates by this problem because from the beginning we understood that putting teachers and markers out of work was antithetical to the educational enterprise. Academic writing tests our ability to generate meaning from the disparate and the abstract and that is preceisely what the computer cannot yet generate. So to generate feedback and authentic assessments and evaluations for students COMARKER relies on a more creative and intelligent system : human teachers. Revolutionary isn’t it.

A: to get to the “does not compute.”

Bad news for those who want to apply completely automated, industrial processes to writing assessment! A recent UK experiment with computerized marking progrmas has given them a failing grade. The Chartered Institute of Educational Assessors recently put the computer marker through its paces by feeding it samples of essays written by the likes of Hemmingway and Churchill. The machine did not take kindly to the test and gave poor marks to some of the best writers of the 20th century. Not that this news is surprising, for computers are still fairly clumsy at natural language processing – just ask your grammar checker in MS Word. To imagine that highschool essays like those required in the British A-levels and CGSEs do not fall into the category of natural language is foolish. Foolish because in many of the rubrics designed to help teachers assess student writing at the highschool level – six plus traits, for instance – to demonstrate mastery, the student must break out of the mold of rigid formula. They must show spark by using metaphor, descriptive language, unconventional sentence structure, and all of the other things that set great writing apart from merely competent writing. How can a parser understand whether a metaphor works, or falls flat? How can a machine see the difference between stylish concision and lack of vocabulary?

COMARKER™ skates by this problem because from the beginning we understood that putting teachers and markers out of work was antithetical to the educational enterprise. Academic writing tests our ability to generate meaning from the disparate and the abstract and that is preceisely what the computer cannot yet generate or assess. So to generate feedback and authentic assessments and evaluations for students COMARKER™ relies on a more creative and intelligent system : human teachers.

We still need beta testers, so contact us if you want to help us make a system that improves on the shortcomings of full automation.

  • Share/Bookmark