Data & AI · Peer-reviewed study
“Everyone wants to do the model work, not the data work”: Data Cascades in High-Stakes AI
Nithya Sambasivan and colleagues · 2021
Interview-based research examines how undervalued data work and unresolved data problems can lead to compounding failures in high-stakes AI projects.
The principle
What the source contributes.
Data collection, documentation and quality work are part of system design, rather than a preliminary task to rush through.
Our interpretation
How it informs our approach.
Informs evidence-readiness checks, source provenance and the decision to improve data collection before selecting analytical models.
This connection is MereSimplicity’s editorial interpretation. It is not a statement of endorsement by the authors or evidence of our product’s performance.
Assumptions & limits
Where the claim stops.
Qualitative findings from particular practitioners and settings do not predict a failure rate for other systems or validate MereSimplicity’s quality controls.
Original source
Trace the publication.
- Title
- “Everyone wants to do the model work, not the data work”: Data Cascades in High-Stakes AI
- Authors / institution
- Nithya Sambasivan, Shivani Kapania, Hannah Highfill, Diana Akrong, Praveen Paritosh, Lora Aroyo
- Year
- 2021
- Publication type
- Peer-reviewed study
- Journal / resource
- CHI Conference on Human Factors in Computing Systems · CHI ’21
- Publisher / issuer
- Association for Computing Machinery
- Sources checked