Paper Summary
Share...

Direct link:

Addressing Defensive Objections: Adversarial Examples in Automatic Essay Scoring

Fri, April 5, 4:20 to 6:20pm, Fairmont Royal York Hotel, Floor: Mezzanine Level, Confederation 3

Abstract

Page and Peterson (1995) identified three objections to the use of automatic essay scoring (AES): humanist objections, defensive objections and construct objections. Defensive objections consider the presence of bad actors submitting intentionally deceptive essays. The proposed paper looks at advances in computer vision and deep neural networks to reevaluate defensive objections and their threat to validity. The proposed paper looks at advances in computer vision and deep neural networks to reevaluate defensive objections. Based on advances in artificial intelligence and machine learning, the paper will identify adversarial examples as an important component of future AES validity research and integrate them into the interpretive arguments for AES derived scores.

Author