When designing algorithms, managers need to recognize two major limitations: 1. Algorithms do exactly what they're told, and may be acting without the common sense of a human. 2. Algorithms are black boxes — they can predict the future, but not what causes it. s.hbr.org/2LX6JfF

There is something broken about this summary. Neither of these things is necessarily true. Also it seems to be simultaneously narrowing and broadening the definition of algorithm. Either way, 2 is definitely false. Many algorithms are not black boxes (eg gradient descent) and causal inference is an entire subfield of statistics. There can be a trade off between predictive accuracy and explanatory power but to suggest models (which it seems is the word that should replace “algorithm” in the OP) can never provide any explanation is just wrong.