2020: A Critical Inflection Point for Responsible AI with Rumman Chowdhury

Explainability

Explainability in AI systems presents significant challenges, particularly when translating complex data science concepts to non-experts. Rumman Chowdhury highlights the difficulty of ensuring that explainability leads to true understanding, rather than just transparency 1. She compares it to legal documents that are fully explained but not understood by the average person. This gap often leaves users without agency, as they cannot act on the information provided by AI systems 2.

Fully explained, completely explained to not at all understood.

--- Rumman Chowdhury

Adversarial models and mimic models are promising, but their complexity can hinder practical application in larger organizations 2.

   

Governance

Governance frameworks are essential for managing AI tools responsibly, yet the abundance of principles can be overwhelming. Rumman Chowdhury notes that there are over 150 sets of principles, creating a challenge in operationalizing them effectively 3. She emphasizes the need for organizations to translate these principles into actionable governance structures.

How do we drive principles into action?

--- Rumman Chowdhury

At Accenture, Chowdhury's role involves creating practical solutions for responsible AI, which includes redesigning organizational infrastructure to support ethical AI deployment 4.