I’ve never kept a big diary past about 6 January (so I know a lot about the early Januaries of my life)

– Brian Eno, A Year With Swollen Appendices

I’m sure you recognize the pattern: new year, new you. So you start something with great excitement. By the second week of the year, reality reasserts itself. Your attention wanders. You’re left with unused exercise gear, a gym subscription, and regrets.

Or, as in Eno’s case, a few pages in a journal.

Do it over and over and you’ll eventually have quite a collection. It’s better than nothing, but undesirably lumpy. Nobody needs piles of exercise clothes, more “Meet our new blog!” posts, or tomes about their early Januaries.

Now picture this happening at scale: lots and lots of people contributing content or metadata to a system, but unwittingly focusing on a single part of the space.

It happens.

Consider the “Popular Highlights” feature on Kindle. Kindle apps and devices let you highlight and annotate books. Popular Highlights shows you the sentences and passages that have been highlighted most. It’s a sort of popularity contest for a book’s content. You can easily add these highlights to your own copy. A classic positive feedback loop ensues.

I find the feature helpful. Often, the highlighted passages are indeed the most important or insightful. But there’s a catch: they’re unevenly distributed. Most books I’ve read on Kindle have more Popular Highlights in the first few chapters than in later chapters.

This is expectable. More people begin books than finish them. And even if they finish, by the time they’re done, they’re likely tired of highlighting. They power through, leaving later chapters unmarked. I don’t have data, but I suspect something like two thirds of highlights and annotations happen in the first third of the book. The early Januaries.

Note that this isn’t because stuff happening after “mid-January” isn’t interesting or useful. It’s happening because most people don’t pay attention after “week two.”

Three takeaways:

  • As a consumer of content in such a system, be aware that useful insights are likely getting buried. Pay special attention to stuff in the “later chapters.”
  • As an author of content, structure information so it delivers value upfront. Realize many folks won’t get to the end.
  • As the designer of such a system, consider mechanisms that compensate for the inevitable attentional dropoff. For example, Kindle could supplement Popular Highlights with “Key Passages” selected by an AI.

Whichever the case, realize that systems informed by user input at scale can produce uneven results. While they might be useful, these “contributions” can unfairly delegate valuable stuff to the long tail. Respond appropriately.