A new predictor model is being developed. If the model shows significant changes in its predictions with slight variations in the input data, how would you classify its sensitivity?

The model would be described as highly sensitive or unstable, because its outputs shift noticeably when the input is only slightly altered. This means the predictor reacts strongly to small changes in data, which can signal a lack of robustness. In practice, a highly sensitive model can be risky because small errors or noise in the data can lead to large prediction swings. For example, if changing a single pixel in an image causes the model to switch its classification from “cat” to “dog,” the model is too sensitive to that pixel’s value. Such sensitivity is usually undesirable, as it can reduce the model’s reliability on real‑world data.