LTI Grading Caveats and Best Practices

[jpoulos]

Overview

As Instructure's Partner Integrations Specialist, I’m often asked how an LTI tool can return a grade to Canvas for an assessment that’s completed on the tool provider side. The answer varies based on a set of requirements the tool provider has. However, I’ve outlined the most common workflows I encounter to hopefully help others find the right place to start—keep in mind that a combination of these workflows can also be utilized, you don’t have to choose just one.

LTI-only Workflow

Requirements

This approach requires an LTI integration to be configured with an assignment_selection placement and a content item selection request. It should also be capable of executing the LTI Outcomes service.

Advantages

• API access not required
• Interoperable
• Can return LTI Launch URL as submission, allowing the tool to iframe a submission on the submission details page and within Speedgrader without the user having to leave Canvas
• Grades applied using the LTI Outcomes service will also mark the student as having completed the assignment; this isn’t the case when a grade is returned via Submissions API

Limitations

• Teacher must manually configure each external tool assignment using content item selection
• Student must launch each assignment so the tool can collect required information
  • As a result, a tool can’t have a series of assessments where grades are returned to separate assignments unless they provide the content as a common cartridge

Instructor Workflow

1. Create an assignment in Canvas
2. Choose “External Tool” as the Submission Type
3. Choose the tool from the list that appears in the modal
4. A properly configured tool allows the instructor to then choose a resource
5. Tool returns a content item message to Canvas with the Launch URL to that resource
6. Teacher saves and publishes the assignment

Student Workflow

1. Launch the assignment triggering a launch to the launch URL provided by the tool when the teacher configured the tool
2. Tool consumes the lis_outcome_service_url and lis_result_sourcedid for the student X assignment combination and stores it
3. Student completes the assessment
4. Tool either automatically grades the student, or waits for teacher to grade student
5. Grade is returned to Canvas via LTI Outcomes service
6. (optional) Using the LTI Outcomes service, the tool can also return a piece of plain text, a basic URL, or even an LTI Launch URL. This will be attached to a student submission object in Canvas, and it’ll be visible on the student submission page and in Speedgrader

LTI + API Approach

There are two flavors of this workflow. One uses the LTI Outcomes service for grading, while the other uses the Canvas submissions API.

Requirements

This approach requires an LTI integration to be configured with at least a course_navigation placement. It also must use OAuth2 to obtain teacher and student tokens.

Advantages

• Allows a single launch point from Canvas (usually from course navigation) where a student can complete many assessments and have the tool provision assignments within Canvas on the fly and grade them as needed
• Less manual work for the teacher
• Organization of assessments can be offered either on the tool side, or by leveraging the Modules page in Canvas
• Content-Item not required
• Tools can create external tool assignments and use the API to simulate a student launching the tool in order to obtain data required for the LTI Outcomes service

Limitations

• Requires OAuth2 to obtain either a Canvas API token for teachers (to grade) and students (to optionally submit)
• Requires heavy usage of Canvas specific API’s, so is not interoperable
• If grading is achieved via submission API, submissions are not marked complete after a grade is applied. The tool must also submit to the assignment using the submission API to mark complete. Furthermore, only plain text or URL’s can be provided

Instructor Workflow

1. Instructor launches the tool from course navigation or elsewhere and assigns the available work to students, or the tool can do this automatically
2. Tool kicks off OAuth2 if it needs to obtain or refresh a teacher token
3. Tool creates external tool assignments using the Assignments API. These assignments should have an LTI Launch URL that has query parameters for the assignment, or the path that launches directly to the assignment, e.g. https://some_launch.com/assignment/<tools_assignment_id>; or https://some_launch.com?assignment_id=<tools_assignment_id>;

Student Workflow

1. Student launches the tool from the course navigation button
2. Tool kicks off OAuth2 if it needs to obtain or refresh a student token

1. Student picks the assignment within the tool and completes it
2. Tool grades and submits the assessment to Canvas
   • Option 1: LTI for grading
     • Tool uses the sessionless launch API as the student to launch the tool
     • Tool consumes the lis_outcome_service_url and lis_result_sourcedid for the student X assignment combination and stores it
     • Tool uses the IMS Outcomes service to return a grade and an LTI Launch URL to their submission
   • Option 2: API for grading
     • Tool uses the Submissions API as the teacher to grade the student
     • (optional) Tool uses the Submissions API as the student to submit so the teacher can verify students completed the assessment

NOTE: The process can be further automated/simplified by obtaining an admin token with the “become other users” permission. This token can then be used to make API requests on behalf of teachers and students by appending the as_user_id=<student/teacher user id>. However, this can be problematic for high numbers of assignments or students as you’ll likely experience throttling issues.

I always like to hear about creative ways for integrations to grade and submit assignments. How has your integration combined API and LTI to grade and submit to Canvas?