Researchers propose 'social life of data' for trustworthy AI
New paper argues trust in data should be social, not just technical...
Researchers Penny R. Atkins and Manish Parashar from Rutgers University and the National Data Platform (NDP) have published a paper proposing a novel approach to evaluating data trustworthiness in the age of AI-driven research. The work, titled 'Exploring the Social Life of Data: Finding Data You Can Trust' and published on arXiv (arXiv:2608.11395), introduces the concept of data-usage graphs as a new layer of scientific infrastructure.
The paper argues that as AI models increasingly integrate data from vast repositories, the challenge has shifted from finding data to finding data that can be trusted. The researchers propose using data-usage graphs to capture the 'social life' of datasets - tracking how they're used across publications, institutions, topics, and workflows. Their prototype implementation within the National Data Platform demonstrates how these social connections can provide context about a dataset's reliability and fitness for purpose, complementing traditional metadata and provenance information.
- Proposes 'data-usage graphs' to track dataset usage across research ecosystem for trust assessment
- National Data Platform (NDP) prototype demonstrates implementation of social trust metrics
- Argues popularity alone shouldn't determine trust - needs context of usage patterns and production quality
Why It Matters
Could revolutionize how researchers evaluate data quality in AI-driven scientific discovery