I would argue that your point is semantics, but even so you still have a choice of whether or not to watch the recommended more "engaging" material. It doesn't change the overall point of my statement.
I'd say it's quite a different point. My own experience has been that the recommended "engaging" material is something in the same genre as whatever I just saw, but with a clickbaitier title, flashier thumbnail, and overall poorer informational quality. It's the different between saying "I see you enjoy sandwiches, maybe you would also enjoy salads or a plate of sushi" and "I see you enjoy sandwiches--here's a candy bar, an off-brand soda made with high-fructose corn syrup, and a carton of cheap grocery store ice cream."
The semantics argument I was pointing out was in regards to "broader" vs "engaging". That's not what my statement was about, it was that no matter what the algorithm recommends to you, you still have the choice whether or not to watch it. The point you are making is purely anecdotal as I assure you the neural network is not simply showing you
>same genre as whatever I just saw, but with a clickbaitier title, flashier thumbnail, and overall poorer informational quality
You can keep telling yourself that you have a "choice", but in the end we all are just humans, with quite predictable behavior. Bias selection of content is since forever one of the more effective ways of shaping opinion. Politics is fighting hard on that front for a reason. For the first time ever are some very few algorithms selecting content for millions of people, with apparently little human oversight. Yes, this should worry us. Simply assuming the results of those will benefit mankind, especially in the long term, would be foolish. It's not quite exactly like the usual ai safety paperclip scenario, but by now it should be very obvious that optimizing watch-time, even with current "ai", comes with significant unintended side effects / drawbacks.
It doesn't gradually introduce broader material, it gradually introduces more "engaging" material.