If we had to categorise interviewing as a science I think it would land squarely in the social sciences.

Difficult to impossible to reproduce results; practitioners with wildly different skills and biases; institutions with systemic biases; individual candidates who can vary wildly from day to day; cohorts that change over time and distance; a lay population with strong opinions.

We'd be better off selecting candidates randomly, but even this only kicks the can down the he road.