Well, I'm not trying to make the case that recommendation algorithms are strictly unhelpful. It would be absurd to imply that there aren't many cases where they have provided a benefit to the end user. I just think that in many cases in my experience they have a bad signal to noise ratio to the point where it annoys me, because of the limitations of surreptitious data collection. Also, Youtube and Amazon don't allow you to turn recommendations off, which to me implies that in a choice between user preference and their business interests, it is clear which one is more important to them.

> That's exactly what these algorithms do. They try to find what YOU would find meaningful and fulfilling and they really do try to give it to you. When that means money will be exchanged they prefer one vendor (that has paid them) over another (which hasn't, or paid less), but does that really change anything ?

To me, this question is asking, "Is there a difference between a data collection algorithm whose primary goal is to help me, and a data collection algorithm whose primary goal is to profit off of me?" While there does exist a subset of transactions where the results would be the same, this whole article upon which we are commenting provides ten case studies of unintended harms or instances where the interests of the data collector and the interests of data collectee are not aligned. If you are looking for a more in-depth discussion illustrating some of the potential harms, I found Robert McChesney's Digital Disconnect to be a convincing first look at the problem, and there are some good documentaries on Thoughtmaybe that may be relevant as well, including Advertising at the Edge of the Apocalypse, Weapons of Mass Surveillance, and Adam Curtis's The Century of the Self.