people are more influenced by an enjoyable recommender than an accurate recommender
Talking with Automation, Like Humans Do.
Austin Beer

Does their study come to this conclusion?

Test subjects were given “the same interface and dataset (a subset of Netflix), but each with a different recommender algorithm.” They found that the perceived quality had no relation to statistical quality, so perceived quality must be influenced by something else.

Can you attribute that to how “enjoyable” their experience was?

Aren’t there other factors that may have influenced perceived quality; for example, perceived sophistication based on the visual design?

Would you be interested in designing “a more natural, pleasurable, and humanistic conversation” with recommenders, even if that doesn’t make them more persuasive?