The NYT unveiled a “For You” option on mobile last week.
A few days later I heard Anna Coenen, an NYT data scientist, present at the NYC Media Lab. She said the Times is starting cautiously by asking clients to pick topics.
Apple and Google have gone further surfacing personalized news streams. The results can be both amazing and underwhelming or even comical. It’s a really hard problem.
While it may seem easier than building a self-driving car, what makes it hard is that previous readership is often not a good predictor.
Google correctly anticipated I’d be interested in articles about machine learning, Brooklyn apartment prices and a body rolled up in a carpet near the Harlem Starbucks.
But having read articles about Jeffrey Epstein doesn’t mean I want more. He’s dead. I may have read about John Legend, but I don’t need to read more.
Google does let me hide the story or indicate that I’m not interested in John Legend as a topic or CNN as a source.
We are in the first inning with personalization, but I’m guess that we may find readership less helpful in predicting interest than search or other data, such as where you live or went to school or work.