by CJ Quines • on
above the noise
on quality and popularity
i remember a conversation in a friend’s discord server, a year and a half ago, where we discussed video games that have been breakout successes. there’s a wide range in terms of how much effort is put into the music or art or gameplay or story—which makes me feel that these things are uncorrelated with success.
conversely, you can put a huge amount of effort into a video game’s music and art and gameplay and story, and still end up with something unsuccessful. i think of games like ōkami or the floor is jelly which were loved by critics and audiences, and yet were still commercial failures. why? a lack of marketing, pretty much.
it annoys me that marketing and publicity has to matter at all. not only that, but it seems to be almost orthogonal to quality. there’s plenty of examples of games that were overhyped and sold lots of copies, only to be ultimately disliked by players; the clearest example for me might be no man’s sky, at least when it was first released.
another example. consider how one of the most popular posts on the mit admissions blogs is how to restore an old whiteboard. it’s not a bad post. it’s just that paolo’s also written so many other great posts that didn’t blow up, like leaning in. i dunno.
it’s a complaint as old as time, i suppose. in stardom and talent, a 1985 article published in the american economic review, moshe adler argues that “discussion-y” goods—goods that people get more utility from when they discuss it with other people—are more prone to bandwagon effects.
with the shift from mass media to social media, and the advent of widespread recommender systems, we could’ve believed that higher-quality content will become more popular, due to wisdom of the crowd or something. yet tons of people i know complain about the enshittification of search engines, content aggregators, and social media, which are all platforms underpinned by recommender systems. these systems are so ubiquitous that people just call them the algorithm!
the model i’ve found that feels the most right here is from azadeh nematzadeh et. al’s article how algorithmic popularity bias hinders or promotes quality. the takeaway is that, while it is possible to have recommender systems that consistently promote higher-quality content, it only happens within a relatively narrow set of parameters. it’s almost inevitable that these effects happen, i guess.
if there’s one bit of consolation, it’s that quality matters, at least in a baseline sense. matthew salganik et. al conducted an experimental study of inequality and unpredictability in an artificial cultural market, and concluded that success is at least partly determined by quality: “the best songs rarely did poorly, and the worst rarely did well, but any other result was possible.”