Erudite. Overview. Recorded video content is However, there is no - - PowerPoint PPT Presentation

▶

Dec 02, 2022 341 likes •474 views

Erudite. Overview. Recorded video content is However, there is no But behavioral viewing an increasingly prominent existing feedback loop for metrics can be predictive of mode of training. improving this content. quality and difficulty.

SLIDE 1

Erudite.

SLIDE 2

Overview.

Recorded video content is an increasingly prominent mode of training. However, there is no existing feedback loop for improving this content. But behavioral viewing metrics can be predictive of quality and difficulty.

SLIDE 3

That’s where we come in.

Erudite provides advanced video analytics for

nline educators.

SLIDE 4

Process.

Instructors and trainees sign up on the platform respectively. Instructors post video assignments that trainees then view. Instructors receive metrics about how the video was viewed and received.

SLIDE 5

What kind of stats?

We provide both time-wise stats throughout the video as well as aggregate summary stats.

SLIDE 6

Watch time (s) Average speed (?x) Pause time (s) Fullscreen time (s) Active time (s) Rewatch time (s) Muted time (s) Fullscreen paused (s)

No. of skips
No. of pauses
No. of rewinds
No. of tab switches

Completion Rate (%) Focus Score (%) Difficulty Score (%)

Much more data provided compared to other platforms, plus the ability to drill down and aggregate by trainees, assignments, etc.!

SLIDE 7

Use Cases.

Tracking quality of completion per viewer as well as on aggregate. Identifying what videos are more intuitive, well-paced and at the right difficulty. Diagnose what videos and viewers are facing difficulties.

SLIDE 8

Data-driven education.

With the advances in AI and data analytics, Erudite can add value even post lockdown.

SLIDE 9

Privacy.

We only share any data and stats with the course specific instructors. Our platform uses fully encrypted and secure communication. We don’t capture off-site markers about what users

pen or view.

SLIDE 10

Future developments.

Some things that I have in mind about where we can go from here.

SLIDE 11

Quizzes

By quizzing viewers, we can build models to predict how these viewing metrics translate to understanding. This also improves quality of data, because they have a more urgent incentive to learn the material well.

A/B Testing

By randomly assigning different versions of a video (e.g. from different YouTubers teaching the same concept) to random subsets of viewers, educators can utilize comparative analytics and what works well and why.

Segmentation

We can use these behavioral metrics to also find groups of viewers who have similar learning styles and hopefully take this a step further to proactively recommend what videos will work better.