It's common nowadays to understand the bigly-ness of Big Data in terms of "Five Vs":
- Volume: there's a lot of it.
- Velocity: the speed with which more is added is accelerating.
- Veracity: not everything on the Internet is true (!).
- Variety: it's in many different (structured and unstructured) formats.
- Value: it's hard to extract value or meaning from it.
I suggest adding two additional Vs:
- Volume (in the auditory sense): it's easier to make yourself heard by more people.
- soVereignty (Yeah, so I cheated it. Sue me.): it's difficult or impossible to control data about oneself that gets online.
These Seven Vs are what make Big Data at once promising and threatening. I want to discuss one particular threat that, as far as I know, hasn't been addressed yet: vicious dynamic feedback loops between predictive analytics and prescriptive analytics. 'Predictive analytics' refers to the use of Big Data to make predictions about hitherto unobserved data. For example, if you have a smartphone, you can turn on predictive analytics to increase the speed of your typing. I just started a message to my wife by typing the letter 'y'. Siri suggested the word 'You'. OK. What's next? Siri gives me three options: 'can', 'are', and 'have'. Let's try 'are'. Now Siri offers 'the', 'a', and my'. I'll go with 'my'. 'Best', 'favorite', 'fav'. What the hell is a 'fav'? Anyway, I'll go with 'favorite'. 'Person', 'and', 'thing'. Thing? That's a bit rude. I'll go with 'person'. So, starting with 'y', through no thought or effort of my own, I get to 'You are my favorite person.' Aw. That's sweet. True too.
This is the power of predictive analytics. It doesn't just diagnose or explain extant data: it makes guesses based on already-analyzed data about what comes next. If the existing data is adequate in terms of volume, veracity, and variety, predictive analytics does a good job. And it can keep doing a better and better job if there's enough computing power to keep churning through new data.
But what if the existing data, instead of being sweet, is distinctly acrid? What if, for instance there's a lot of sexist trash in the data-stream? Then we get Google suggesting search queries like this:
I'll return to 'why are women so mean' and 'why are women so emotional' in a sec. Let's start with the last suggested query. I hadn't considered asking why women are always cold, but now that you mention it, Google, why are they always cold? Let's find out.
Huh. Apparently women are always cold because they don't have balls. Weird. Thanks qualityhealth.com (and Google)!
More importantly, note that little box around the search result. This wasn't just the top result out of the ten Google showed me on its first page. It also got the Google seal of approval. That box means a lot. It's part of Google's effort at prescriptive analytics. 'Prescriptive analytics' refers to the use of Big Data to recommend not just links but verified, gold-standard answers to queries. For example, if you enter the search query 'sunset today' and have allowed Google access to your location, it'll pop out something like this:
Thanks, Google! That sucks, but at least now I know.
Not every query returns a recommended answer with a little box around it. When I googled 'rabbit domestication', Google's prescriptive analytics didn't tell me the answer, which is that bunnies were first domesticated in France by some medieval monks to get around the Catholic Church's meat prohibition during Lent. For some reason, the meat of young rabbits had been declared fish for religious purposes. The more you know....
Let's return, though, to sexism. I was just playing around online, then Google suggested I try a misogynistic query. OK, Google, you've been so helpful up until this point, I'll give it a shot: why are women so mean? 302 million results! Wow, I guess the internet hivemind has been puzzling over this question a lot. In the top 10 results (which almost no one ever clicks past) I find three more specific versions of my (well, Google's suggested) query:
- Why are hot/beautiful women so mean?
- Why are women so mean to men?
- Why are women so mean to each other?
Let's find out. According to someone who goes by the name of 'Dan Bacon' (really, dude?) writing for themodernman.com,
Attractive women often act cold and uninterested towards guys because they don’t want nervous guys to feel comfortable about approaching them.
Uhh... OK. If you say so. What does 'Joe Valentine' (come on, really?) at thoughtcatalogue.com have to say?
It’s important to understand that most beautiful women have been approached by hundreds, maybe thousands of needy guys, depending on their age. Over time, they develop a keen sensor for beta-males. Just like a baby grizzly bear, they can smell fear, and they think it’s pathetic.
It seems that Joe agrees with Dan. In case any of y'all were wondering, hot chicks despise betas. At least according to Google.
When I went down one of the other rabbit-holes, I got answers like this:
So, #NotAllWomen but #WayTooManyWomen. Thanks for mansplaining this, @conceptualclarity (really, dude?)!
The problem of predictive analytics suggesting hateful queries that people might not have made otherwise has been discussed before. Here are a few further such suggestions I found by playing around for just a couple of minutes:
Someone who wants to know... something about jews, muslims, or the black lives matter movement is liable to have their query directed, shaped, or outright constructed by the predictive analytics that drive Google's search fill-ins. It's worth noting that this can only be because lots of people have entered these queries. AIs are really good at learning, and learning quickly. So if you feed them hate, they learn to hate.
By the way, when I took Google up on its suggestions to ask whether jews are white, muslims bad, and BLM racist, I learned, first...
that Jews are white for now. I also learned...
that Muslims are most certainly bad, and finally that...
Yup, that's a little box you see around the answer from truthrevolt.org. It's got Google's gold-standard approval. So, unlike the facts of rabbit domestication, Google assures me that it's certain -- as fucking certain as the fact that the sun will set at 4:32 PM today -- that BLM is anti-American, Marxist, and racist. Of the other nine hits on the first page of results, six results agreed with truthrevolt.org, two disagreed, and the last was a link to the satire website clickhole.com:
Good question, Jerry Metz (why is the satire guy the only one with a real name?!)!
So this is the world we live in now. If memory serves, the psychoanalyst Adam Phillips (it might have been John Forrester) has argued that every seduction begins with I know what you want. My contention, then, is that we are being seduced to hate by the vicious interaction of predictive and prescriptive analytics. Google knows what you want... to know, and it has ready-made answers to the questions that it itself prompts you to ask.
Terrified? So am I.