AI-Assisted Grading with Gradescope

Gradescope is an assessment platform that leverages artificial intelligence (AI) to streamline grading and provide detailed feedback.

AI-Assisted Grading with Gradescope

Gradescope enhances student learning by assisting educators with assessing student work, whether handwritten or digital, with timely and meaningful feedback. Its capabilities include AI-assisted grading, flexible rubrics, detailed analytics, and more.

Key Features of AI-Assisted Grading

  • Answer Groups: This feature allows instructors to form and grade answer groups, instead of individual student answers. For some question types, Gradescope AI automatically forms suggested answer groups for instructor review.
  • Formatting Guide: Gradescope provides a formatting guide to help make suggested answer groups the best they can be.
  • Auto-Grading: Gradescope also offers an auto-grading that supplements the grading workflow, for online, programming, and bubble sheet assignments. Multiple-choice questions or clearly defined answers, are automatically assigned scores based on pre-defined criteria.
  • Various Supported Question Types: Four question types are currently supported: Manually Grouped, Multiple Choice, Math Fill-in-the-blank, and Text Fill-in-the-blank.

Access Gradescope

Gradescope is accessible via Blackboard, and through the Gradescope (gradescope.com) portal.

Login to Gradescope

Using AI-Assisted Grading

Gradescope is available to both experiences in Blackboard (Original and Ultra Course View).

Tutorial Description
Create an Assignment Outline This guide provides detailed instructions on how to create an assignment outline for AI-assisted grading.
Formatting Guide for AI-Assisted Grading General recommendations and tips for formatting assignments for AI-assisted grading.
AI-Assisted Grading and Answer Groups An in-depth guide to set the question type, how to form answer groups, and how to grade each submission individually.
Guide for Auto-Graded Programming Assignments A detailed overview to create a programming assignment, and use a combination of autograding and manual grading workflows.

Considerations

When using the AI-grading feature in Gradescope, there are several considerations to keep in mind:

  • Availability: AI-Assisted Grading and Answer Groups are only available on fixed-template PDF assignments and not available on any question or answer fields in Online Assignments.
  • Template / Assignment Outline: Ensure a blank template is uploaded, and identical to the one shared with students. The template is used as a negative to extract student ink from their submissions. A template can be created in an external application like Microsoft Word.
  • Space Between Questions: Within a template/outline, leave some space between questions, so that the region boxes within the outline does not overlap.
  • One Answer Area Per Question: Try to use just one answer area per question in a template/outline. Increasing the number of answer areas can result in many permutations of answer groups.
  • Question Types: Four question types are currently supported: Manually Grouped, Multiple Choice, Math Fill-in-the-blank, and Text Fill-in-the-blank. Each type has its own considerations for optimal results.
  • Answer Groups: The ability to form and grade answer groups, instead of individual student answers, is available for all question types. This allows instructors to grade groups of similar answers at once.
  • Assist the AI Assistant: As Gradescope's AI assistant learns to identify and group answers, treating them as batches. While the AI assistant automatically forms suggested answer groups for review, it is important to review the suggested groups and make adjustments as necessary.
  • Student Submissions: If students are taking photos of handwritten work, converting them to a PDF, and uploading the PDF to Gradescope, inform students to lay pages flat on a table and ideally use a scanning application. Students can scan documents with the Gradescope app, or use a native application available in with iOS (iPhone/iPad), or like Adobe Scan (login with UM credentials).