Algorithmic Feeds Trend to Homogeneity

Republished from the April 23, 2018 Newsletter.

The medium is the message and our mediums are shaping the messages we’re willing to share.

The allure of the algorithmic feed is that it shows more relevant information to individual users. However, it maps very poorly with how people WANT the services to act. For example, I want to see everything my wife posts on any of the services we’re both on. This is important because they are often about her, our daughter, or our family in general. As far as Content I’d Like to Consume this ranks up so so so high on the list that it’s shocking to me whenever the Algorithm doesn’t surface this content. Surely with all of that Big Data they could make that connection that, yeah, these two people enjoy each other’s company.

When you trust the algorithm you start to use the only signals you have from the service to inform that algorithm. If those are based on things like likes and retweets…well…guess what’s not going to show up as often? A tweet that is perhaps highly relevant, but didn’t get enough engagement to push it over the edge to appear in your feed.

Most of my followers and people I follow are (by design) people with sub-1000 numbers. They are likely to never see a tweet go viral and most tweets might go by “unliked”. Then they get suppressed and likely don’t want to tweet similar things again. Imagine a twitter that had more controls over what you see. Imagine if you could go back to a chronological timeline!