What is artificial intelligence explainability?
A working definition from MIT Sloan
artificial intelligence explainability (noun)
A quality that enables users of artificial intelligence programs to understand and trust how models operate and make decisions.
Creating successful artificial intelligence programs doesn’t end with building the right system. Stakeholders must have confidence that the programs are accurate and trustworthy.
According to research from Ida Someh, Barbara Wixom, and Cynthia Beath of the MIT Sloan Center for Information Systems Research, artificial intelligence explainability helps by ensuring users that models are “value-generating, compliant, representative, and reliable.”
There are several reasons stakeholders hesitate to trust AI. Because AI is relatively new, there isn’t an extensive list of proven use cases. Models are often opaque — AI relies on complex math and statistics, so it can be hard for average users to tell how a model works, whether it is producing accurate results, and if it is ethical and compliant.
Models can produce biased results if trained on biased data, and they also “drift” over time, meaning they can start producing inaccurate results as the world changes or incorrect data is included and replicated.
AI explainability is an emerging field, and teams working on AI projects are mostly creating the playbook as they go, the researchers write. Organizations can start by identifying units and organizations that are already creating effective AI explanations, continuing to test the most promising practices, and institutionalizing the best ones.
Working Definitions: Artificial Intelligence
MIT Sloan's Working Definitions explore the words and phrases behind emerging management ideas.
Leading the AI-Driven Organization
In person at MIT Sloan
Register Now
Pro-worker AI, explained
Artificial intelligence can make workers more capable and productive, but only if leaders design and deploy it to augment human judgment.
5 things to consider when working with AI
Researchers at the MIT Initiative on the Digital Economy share the latest insights about getting the most from working with AI, such as personality pairing and reorganizing job tasks.
Balance AI innovation and risk with ‘minimum viable governance’
As organizations scale generative AI, traditional governance models prove to be too rigid or too loose. Minimum viable governance calibrates oversight to risk, enabling responsible innovation.
The surprising power of warmth in AI negotiations
In MIT’s international AI Negotiation Competition, “warmer” agents achieved better outcomes in negotiations with other AI agents.
Seeing real value from AI depends on being able to verify its outputs
A new paper explores how seeing economic value from artificial intelligence hinges on closing the gap between what AI can do and how humans can verify its outputs.
‘AI gravity’ is pulling you toward dependency. Here’s how to push back
AI systems hold the promise of competitive advantage, but they can usher in cognitive decline among workers, says MIT Sloan School of Management’s Eric So. Learn how to protect cognitive capital.
Heeding the pope’s call to ensure AI protects human dignity
Following Pope Leo XIV’s encyclical on AI, MIT Sloan professor emeritus Thomas A. Kochan argues that firms should partner with workers to ensure AI augments human skills and knowledge.
What senior leaders want to know about AI
Leaders are turning to MIT Sloan Executive Education to learn more about AI, including managing humans amid technological change and rethinking their relationships with IT departments.
What leaders still get wrong about AI
Organizations are struggling to succeed with AI. Research from the MIT Center for Information Systems Research shows common mistakes and how to overcome them.
How ‘learnrights’ would compensate creators for AI model training
Learnright laws would give copyright holders the exclusive right to license their content for artificial intelligence model training.