Wevideo
2024
Helping teachers create higher-quality video questions
Designed an AI-powered workflow that generates Bloom's Taxonomy-aligned comprehension questions, reducing teacher effort while improving the quality of video-based learning.
Overview
Helping teachers create higher-quality video questions in seconds with AI
Teachers spend a significant amount of time creating questions for educational videos. Many default to simple recall questions because writing thoughtful comprehension questions takes time they don't have. We explored how AI could reduce that effort while keeping teachers in control of the final learning experience.
The Impact
Following the launch of AI Interactions, usage of WeVideo's Interactivity product continued to grow, with active educators increasing by 76.13% and active students by 75.27%. While multiple factors contributed to this growth, the feature became part of a broader investment in expanding interactive learning experiences.
Problem
Creating meaningful questions shouldn't take longer than teaching
Teachers often create comprehension questions after recording or assigning videos, but writing thoughtful questions requires significant time and mental effort. Because of packed schedules, many settle for simple recall questions or skip assessments entirely. We needed a way to reduce the effort without removing teacher ownership of the learning experience.
Creating interactions involved multiple manual steps repeated throughout each video, increasing both the time required and the cognitive effort needed to produce effective learning experiences.
Research
Understanding how teachers create assessments
To better understand how teachers created video-based assessments, I analyzed the existing interaction workflow and synthesized feedback from educator interviews, and cross-functional stakeholders. Rather than focusing solely on how long the process took, I wanted to understand where teachers experienced the greatest friction, what prevented them from creating higher-quality questions, and how AI could reduce effort without replacing instructional expertise.
I organized recurring themes into a series of "How Might We" questions, helping the team translate research insights into actionable design opportunities and align on the problems worth solving.
What we learned
The research revealed that the biggest challenge wasn't writing questions—it was the amount of time and mental effort required to create meaningful comprehension checks throughout an entire video. Teachers repeatedly watched videos, identified teachable moments, selected interaction types, and authored questions from scratch, making it difficult to create high-quality assessments consistently.
Three key opportunities emerged:
Reduce repetitive work by streamlining the interaction creation workflow and eliminating unnecessary steps.
Use AI to generate a strong first draft while keeping teachers in control through editing and personalization.
Improve learning outcomes by generating Bloom's Taxonomy-aligned questions that supported deeper comprehension rather than simple recall.
Ideation
Exploring the right balance between AI and teacher control
After identifying our key opportunity areas, I explored how AI could reduce the manual effort of creating interactions without taking away teachers' instructional expertise. Rather than treating AI as a replacement for question writing, I focused on designing it as a collaborative tool that generated a strong first draft while allowing teachers to review, edit, and personalize every interaction.
Early concepts explored where AI should fit within the existing workflow, how much control teachers should retain, and what information was needed to build trust in the generated content. This exploration ultimately shaped an experience that accelerated content creation while keeping educators in control of the final learning experience.
Testing interaction flows early
I evaluated multiple interaction flows to determine the most intuitive way to introduce AI into the existing authoring experience. These explorations focused on reducing unnecessary decisions, keeping editing within context, and ensuring teachers could easily refine AI-generated interactions before publishing.
Designs
Generate questions where teachers already work
Rather than introducing a separate AI workflow, I integrated question generation directly into the existing interaction editor. Teachers could generate interactions for an entire video or a specific timestamp without leaving the authoring experience, making AI feel like a natural extension of their existing workflow instead of another tool to learn.
Keeping AI within the editor reduced context switching, streamlined content creation, and allowed teachers to immediately review and refine generated questions.
AI generation is embedded directly into the interaction workflow, allowing teachers to generate questions without interrupting their existing authoring process.
Regenerate, edit, Bloom's Taxonomy dropdown.
Lessons
Designing AI that supports expertise rather than replacing it
One of the biggest lessons from this project was that teachers didn't want AI to replace their instructional judgment—they wanted it to eliminate repetitive work. Giving users control to regenerate, edit, and adjust Bloom's Taxonomy levels created greater trust than fully automated question generation. The project reinforced that successful AI experiences are often less about automation and more about helping experts move faster while staying in control.