Popularity does not always win: pondering signals of relevance

A quick post-omelette thought…

Over lunch today, I spent some time swirling around a YouTube time sink. Having watched a video about the Black Lives Matter protests, I noticed that my YouTube guardians recommended a Featured playlist on my home page. This included some, what seemed to be popular, vidoes on the theme of “Understanding Racial Injustice” (something about that title puts my teeth on edge). I applauded this. It was responsive to my behaviour (pretty much in real time) and topical.

Then I noticed that the selected videos were all pretty old. Their age is probably a good reason for their view count, a major reason for their inclusion. Given the date of my browsing, something current and topical seems an equally, if not more, powerful signal of relevance. Popularity, in itself, lacks an editorial sensibility as a selection criterion for inclusion.

Choice of expertise is also something I greatly appreciate in editorial judgement. It is a complex choice in curation decisions – a difficult feature to replicate algorithmically. The inclusion of Akala as an author of the one of the films is right on the money for me. The omission of someone like Afua Hirsch feels clumsy, however. This video, to my mind, would be a useful addition to the playlist. It is both current and adds to the thrust of the playlist. From a more obscure channel and with a lower viewer and subscriber count, it misses the cut.

I don’t profess to know the objectives of the playlist and the rationale for its presentation and population. It does strike me, though, that care should be taken in the use of popularity as a strong signal for choice of content in a curation of resources. It’s probably necessary but not sufficient. I still favour editorial judgement as a balancing factor and judgement borne from expsertise and experience. This is still a problematic trick for machines to pull off.


Leave a Reply