The social network routinely used surveys in the past to collect user feedback on topics including News Feed, the validity of news stories, friends’ reactions to their posts, improvements to Messenger, experiences with pages, the stand-alone Facebook Groups application (shuttered in August 2017), Ads Manager and Facebook Login and apps.
The two changes being introduced to ranking in Facebook’s News Feed algorithm are prioritizing the friends a user might want to hear from the most and prioritizing links they might consider most worthwhile.
Product managers Ramya Sethuraman and Jordi Vallmitjana and technical program manager Jon Levin said in a Newsroom post, “We’ve historically predicted who people might want to hear from based on signals like how often they interact with a given friend, how many mutual friends they have and whether they mark someone as a close friend.”
The social network is taking this one step further by surveying people to ask them which friends they are closest to, saying that it will examine patterns that emerge from those results—being tagged in the same photos, reacting to and commenting on the same posts, checking in at the same places—to inform its algorithm.
They stressed that the algorithm tweak does not mean that those people are the only people whose content will show up in News Feed, saying their posts will appear higher, and they added that Facebook’s prediction models will be continuously updated based on users’ interactions with their friends, promising further surveys and noting, “We also know that whomever you want to hear from right now may not be who you want to hear from in a year, or even a few months.”
Facebook has also been surveying people to learn which posts they felt were worth their time, and it is using those results to try to surface links that it predicts people will find worthwhile.
Sethuraman, Vallmitjana and Levin explained, “Similar to the above update, we use these responses to identify signs that someone might find a link worth their time. We then combine these factors with information we have about the post—including the type of post, who it’s from and the engagement it’s received—to more accurately predict whether people are likely to find a link valuable.”
Finally, they stressed that the impact on pages will be minimal, and that the changes are not being made to show more or less content from pages or friends.