The term "algorithmic" in this context is frustratingly, but obviously, not the general CS usage. What's meant, more or less, as indicated by the "addictive",

The term "algorithmic" in this context is frustratingly, but obviously, not the general CS usage. What's meant, more or less, as indicated by the "addictive", are feeds that: 1. Optimize for user engagement. 2. Are opaque to the user. 3. Are not controlled by the user. A sort by a field does not constitute such an "algorithm". One might argue that even having a default to sort by "most views" or "popularity" is a gray zone, but I'd say they differ per #2/3 in that the data used to order the items are completely visible to the user, and they are making an affirmative choice to sort by that data. I would prefer the use of terminology such as "metric-driven/metric-independent", "user-opaque/user-transparent" and "user-controlled/service-controlled". I would also prefer we refer to "predictive models" rather than algorithms, as that immediately implies the presence of some likelihood/metric being optimized for (by the service, not the user) in an almost certainly opaque or incomprehensible way. That said, I think for now we're stuck with "algorithm" having a very overloaded lay definition and just have to play ball rather than split hairs. There's too much at stake.