Breakout Learning is an interactive discussion-based learning platform. Students meet synchronously over video to answer discussion prompts based on an assigned topic, scenario, or case, and AI evaluates each student’s contribution to the discussion. Results sync back to the D2L gradebook.

Two assignment types are available. The peer-to-peer group discussion is the primary format. Solo experiences are also available.

 

Pedagogical Value

Students are incentivized to prepare. A live discussion with peers creates accountability that written assignments no longer produce reliably. Students complete the pre-work because they have to speak to it in front of people they know.

The assessment reflects what the student actually knows. Generative AI can produce a discussion post or a short essay in seconds, which has weakened written submissions as a signal of understanding. A student reasoning out loud, in real time, answering challenges they did not anticipate, produces evidence that cannot be outsourced to a model. This is authentic assessment: the student demonstrates comprehension in a format AI cannot complete on their behalf.

The discussion is human-to-human. Students talk to each other and deepen their learning while practicing essential career readiness skills like communication, collaboration, critical thinking, and creativity.

 

Benefit to large-enrollment courses

In a 40-person section, an instructor can facilitate discussion and assess it directly. At 200 or 400 students, the time required to run breakout groups, observe them, and grade participation with any consistency exceeds what a faculty member and a TA pool can supply. Most large courses fall back on written assignments, multiple choice, or participation credit based on attendance, because those are what scale.

Breakout enables the facilitation and evaluation of discussions at scale.

 

How the AI Works

 

What is the system doing?

The AI analyzes the transcribed content of student discussions and scores each student’s contribution against the rubrics attached to that experience.

What it evaluates.

Three rubric types are in use, at the instructor’s discretion:

• Discussion Quality Rubrics (default) score students on criteria such as how well they apply concepts, articulate a reasoned position with evidence, and engage with their peers’ ideas.
• Bloom’s Taxonomy Rubrics assess the Bloom’s level of cognitive work in a student’s contribution, based on specific learning objectives.
• Pass/Fail Rubrics render a binary result against pre-defined criteria. In a negotiation role play, for example, a pass/fail rubric might ask whether the student stayed truthful to the facts in their role sheet.

How it scores. Discussion Quality Rubrics produce a score from 0 to 4. Bloom’s Taxonomy Rubrics return a rating on the taxonomy’s levels (remembering, understanding, applying, analyzing, evaluating, creating). Pass/Fail Rubrics produce a pass or a fail. Instructors control how each rating maps to a percentage in the grade calculation, with the highest rating fixed at 100%. Every result comes with a detailed justification, and students see more than the number:

• The AI’s reasoning for the score
• The key arguments it identified in their contributions
• Direct citations from the transcript supporting those arguments

Instructors and students both have access to this record.

Scoring process.

Breakout uses GPT-5.1 to evaluate discussions, and continuously tests other models against exemplar discussions. Evaluations run in batches: for each batch of rubrics the model is asked for 9 independent scores, and the batch is retried up to 3 times (27 completions at most) until at least 7 usable scores have been collected. The collected scores are sorted and the final score is the one sitting just above the midpoint, a deliberate slight upward bias; zero scores are set aside unless they make up most of the results. If no clear consensus emerges, the system flags the rubric for investigation by our team.

Context. Assignment-specific information is injected into the evaluation, including learning objectives, key concepts, and the assignment description, so the AI assesses what is relevant to that particular discussion rather than everything students said. Before an assignment goes live, instructors typically review these learning objectives with our instructional designers to confirm they align with what is being taught in class.

The AI does not assign the final grade. It is one component of the score. By default it accounts for 30%, with the remainder from non-AI components: quiz performance (40%), attendance, meaning whether the student joined the discussion room (15%), and completion, meaning the duration the student remained in the session (15%). Instructors have full control to adjust these weightings or to exclude the AI evaluation entirely.

 

What information is it using?

For evaluation, the AI uses:

• The transcript of the discussion, attributed by speaker
• The learning objectives, key concepts, and assignment description supplied by the instructor
• The rubric definitions for that experience

Quiz responses, attendance, and time in session are recorded separately and are not part of the AI evaluation. They feed the grade calculation directly.

Student data is not used to train AI models. Breakout Learning does not use student information or session transcripts to train AI models, including when that data has been de-identified. Discussion data is used to produce the evaluation, feedback, and insights for that learning experience, and for nothing else.

Breakout complies with FERPA and with the EU–U.S., UK, and Swiss–U.S. Data Privacy Frameworks. Student data is not sold, shared with advertisers or data brokers, or used for marketing.

Compliance documentation is available at trust.breakoutlearning.com and the full policy at breakoutlearning.com/legal/privacy-policy.

 

Who monitors it?

Our team monitors results through an internal dashboard and investigates when issues surface. The consensus mechanism routes low-confidence rubrics to that team automatically.

Additionally, faculty see every score with its supporting evidence and the same justification the student sees, and can review evaluations before grades are released.

 

What happens when it is wrong?

A human is notified. If the AI is unable to provide a consensus-based evaluation, the session is flagged for human review. Solutions are handled on a case-by-case basis.

Errors should be reported. Send anything that looks like a systematic problem to support@breakoutlearning.com.

 

When and how often does it change?

Breakout Learning is continuously updating and improving its system. Any updates that fundamentally impact the instructor or student experience will be communicated.

 

What Training Materials Do We Provide?

What is Available for Building Custom Activities?

Building a custom discussion from an institution’s own case content, prompts, or video is currently done in partnership with our account management and instructional design team.

 

Want to learn more?

We welcome the chance to talk through what a lightweight “build your own module” guide could look like for content developers, if that would help reduce the back-and-forth on your end.

 

What is the Onboarding Process and Timeline?

Once a course is ready to move forward, your designated account manager will meet with the instructor to configure the class setup. The onboarding is usually done 2-3 weeks before the course is set to begin. Exact timeline depends on the term start date.

During this time the account manager will also share the faculty and student prep materials above, and confirms LMS integration is passing rosters and grades correctly before the assignment goes live.

A follow-up check-in call is scheduled after launch to review how the first assignment went and troubleshoot anything that comes up.

 

Faculty’s Role & Self Service

Instructors introduce Breakout using the materials we provide.

If an instructor would rather we run a live session with their students directly, our account management would be happy to discuss this option further.

Within their LMS, faculty can independently adjust assignment dates and grade settings even after an activity is published. These live in the same configuration screens used during initial setup.

Changes to the discussion content itself currently go through our account management and instructional design team rather than being directly editable by instructors.

 

Scaling Training & Sustainability

All of the core materials above are static assets designed to be reusable, not delivered one-off per class.

The Student Onboarding Guide and self help resources can be uploaded into a “Getting Started with Breakout” module in D2L for students and faculty to review.

 

Accessibility and Accommodations

We take accessibility concerns very seriously and welcome feedback from faculty. Our team is committed to addressing any necessary accommodations to ensure an accessible experience for all users.

Breakout Learning is fully compliant with WCAG 2.2 Level A and AA standards. Additional information and accessibility documentation can be found on our Trust Center.

 

Get Help

Account Manager- Eduarda Vieira eduarda.vieira@breakoutlearning.com for any specific questions, course or content setups, and course management

Support Team— support@breakoutlearning.com for questions, access issues, or to be connected with our tam

Knowledge Base — for detailed articles on using the instructor portal and reviewing results.