LearningSpace AI tools - FAQ

◾ Why use LearningSpace’s AI tools?

What to consider when subscribing to LS AI tools?

• Significant time savings for faculty and staff.seudo

• Quality and consistency improvements in evaluation workflows.

• Cloud-based service, incurring operational costs.

◾Can admins control which features use AI?

Yes. AI-related features can be globally toggled via the System Settings by users with the appropriate privileges, if the institute has an active AI License for LearningSpace.

◾Can AI features be disabled per program or assessment?

Yes. Each AI feature (e.g., Checklist Prefill, OEQS Prefill) has an individual on/off toggle, allowing granular control at the feature level.

◾Is customer data used to train AI models?

No. Customer data is not used to train any AI models.

◾How long does AI processing take (Transcription + Checklist AI)?

Actual times may vary slightly depending on audio quality and recording length, but overall performance is consistently fast. Both AI-generated transcripts and AI checklist suggestions process quickly:

• Shorter recordings & standard checklists: typically 10–30 seconds

• Longer recordings (1 hour+): usually under 3 minutes for full transcription + checklist processing

◾Is there an upper limit to how many speakers the AI transcription is able to identify?

For speaker diarization, LearningSpace automatically identifies speakers, with no strict limit on the number of speakers. The system will dynamically handle the number of speakers based on the conversation.
The number of participants can affect diarization accuracy. If there are too many speakers, the system might struggle to accurately distinguish between them, especially if voices are similar or overlap.

◾Are proprietary or third-party models used?

Yes.

• We use OpenAI models, hosted via Microsoft Azure

• We also use Azure's transcription services

◾Where is AI processing performed?

All AI processing takes place in the Microsoft Azure Cloud.

◾What type of data is processed by the AI features?

Educational assessment content only — specifically audio from video recordings, checklist rubric items, and open-ended question responses, associated with learner identifiers. No clinical data, ePHI, patient records, or financial data are involved.

◾Is the AI model a closed-loop system?

Yes. The AI features use fixed models that do not learn or adapt to customer data or ongoing use. No customer inputs are used to retrain or update the underlying model.

◾Is data anonymized or pseudonymized before sending it to the cloud?

We do not send user profile information directly.
However, if audio or content (e.g., checklist items or free-text responses) includes personal data, that is not anonymized.
It is the customer’s responsibility to manage or redact sensitive data if required.

◾Is the AI hosted on a specific cloud provider?

Yes. All AI processing takes place on Microsoft Azure. OpenAI models are accessed via Azure OpenAI Service, not directly through OpenAI's consumer infrastructure.

◾What logging or audit trails exist around data sent to AI components?

We do not log the content itself, but maintain audit trails and metadata logs for troubleshooting and monitoring.

◾How long is the data retained, and under what conditions is it deleted?

After receiving a response from Azure, all processed data (e.g., submitted text answers, special section questionnaires) is promptly deleted from Azure's environment.
Only the processed result is stored on the customer's server.
For more details, refer to Microsoft’s Trust Center.

◾What happens to already generated data if a trial or subscription expires?

Previously generated data (e.g., transcripts, AI suggestions) remain stored and accessible.
New AI-generated data cannot be produced until the subscription is renewed.

◾What happens to a transcript file if the related video file gets deleted?

Once a video is deleted, the related transcript will be deleted as well.

◾What happens if AI features are unavailable?

There is no interruption to assessment workflows. All checklist and open-ended question evaluations can be completed manually. The AI suggestions are an optional augmentation, not a dependency.