controllability
What It Means
Controllability means humans can step in and override or redirect an AI system when needed. It's the ability to maintain human authority over automated decisions, ensuring the system doesn't operate completely independently without the possibility of human intervention.
Why Chief AI Officers Care
Without controllability, AI systems can cause business damage that's impossible to stop once initiated, creating massive liability and regulatory compliance risks. Many regulations require human oversight capabilities, and investors increasingly demand proof that AI systems have appropriate human safeguards built in.
Real-World Example
A credit approval AI system must allow loan officers to override automated rejections for qualified applicants who don't fit standard patterns, such as self-employed borrowers with irregular but sufficient income, preventing the loss of good customers due to algorithmic blindness.
Common Confusion
People often confuse controllability with explainability - being able to control a system is different from understanding why it made a decision. You can have an override switch without knowing the AI's reasoning process.
Industry-Specific Applications
See how this term applies to healthcare, finance, manufacturing, government, tech, and insurance.
Healthcare: In healthcare AI, controllability ensures clinicians can override automated recommendations, modify treatment protocols,...
Finance: In finance, controllability ensures human oversight over AI-driven trading algorithms, credit decisions, and risk assess...
Premium content locked
Includes:
- 6 industry-specific applications
- Relevant regulations by sector
- Real compliance scenarios
- Implementation guidance
Technical Definitions
NISTNational Institute of Standards and Technology
"property of a system that allows a human or another external agent to intervene in the system’s functioning; such a system is heteronomous."Source: ISO/IEC_TS_5723:2022(en)
Discuss This Term with Your AI Assistant
Ask how "controllability" applies to your specific use case and regulatory context.
Start Free Trial