Robograder: Can Computers Grade Papers?

Too much thinking around educational technology reflects a lack of familiarity with how education actually works and what technology can and can’t do. A wide-eyed NY Times article marvels at a university level computer programming course in which student assignments are graded by artificially intelligent software. The article explains that the AI grader shows students where they made mistakes in their code and that students are satisfied with the automatic feedback. However, the integrated development environments (IDEs) computer programmers use have offered a roughly similar feature for years - a “type safe” function that flags possible errors and offers resolutions. Moreover, insufficient distinction is made between assessment types suitable for automated grading (multiple choice, formative) and assessment types well beyond the reach of AI (essays, summative and creative). After you read the NY Times piece, check out this more realistic explanation of how students easily trick AI into awarding “A” grades on written submissions. (Hint: keyword salad.) Also: this Motherboard piece explains that implicit bias is embedded in the patterns that automated essay graders scan for when assigning scores to written work.

Remember: Ed tech is for humans.