On Recommendations


Social media recommendations make sense. Someone new joins a network and you think “we want them to stay, let’s find them some friends.” From its inception, Twitter has always defaulted to offering you extremely famous – verified – people to choose from. LinkedIn’s transition from simply an online job social network to an “influencer engineer” gives you the opportunity to listen to the musings of Bill Gates and a bunch of other very famous people who’ve had success.

The issue here is less about the people who might find themselves never having access to the wisdom of very famous people who achieved success; it’s deciding by fiat which voices have merit and establishing a preemptive tone for the sorts of content you want to reward. There’s something problematic about a place that’s built upon sharing “what’s going on” that enables people to use it purely as a broadcast network; whereas others sharing deep thoughts about whatever is going on.

What’s worse are the perverse incentives that occur on far less dense networks that model themselves on this thinkfluencer strategy, and exist solely to mimic these models. Telegraph feeds work because we can see what strangers are thinking, and it brings us closer to people we’ll likely never meet. The veener of access imposes framing on experiences that might happen more organically without recommendations. Doing user research for guided experiences that presupposes what a user needs at scale, just reinforces our own value structure and encodes the harm baked into recommendation engines.