The reality is there is a great deal of overlap of races and socioeconomic status. High crime areas for instance tend to be not white.

I've been involved in a business that was sued for discrimination simply because the target market was low income. Which happened to mean the majority of customers were not white even though we didn't track any customer demographics. They won the suit based simply on that fact. Literally zero proof of any bias occurring. They couldn't even come up with how the business were exploiting them and what the benefit was to the company. Simply that there were a disproportionate amount of customers that are not white is good enough.

The DA(district attorney) kept asking for our racial data and would never accept that we didn't have any. I had to sign so many things saying that I was indeed providing all the data we had and that in fact we did not keep racial data on customers.

Socioeconomic geolocation bias of algos.

Eventually you will start locating places that have racial bias. For instance a heavily ethnic restaurant. That means its customers will be by default mostly a minority. You are now tracking and exploiting minorities/protected class.

See my other comment on this comment thread for a personal anecdote.

You awoke a question:

What sort of dark patterns are we (the public) _NOT_ seeing in algo behaviours, and further - how can Dark_Patterns be mapped.

ELI5 is there anyone working in this problem space thats open to speak without NDA issues/state-actor repercussions....

Whats the nomenclature in this space to address the data? And where is the data?