People are not algorithms. They are incredibly more complex and work by a mish mash of analogue, messy, unpredictable rulesets hugely affected by circumstance. To make proclamations like:
‘The likelihood of a person changing or maintaining a certain opinion is directly proportional to their exposition to it.’
.. is an extreme over simplifcation that allows a model to be built, and therefore the model is largely an approximation that will be used as part of another practice. (It is clearly not a great model, e.g. my massive recent exposure to Donald Trump has not left me reflecting his opinions)
To treat humans directly as models is a practice to remain extremely cognizant of, it has deep ethical implications which absolutely must be considered in parallel and as engineers we MUST be considering these ethical considerations as an integral part of the development and use of these models/algorithms.
It is a huge red flag that in articles like this the ethical application of models to not be at least referenced if not considered specifically for the algorithm/application.
The IZombie tv series did an interesting reversal of the “Once a zombie, always a zombie” statement.