Overfit Kaggle Models Discouraged Data Science Competitors

May 1, 2017

In the 'The Nature Conservancy Fisheries Monitoring' competition on the data science platform Kaggle, several participants overfit their image classifier models to an unrepresentative validation dataset. This incident highlights the importance of trustworthy AI practices and guardrails for safe and secure AI development.

Project Cerebellum, a responsible AI governance initiative, invites you to explore its AI incident database and help prevent harm through mapping and managing such incidents via HISPI Project Cerebellum TAIM (Govern/Map). JOIN US.

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

MEASURE 2.6 — similarity 0.741, rank 1. TAIM detail and related incidents →
MAP 4.1 — similarity 0.735, rank 2. TAIM detail and related incidents →
MEASURE 2.10 — similarity 0.720, rank 3. TAIM detail and related incidents →

Alleged deployer: individual-kaggle-competitors
Alleged developer: individual-kaggle-competitors
Alleged harmed parties: individual-kaggle-competitors

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/61

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.