For the health of society, all recommender algorithms have to be daylighted. For a start. Yes, the gamification will be terrible. So what? Is maximal advertising revenue for monopolies societally important? Every other efficient open market is heavily regulated. Accountability, transparency, fair play, information symmetry, prohibit self-dealing & conflicts of interest, tort, etc. Right now we can't even imagine appl…
1. the quality of the recommendations is the real competitive advantage. It's like saying that all the sports scouting agencies need to fully reveal their algorithm to finding top talent to ensure fairness, when that exact algorithm is the thing that makes them a top scouting agency.
2. The company itself doesn't really know how the recommender works.(source: I built explainable ML models for a living).
The inherent black box nature of any non-linear function approximator makes it near impossible to enable Explain-ability and gauge fairness for any system. In fact, I'd go as far to say that a sufficiently complex system of any kind quickly goes beyond human understanding. There are some decent proxies for it (SHAP, attention maps, gradient maps), but many recent papers raise questions on whether they really capture what we think they capture.
I think regulation of some form is certainly a good idea for tech ethics. However, it does appear that laws are being drafted by politicians, lawyers and social scientists. There is a glaring lack of qualified ML/CS researchers and practitioners on the implications of vaguely worded policy to their current processes and pipelines.
I was at ACM FAccT (ACM's fairness,accountability and transparency conference) this year and too many of the attendees seemed to be more concerned with the 'what needs to be achieved' rather than the 'how we can go about achieving it'. Terms such as fairness and accountability do not have widely accepted or principled meanings, which makes it near impossible to make a structured discussion around how anyone would go about implementing these in practice.
In conclusion, what you state is necessary, but we are very far from coming up with actually useful laws to implement it. Moving too quickly risks haphazard rules that do more damage than help.