Real-world experiments and the Facebook controversy over mood manipulation

Big Data Evangelist, IBM

Manipulating somebody else's mood is fairly easy to do. Just show them something they like, and watch them smile. Or do something you just know will tick them off, and watch the sparks fly.

Mood manipulation is something that whole industries, such as media and entertainment, specialize in. That's why the big stink over the ethics of Facebook's attempts to influence moods through tweaks to its newsfeed algorithms is so deliciously absurd. Facebook's very business model depends on people liking the experience enough to return again and again. More than that, Facebook's revenue model depends greatly on: 

  • People clicking "like" on content they groove to
  • Sponsored ads
  • Continued user polls and requests for members to declare their favorite this and that
  • Third-parties distilling the sentiments, the likes and don't-likes expressed in the never-ending stream of chatter emanating from that world-crushing community.

Facebook has been a big "happy-happy-joy-joy" machine from day one. The ongoing brouhaha has simply exposed world culture's queasiness over its reliance on this one community to make or break their day. You want to know how powerful an influence Zuckerberg's baby has on the world's mental climate? Just consider all the real-world family and friend relationships that have broken up or been severely strained by something said (or not said) on Facebook. 

Images courtesy of Openclipart and used with permission

What exactly did Facebook do to spark this recent controversy? Here's one of many recent articles analyzing the controversy from many perspectives. Essentially, Facebook data scientists conducted one of many real-world experiments that are standard operating procedure with them and with most online businesses these days. More specifically, they did A/B testing on newsfeed-optimization models to gauge whether exposure to more positive words caused users to post more positive (and less negative) sentiment, and conversely whether being exposed to negative words had the opposite impact. In fact, the test, which took place more than two years ago, confirmed that hypothesis.

Was this hypothesis as controversial as some people have made it out to be, or has it been sensationalized and distorted beyond recognition? My feeling is the latter: it was just a routine real-world experiment in big-data-driven sentiment analysis, content optimization and customer experience management. Let's parse these practices into their respective spheres and examine whether any of them are unseemly:

  • Sentiment analysis: Nobody doubts that Facebook, or any other business, has every right to gauge whether customers feel positively or poorly about the value they're receiving. Facebook customers express those feelings in various ways, such as "likes" on individual content postings, as well as the tenor and substance of comments inline to their wall posts. And if a user has published his or her feelings (explicitly or implicitly) in any of those ways, Facebook can certainly leverage that content to infer what their sentiment (their mood) might be at any point in time. No one's questioning the ethics of them doing so.
  • Content optimization: Above all, Facebook is in the content-delivery business. The core content under its control is the news feed: personalized renderings of the consolidated content items posted 24x7 by each member's "friends," and by other sources to which they subscribe. To boost the value provided by news feeds, Facebook continues to tweak the machine learning, predictive and other models and artifacts that determine the arrangement, sequencing and rendering of subscribed-to content on browsers, mobiles and other platforms. No one questions the value of Facebook continuing to conduct real-world experiments designed to incrementally improve its content-optimization models. Customer sentiment, whether positive, neutral or negative, is a valid metric of whether one experiment is more successful than another.
  • Customer experience management: Like any business in today's hypercompetitive world, Facebook knows that it must improve overall customer experience to hold onto its members and to deepen their engagement with the service. Experience improvements are the chief outcomes that Facebook and other modern businesses attempt to realize, incrementally, from real-world experiments such as the Facebook initiative. Facebook, like many organizations, employs or engages with psychologists and other scientists to work with data scientists on the nitty-gritty of building and testing analytic models that describe what might be going in customers' minds as they interact on the service. If the real-world experiments in 2012 had shown, counter-intuitively, that viewing more negative content had actually, among some customer segments, boosted time on site and ad clickthrough rates, this intelligence could have been useful in optimizing those specific customers' experiences. If nothing else, even if nothing operationally useful had come from such experiments, Facebook should have at least been lauded for exploring counterintuitive scenarios. After all, "disruptive" business insights often come from bold thinking such as that. And if no customers were harmed in the process of conducting the experiments (moods being the very definition of ephemeral) Facebook's experiment could at least have been chalked up as a nice try.

Come to think of it, Facebook's attempt at mood manipulation was far wimpier than practices that are standard operating procedure in most industries: marketing, advertising and customer service.

  • Marketing exists to put customers in a buying mood, with mood manipulation tactics ranging from subtlety to sledgehammer
  • Advertising and promotions use every creative trick in the book to foster of warm glow of brand love in the customer's heart
  • Proactive customer service is all about keeping customers feeling loved, cared for and pampered to the nth degree

None of the practices outlined above are unethical, when conducted within the bounds of the law, privacy sensitivity and established ethical norms. And none "controls" the customers' moods to the extent that they lose the will to resist.

But every one of those practices is designed to make a modern brand practically irresistible to its target market.