A Friday post – Guns and Beer

Some time ago, I noticed how Facebook has a ‘recommended pages’ widget and it was not very good at estimating my taste. Anyone who knows me, will be able to tell you that ‘guns’ and ‘beer’ are not things I like very much. From this small observation, I dared draw a few Big Data conclusions:

1. Ad-blockers and anti-tracker-tools work. Apparently Facebook doesn’t know enough about me to give me any recommendation that is based on anything other than perhaps ‘he’s a guy’. (Incidentally, some will actually argue that my use of ad-blockers is breaking the internet, and that if everyone were to use them, companies like Facebook could no longer exist for free)
2. It’s easy to be wrong. The more precise we want the predictions to be, the more obvious it will be when they are wrong. If facebook had presented a list of twenty pages, I would have probably found five to ten that indeed match my interest, but because they attempted to only present the ‘most relevant’, they missed the goal completely. In general, drawing truly personal conclusions from Big Data may be much harder than the overal trends and dynamics. It may be easier to measure the sentiment of an entire continent than it may be to measure mine.
3. It’s easy to be right. On the other side, using big statistics, this may have been a safe bet for them: perhaps _most_ guys (of my age?) would indeed like these pages. Evidence the millions of ‘likes’ that these pages have already collected. That brings the question for yourself: what big new insights are you expecting that you don’t already know? Is the information you want to extract from Big Data truly revolutionary or are they frills, barely noticable compared to the big demographics that you already know?
4. I am apparently not ‘an average guy’ (but is anyone?). It did lead me to consider a potential scenario where, based on my data, every app and website would want to recommend me something that I really don’t like, or that is really not like me. If every website were to continuously promote ‘guns’ to me, because of some statistical correlation between me and true gun-lovers, what would that do to my relation to these websites, or technology in general? Would I start to love guns a little bit more, making it a self-fulfilling prophecy? Or would it increase my negative feelings towards guns? Or to reflect it back to a company setting: how can we leave the possibility open that some people truly are individuals that do not fit their statistical mold?