Opacity, where the internal reasoning of a model is: Among the unique characteristics of AI that make

Opacity is the characteristic capturing the difficulty of understanding how a model produced its output.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Data dependency, where system behavior is shaped heavily by the data used to train the model.

   Plausible but wrong: data dependency concerns reliance on training data, not explainability of a given output.
2. Speed and scale, where a single system can affect very large populations almost instantaneously.

   Plausible but wrong: speed and scale describes reach and velocity of impact, not interpretability.
3. Opacity, where the internal reasoning of a model is difficult for humans to inspect or explain. ✓

   Correct: opacity (the 'black box' problem) is the characteristic describing hard-to-interpret model reasoning.
4. Autonomy, where a system takes actions or decisions with limited real-time human intervention.

   Plausible but wrong: autonomy concerns independent action, not the interpretability of outputs.

How to remember it

Opacity is the characteristic capturing the difficulty of understanding how a model produced its output.

More Understanding the Foundations of AI Governance questions

• Ethics and bias risk: Which of the following is such a category commonly attributed specifically to AI →
• Transparency and explainability: Which responsible-AI principle does this most directly reflect? →
• So that accountability for AI decisions and oversight is: When establishing an AI governance program, why is →
• All 337 Understanding the Foundations of AI Governance questions →