ExplainabilityTransparencyTrustworthinesscs.AI
xLP: Explainable Link Prediction for Master Data Management
Balaji Ganesan, Matheen Ahmed Pasha, Srinivasa Parkala, Neeraj R Singh, Gayatri Mishra, Sumit Bhatia, Hima Patel, Somashekar Naganna, Sameep Mehta
March 14, 2024
arXiv.org
2 citations
Key Takeaways
-
Multiple explanation formats improve user trust: The paper demonstrates that offering users different types of explanations for AI predictions—rather than a single standard explanation—allows people to choose formats they find most understandable and trustworthy, which is critical for enterprise adoption.
-
Explainability is essential for business applications: In enterprise settings like master data management, the cost of user time and the need for model trust make explainability a business requirement, not just a technical feature. The authors emphasize that users must understand why the system makes predictions before they'll act on them.
-
Combining multiple AI techniques improves explanations: The solution draws from diverse approaches including interpretability research, fact verification, path ranking, and neuro-symbolic reasoning, suggesting that responsible AI explanations benefit from integrating multiple methodologies rather than relying on a single approach.
-
Link prediction explanations require creative design: The paper highlights that explaining neural network predictions for complex tasks like identifying related records in databases requires thoughtful, creative design work—not just applying standard explanation techniques off-the-shelf.
Abstract
Explaining neural model predictions to users requires creativity. Especially in enterprise applications, where there are costs associated with users' time, and their trust in the model predictions is critical for adoption. For link prediction in master data management, we have built a number of explainability solutions drawing from research in interpretability, fact verification, path ranking, neuro-symbolic reasoning and self-explaining AI. In this demo, we present explanations for link prediction in a creative way, to allow users to choose explanations they are more comfortable with.