This week, tech news and social media were all up in arms about a test that Facebook had performed a few years back.
The test was very simple: for a small subset of users, Facebook tweaked the algorithm that decides what posts to show you. Group A got slightly more negative posts displayed in their feed, and group B got slightly more positive posts. The net effect is that people in group A went on to share slightly more negative posts themselves, whereas the folks in group B posted slightly more positive things.
Now, I’m not going get into the ethics of the test, or the tap dancing around Facebook’s terms of service that apparently occurred. This isn’t even about whether Facebook’s actions were good or evil, right or wrong or something else (though I will touch on one of the outcomes of the test and how it has direct relevance to something I’m doing.)
Nope. This is about a greater reality that I think you need to be aware of.
You’re being experimented on all the time.
Become a Patron of Ask Leo! and go ad-free!
If this then that
The Facebook test boils down to a simple concept:
If <this> then <that>
The “<this>” was altered, and the resulting changes to “<that>” were observed.
Web sites, email marketers, newspapers, advertisers and the mass media do this all the time. In its simplest form it’s called “A/B testing”. Have two different “<this>’s” and see how each impacts the “<that>”.
A few examples:
- Two different headlines for the same article: which one generates more readers?
- Two different headlines for a social media post: which one generates the most clicks?1
- Two different prices for the same product: which one garners more total revenue?
- Two different ads: which one generates the most sales?2
- Two different paragraphs of text: which one causes more people to take action?
- Two different articles on the same topic: which one are more people clicking through to read?
It’s not that much of a stretch to say “Two different collections of posts, which one makes people happier?”
Of course, tests can be more complex than simple A/B testing.
Tracking need not be involved
One of the topics that’s a big source of concern is – of course – tracking. People are worried that they, personally, are being tracked.
Tracking need not be used at all to perform these tests. It’s purely a numbers game.
Let’s say I have a form that asks you to subscribe to a newsletter. I create two versions of this form – A and B. I run each for a given amount of time 3, or I randomly show some people form A and some people form B. If 5% of the people that see form A sign up, and 10% of the people that see form B sign up, then my test tells me that form B is more effective. No individual information needed whatsoever – purely counting the number of times shown, and the number of signups that resulted from each.
Most testing is just this simple.
And it’s ubiquitous.
That ethical line
I’m not saying that testing is always that simple or anonymous. The Facebook test tracked the behaviour of a set of individual accounts, for example. And yeah, that’s where things get sticky.
A few years ago Amazon got into some hot water for showing different people different pricing for the same item. It’s difficult to know whether that’s testing, or a more dynamic pricing model where an item’s price can, quite literally, change over the course of a day in response to … well, one would expect supply and demand. But it’s likely that companies may very well consider many additional factors in determining and adjusting prices.
Is it ethical? I don’t know. I believe that most of the time the simple cases of A/B testing are, and honestly, they don’t concern me. Heck, I should be doing more of them myself. (Newsletter signups are, after all, the lifeblood of a site like Ask Leo!.)
But it’s also quite possible to construct tests that cross the ethical line for people. Perhaps Facebook’s was one of them.
What’s important to know, though, is that this kind of thing happens, and more often than you probably realize.
Using the data
I mentioned that one of the outcomes of Facebook’s test had direct relevance to something I’m doing.
That outcome was this: people post more positively when they see more positive posts. Part of me wants to yell out “Well, Duh! OF COURSE they do.” It seems obvious to me that things that even slightly elevate (or depress) our mood cause us to act in accordance with that now slightly-changed mood.
And ultimately this is exactly why I took on HeroicStories.