If you you haven’t been following, well, trending new topics lately than you’ve probably missed the recent hoopla surrounding Facebook’s Trending Topics.
Here’s a quick recap: earlier this month Gizmodo reported on life inside Facebook’s Trending Topics followed by a subsequent report claiming Facebook contract employees routinely suppressed news stories of interest to conservative readers. Mark Zuckerberg replied directly to the allegations and Facebook published their 28-page internal Trending Review Guidelines document to provide additional transparency.
Justin Osofsky, Facebook’s VP Global Operations, outlined how Trending Topics works in a post on the same day. Here’s the gist: topics are first “surfaced by algorithm.” I love this eloquent vignette and I’ll discuss the importance of this in more detail shortly. The “surfaced” topics are then reviewed by the Facebook Trending Topics team who then provide several layers of curation and customization. Finally, algorithms are again employed to determine who receives which Trending Topics suggestions.
The revelation that Facebook currently relies more on humans and less on algorithms than presupposed has been jarring for some. Perhaps at the heart of this issue is perceived transparency. I think it’s more accurately, the perception that an implicit contract was broken.We make different implicit contracts with businesses, service providers and platforms and when those implicit contracts are broken, users become unsettled. I think one of the implicit contracts users have with Facebook is the unfiltered flowing of information.
What’s fascinated me through all of this dialogue is the dichotomy that exists. Many people are extremely alarmed by how much we are turning over to machines but when it is revealed that humans are significantly more involved at different layers of something like news curation than we previously realized, we become equally berate. I think above all else, this is the issue that is going to stay with us in the years to come. What’s the right blend between humans and algorithms? Where do we want humans more involved and where do we want algorithms and machines more involved. To get the mix right will involve experimentation.
Over time, human involvement will be driven by two factors. First, we’ll employ human capital where it still has a comparative advantage. Economists differentiate between absolute advantage and comparative advantage. There will be times when machines will have an absolute advantage, but humans might still have a comparative advantage. Secondly, employing human capital will also be a decision we make explicitly. I envision a myriad of places we’ll want to use humans over algorithms for any number of reasons. Our tastes and preferences evolve continuously and in non-predictive ways for example. Even as algorithms can decipher some of these trends, we may want humans involved. I think humans will continue to play key roles in settings with customer service elements. I also think customer service will evolve as we deploy more machines (a topic for a future post). Despite how little we sometimes trust other humans, we seem to trust machines and algorithms even less. Developing greater trust will be a key component in the further deployment of algorithms and machines.
I think Facebook’s process, as outlined by Osofsky in his post, is probably a realistic guide for our future. Algorithms will be used in tandem with humans. We’ll pass work product between humans and machines, each adding additional value along the way. Work flow will involve humans and machines at different points in the process. Work might begin when something is “surfaced by algorithm.” From there it might be passed along a designed work flow until a human intervenes and adds additional value, at which point it cycles back through machines. Over time, the jobs performed by both the human and the machine will likely change, but throughout, I think both will provide important value to work products.
