In a machine learning dataset, two features are recorded: temperature in Celsius and the same temperature in Fahrenheit. What does linear algebra say about the model that follows?

C. The columns are linearly dependent, so the fit is unidentifiable

One column is an affine function of the other, so it contributes no new direction and the weight can be split infinitely many ways. Predictions stay fine and the coefficients become uninterpretable, which is multicollinearity, and why 'the model works but the coefficients flip sign every refit' is a dependency problem rather than a mystery.