Their authoritarianism violates a fundamental principle of the very nature of algorithms.

These algorithms can't bias their input, and they can't manage output once it leaves the function. The input is simply the input. The ouptut is no longer part of the function. That's why they have to introduce functions to intentionally bias the program to generate the results they want, and why they have to have a backdoor into the function to make it bias results differently to generate the outcomes they want.

This is precisely what "Machine Learning Fairness" is about. The intentional corruption of data to generate preferred results.