AIAI (Artificial Intelligence for Advancing Instruction) Challenge - Phase 2

Classify instructional activities using multimodal classroom data. #education

$70,000 in prizes
Completed aug 2025
25 joined

Overview

Welcome to the Artificial Intelligence for Advancing Instruction (AIAI) Data Science Challenge! In this challenge, you will build models that classify instructional activities using multimodal classroom data.

Classroom observation videos provide valuable insights into a teacher's instruction, student interactions, and classroom dynamics. Over the past 15 years, their use in teacher preparation and the study of teacher quality has increased significantly. Classroom videos are also a common source of data for educational researchers studying classroom interactions as well as a resource for professional development. Despite this growth, using video at scale remains challenging due to the time and resources required for processing and analysis. In this challenge, you will build models to help automate classroom observation so that it can be offered at scale and inform future teaching.


Phases

There are two phases to this challenge:

  • Phase 1: Model Development (June 9th, 2025 – August 5th, 2025)
    Participants develop and refine their models using the training dataset, then generate predictions for the Phase 1 test set. Submissions are evaluated against the test set labels and ranked on the Phase 1 leaderboard. These scores contribute 25% to final prize rankings.

  • Phase 2: Final Scoring (August 7th, 2025 – August 11th, 2025) ← you are here
    Participants make one submission against a new, unseen test set. Submissions are evaluated against the Phase 2 test set labels and ranked on the Phase 2 leaderboard. These scores contribute 75% to final prize rankings.

Prizes

Phase 2 End Date:

Aug. 11, 2025, 11:59 p.m. UTC

Place Prize Amount (Phase 2)
1st $40,000
2nd $20,000
3rd $10,000

How to compete (Phase 2)

Note: You must have successfully submitted a solution in Phase 1 to participate in Phase 2.

  1. Click the "Compete!" button in the sidebar to enroll in the competition.
  2. Access the Phase 2 competition data via Globus and use your chosen model from Phase 1 to generate predictions that match the Phase 2 submission format. You are not permitted to retrain your Phase 1 model.
  3. Click “Submit” in the sidebar, and then “Make new submission”. You are permitted only one submission for Phase 2.

The challenge rules are in place to promote fair competition and useful solutions. If you are ever unsure whether your solution meets the competition rules, ask the challenge organizers in the competition forum or send an email to info@drivendata.org.

Phase 2 Dataset Updates (August 7th, 2025)

Please see the following for notes on the Phase 2 dataset and the effect on your submissions:

  • The video for 210.048_ELA2_Year3_190415_PART1 contains approximately 27 seconds of data beyond the submission format. You should ignore the final 27 seconds of the video file when generating your predictions and use the submission format as a guide for the timestamps to use in your predictions.
  • The submission format for 210.048_Math1Continued_Year2_20171115 contains only one line for one second of video, whereas the video file is 704 seconds long. There was an error with the annotation file for this video. The ground truth for this row contains entirely zero values, which you should use as your prediction for this video.
  • The submission format for 211.053_Math2_Year1_20180611 contains only one line for one second of video, whereas the video file is 657 seconds long. There was an error with the annotation file for this video. The ground truth for this row contains entirely zero values, which you should use as your prediction for this video.
  • The video for 211.053_Year2_Math1_20190419_Part2 contains approximately 19 seconds of footage not present in the submission format. You should ignore the final 19 seconds of the video file when generating your predictions for both audio and video labels, and instead use the submission format as a guide for the timestamps to use in your predictions.

We are thankful to the participants who brought these issues to our attention - please continue to let us know if you find any other issues with the dataset or submission format.


Challenge sponsor

Prize generously supplied by our friends from the Artificial Intelligence for Advancing Instruction (AIAI) project at the University of Virginia.

University of Virginia Logo UVA AIAI Logo

Image courtesy of the University of Virginia.