Facebook performs research, internet complains…

Facebook performs research, internet complains…
3 Jul 14

So Facebook has been getting quite a bit of heat this week for some experiments they conducted back in January 2012 on users of Facebook.

For those of you who have been living under a rock and are not aware of this study you can find out more about the study on PNAS.org. However in a nutshell Facebook edited the stream of some Facebook users to only show them positive messages their friends have posted, and showed others only negative messages their friends have posted. The result was simply that people who see mainly negative posts in their facebook feed are more likely to write negative posts and people who saw positive ones are more likely to write positive posts. I have of course simplified the results and you can read more about it on the link above.

The internet, as it seams to be with countless things these days, is appalled. I am less so and here is why…

Although most of you might know me as that Scottish guy who has crappy bandwidth, hosts a few shows on the network and is the Barry Chuckle to Ewen’s Paul Chuckle, by day I am a UX (User Experience) consultant. I work with many large government, financial and retail companies to help them increase sales and/or happy customers.

I, like the hundreds of thousands of other folks in the same career as me, do this in a number of ways, and all of them include running tests on customers to prove or disprove hypotheses. For example… “By changing this button from Red to Green more people are likely to click on it” we would then run tests to see if that hypotheses was true or false. These tests take many forms, the main ones you might be familiar with is the User Experience Testing, which is achieved by enticing a willing participant into a room with a laptop and a moderator. The willing participant (who likely will have signed some kind of consent document) will be asked to perform tasks on the website, how these tasks were performed is then reviewed and changes may be made to the website, app or service based on this feedback. There are many variations of this e.g. focus groups, surveys, etc… but these are all done with the full consent of the participant and everybody leaves the tests friends and in some cases the participant is £40 or £50 better off just for providing this feedback.

If the project or hypothesis requires it we may run more ‘behind the scenes’ tests called A/B tests (there are others but i am not going to go into these here). A/B tests are performed on a live website and (generally) visitors are unaware they are being tested. A real world example of a A/B test might be the British Tech Networks Bash Banner. Lets say the hypothesis is that “Having Ewen on the banner image would get more people to come to the bash as people want to meet Ewen” so we would create a banner with me on it and one with Ewen on it. Visitors to the site would either see me or Ewen, people who saw me would never see Ewen, they would never be aware that a banner with Ewen existed and vice versa. We would then check the stats after a few days/weeks and see whose smiley mug sold more bash tickets. The hypothesis would either be true or false. Now lets take it a stage further and dig a little more into the psyche of our visitors… Lets say we thought that the majority of the people likely to come to the bash were Hetrosexual Male Geeks i.e. that was our target demographic, we might then run a test that said that “Having a picture of **Insert name of pretty female geek** on the same page as the button to buy the bash tickets would increase sales” Note: not on the banner, but on the same page so not to be accused of suggesting she will be at the Bash. We would then make a banner with Ewen and one without Ewen but with a photo of **Insert name of pretty female geek** below it. Show them to customers as above and check the stats. You can see how this could get very detailed, very quickly.  Just for the record, we don’t run any A/B testing on the Britishtechnetwork.com.

Most online retailers do this as well (as do offline retailers, but i am not getting to that), as do banks, charities, public services, everyone.  For example, a charity might have a website to collect donations for a disaster in a third world country or to collect money for homeless cats.  You can be sure that some of them will be A/B testing images of sad children in this disaster ridden area or photos of neglected cats to find out which images make you; the visitor donate more cash.

“But wait” i hear you say… “Thats messing with our emotions and thats bad!!”

Perhaps, and its been done fore years, similar things were getting done before the internet in print, TV and Radio, this is nothing new. Don’t get me wrong, i think Facebook has gone into some morally quite grey areas of A/B testing. But this is simply A/B testing, you will have been a part of countless A/B tests in your time on the internet, in the last week, even today. Facebook wanted to find out information about customers to prove a hypothesis presumably to better target adverts. Thats all.

This is just research people (extreme research i know, but still just research), I am sure Facebook has done much worse with your data over the years and as always, if you don’t like it close your Facebook account. Now lets collect all our toys from the exterior of the pram and move onto the next wide scale internet panic.


Paul Wheatley



I totally agree with the comments here.

I’m sure that if we scratch further on a lot of what Facebook and other organisation and yes, governments, have done with the data that they have collected on us.

I might even suggest that Facebook is gaining data from the responses to the ‘leak’ of this information.

Would you really rather come and see me or Paul at The Bash tho? Be honest. Not insecure or anything. Just interested.

I completely agree wholeheartedly Paul. I feel a modicum of anger towards the powers that be, whether they be regular consumers, members of the ‘informed’ press or other Internet organisations when a powerhouse like Facebook uses methods like this to conduct research.

The fact they are as large as they means that anything they do has to be on a larger scale than organisations we are members of, or affiliated with, can only dream of and yes, I understand they have put themselves into a position where they are open to attack due to their success however if you were to compare the research conducted in this manner by Facebook in relative terms to ‘nefarious’ research conducted by other companies, tech or otherwise, as a percentage of their user base, you would see a set of comparable figures.

Oh, silly me, they are not as successful as Facebook and we don’t feel as threatened by them so it’s OK

Paul – you mention in #1 that you get the subject’ consent. We know that didn’t happen here, and the TOS to which FB refers were not in place at the time of the testing. The other point is the intent of FB here. The test they ran was not about which would make you click more often, it was whether they could make you sad. I think that’s the crux here. Luckily they were not particularly successful, much less than 1% change in behavior, but what if they had been successful? What if highly successful, and people committed suicide? I understand and agree with everything you said, but it doesn’t include the type of test FB actually ran, and goes against point #1 of yours. And thanks for not using my picture for the bash…

PaulW PaulW

Hi Alison, thanks for taking the time to read and comment on this.

Point #1 was to give context, in the scenario where i sit with a user and video them while working through a site or app i would expect to get consent, mainly as i am recording video of them and asking them direct questions. I would also expect to include their name in reports and recommendations based on their feedback. I included this information mainly to say that UX testing isn’t just about A/B. As with A/B (as stated above) i would not expect to get their consent.

In regards to your point about FB’s intention to make you sad (or happy) and that makes it different from A/B testing I disagree. Perhaps you don’t get the same kind of TV advertising as we do but there are a number of TV adverts (which can be viewed on YouTube and the charities website) which have the sole goal of making you incredibly sad so that you donate more. There was an advert on UK TV a couple of years ago for a Dog rehoming charity which was heartbreaking. I honestly had to change channel when it came on… and i also donated despite being a cat person!

And i can tell you without a shadow of a doubt that a list exists with colours of cats ranked based on how much extra donations they drag in. e.g. Cute ginger cats work best with single women from the north east, Black and white long haired cats work best with married men from the south, etc…

Companies play with our emotions all the time, thats how advertising works and FB is big enough and has enough users to be able to do testing on a large scale, thats all this is.

I have mixed feelings about this. When I go to Facebook, like other social media sites, I expect to be able to see all of what people post (providing the user decides they want me to see it). I do not expect the site owner to decide which of their posts I can and cannot see. Otherwise, in Facebook’s case, it is not an accurate timeline and devalues the content. This is why people, if possible, should always have their own web site rather than relying on just a social media web site so that you can control what people see.
I take your point Paul that this is about working out what works best for a web site to make it most effective and, if we are to be given ads, I would always prefer targeted ads rather than random ones.
However, I think a lot of the anger about this is that it is being done without permission or awareness of the users. It makes no difference which site it is, be that Facebook or a relatively unknown, small web site. And just because it may have been done for years on other mediums, does not necessarily make it right …

Leave a comment:

* Required.