Curation is my entire business and this article only casually mentions what I think is the most important point about why human curation works so well: trust. And, in particular, trust when things get weird.
When there's a clearly identified brand or person standing behind curation, it has a major effect on the audience and their response, versus a "blameless" automated job. Humans explicitly trust or distrust other humans, whereas trust in software or algorithms is either implicit or of dubious nature.
If I subscribe to a magazine, listen to a radio station, or attend a festival curated by a known figurehead, group of people, or a brand like "The New York Times", that I trust, I know things that I experience that are out of my comfort zone were likely designed to be there. If a news recommendation algorithm throws up something weird (and they always do, alas), I have no idea if it was being smart or just making a mistake.
Not just trust, but also social proof. Humans are hard-wired to find interesting what other humans seem to be interested in, regardless of whether it's a positive or negative interest.
I have no idea if it was being smart or just making a mistake.
When does a bold recommendation become a mistake? Can human recommendations be mistakes? (Both algorithmic and human recommendations are based on models, assumptions, and partial information.)
I would think, by strict definition, a bold recommendation would be one you doubt you'll like but that turns out well, and a mistaken recommendation would be one that turns out poorly. In other words, the value of a recommendation is not something you can evaluate at the time, before following the recommendation; instead, it's a measure of regret. (And by that vein, yes, human recommendations can totally be mistakes, too.)
On the other hand, people do tend to try to filter on the recommendations themselves before knowing their regret-value. People will subconsciously multiply the likelihood that something was a "bold recommendation made by design" by the relative status of the recommender.
Celebrities, leaders in your industry, people you're attracted to, etc. can get away with recommending all sorts of crazy things. Just because of their source, these things seem much less crazy. And this doesn't seem to correlate at all with how well they would know you or your tastes as an individual. Kind of a mysterious behavior, though it reeks of evo-psych tribal status dynamics stuff.
Oh absolutely. I certainly have. And boy do you hear about it! Due to the trust people have in manual curation, mistakes are quite painful. This is one of the reasons I think people subconciously like manual curation. If you can blame and chastise the curator, that makes it more appealing.
If The New York Times runs some vile headline on their front page, it would do incredible damage to the brand. If Google News automatically ran a similar thing, fingers would get pointed at some blameless algorithm (like with the black people/"gorilla" auto tagging debacle last week) and we'd all shrug at how funny technology can be while remembering just how uninteresting Google News usually is at the same time.
When there's a clearly identified brand or person standing behind curation, it has a major effect on the audience and their response, versus a "blameless" automated job. Humans explicitly trust or distrust other humans, whereas trust in software or algorithms is either implicit or of dubious nature.
If I subscribe to a magazine, listen to a radio station, or attend a festival curated by a known figurehead, group of people, or a brand like "The New York Times", that I trust, I know things that I experience that are out of my comfort zone were likely designed to be there. If a news recommendation algorithm throws up something weird (and they always do, alas), I have no idea if it was being smart or just making a mistake.