These days everyone has an algorithm. It’s all about predicting our behavior, our impulses, our purchases, our viewing habits, surfing (web) habits, and anything else that some potential advertiser, obscure government agency, or fad non-profit organization might pay for in order to target us with some pitch, product or panacea that 99% of the time we’re going to say no to. They’re gaming for the 1%-ers. The problem is, none of these algorithms seem to work.
Not that they don’t do something predictive. They do. A recent conversation over the dinner table involved one-upping stories of the sudden onslaught of targeted advertising after clearly precipitating events, that simply weren’t substantive. One person related having responded to an email with a hearty congratulations to a friend who had just bought a condo on a beach in Spain. Within 24 hours he was receiving both email and sidebar ads as he surfed, offering up connections to real estate agents on the Spanish coast. Another had clicked “Like” on a picture of a friend’s dog doing some sort of trick and found Facebook suggesting pages of pet training services on a daily basis. And more.
Like everyone else, I’m subjected to algorithms from the moment I wake up in the morning….
I make my coffee and then sit down to peruse my email. According to Google, the gmail application learns over time to predict what sorts of mail should be shown as priority, what shouldn’t, what should go to spam, and what not. I can count on that at best, maybe half of my mail will be sorted correctly. I have a filter for mail coming from the Casa SaltShaker website reservation system to automatically put those into priority mail. Maybe 2/3 of them will be, despite a very clearly worded filter. I’m on several mailing lists for which I get a daily email. Roughly half the time, those emails are marked as spam – despite the fact that they arrived from the same address and with the same subject line, daily, and every time, for years now, they go to spam, I mark them as “Not Spam”.
On the flipside, I get junk mail that I regularly mark as spam which continue to arrive from the same addresses and with the same subject lines, daily. Well over half the time they end up in either my regular or even priority email boxes. I have whitelists and blacklists, neither of which seem to have any effect on where mail ends up. One algorithm down.
My favorite has to be the Netflix “Top Picks for Dan” queue. Yours probably says “Top Picks for Joe” or “Top Picks for Shannon”, or maybe even “Top Picks for Elspeth”, a charming personalization touch meant to make sure we feel loved and cared for. By Netflix. Mine currently has 40 “top picks”, I guess that hearkens back to the days of “Top 40 Hits” or something. Now, if you use Netflix, you know it has a rating system, of 1 to 5 stars. After you watch something, you’re invited to rate it, and that goes into an algorithm, along with the genre and style of film or show, to help create a predictive algorithm. Now, I’ve rated a lot of films over my time on Netflix, so it ought to be narrowed down pretty well, right?
When they show you their recommendations for top picks, they also show you what the algorithm predicts you’ll rate the film or show. So one might, reasonably, predict that almost all of the top picks would have either five or four stars, no? No. Of those forty, currently, seven of them have five star predictions, fifteen are guestimated to receive a four star review from me, three for three, a whopping nine for two, and even five for one star. Oh, and one television series which I’ve already watched and rated. Four stars. So, why recommend anything that they predict I’ll only give one or two stars? Given the vast catalog, why even three stars?
They are good at picking out some four and five star recommendations that I might not otherwise have picked. It recently led me to binge watching through two different TV series. Nowhere Boys, which I only clicked on because it looked like mindless entertainment with cute boys, I figured I’d watch an episode or two of one evening when I just didn’t want to think. It turned out to be an Australian series, in the all-popular genre of teen witches, but with some dark twists, and it’s well written and well acted, and, well, the boys are cute.
It also led me, inexplicably, to a one-season Korean medical telenovela (soap opera), titled Good Doctor, that likewise looked like there were some cuties in it (maybe that was the predictive element?), which turned out to be the travails of a young surgical resident who has more or less overcome being autistic, enough to just barely function in day to day life, but he’s a savant in the world of medicine. It was, at times, sweet, at others painful, and not only was it well written (the English subtitles probably missed a lot of nuances, but even so), but it didn’t stint on the medical stuff – getting right in, close-up to the blood and guts in the operating theater (I was, at one time in my life, a paramedic, considered going to medical school to become an ER doctor, and am still fascinated by that whole world). But, those were also predicted to, respectively, receive four and five stars from me. Not one or two.
Until recently I was using a Samsung Galaxy S4 phone. When I would start to type a message, it would automatically detect whether I was starting off in English or Spanish, and switch dictionaries for predictive entry accordingly. I recently “upgraded” to a Galaxy S6. It doesn’t do that. Apparently (despite the fact that the S4 did it really well, it rarely made a mistake in which language I was working in after the first couple of words), too many people complained that it either a) didn’t pick up the correct language within one or two keystrokes (seriously?) or b) switched to the incorrect language.
So Samsung scrapped it because it wasn’t a good enough predictor, and now has a button to select your language (but every time you start typing a new message, it switches to your main language, which means if you’re having a back and forth conversation, you have to re-select, in my case, Spanish, if you want to use predictive entry, over and over again. It also, annoyingly, sometimes tries to switch me to Chinese (apparently the “international version” phone I bought was pre-programmed with Chinese language bloatware that while I can disable, I can’t remove, and now and again, some of it resurfaces). Thankfully, there are apps you can download to overwrite the Samsung keyboard and give it back automatic switching (thank you SwiftKey).
Foursquare is another one. There’s a bit of rating involved – Liked, Neutral, Didn’t Like – that goes into it. But more, they have a service of recommending restaurant alternatives – “If you like that place, you’ll like this place.” It’s a mystifying selection process. If I look at the listing for my restaurant, a reasonably upscale, albeit casual, experience, with a five course tasting menu, paired wines, at a communal table, FS recommends on the basis of having liked us (we have an 8.6 rating out of 10), one will like: a Mediterranean café and bistro on the opposite side of the city (rated 7.4 out of 10); a Mediterranean restaurant in Palermo (7.0 out of 10); and a bar in Palermo (5.3 out of 10). They’re not in the same neighborhood, they’re not, well, in the same class, nor even the same style of food, or even cuisine. Go figure.
LinkedIn – I’m not sure if this is the same as a predictive analysis, but I get a weekly email from them with “job openings you might be interested in”. The only predictive part of that seems to be that they’re located somewhere in South America, not even always in Argentina. Not once, over more than a decade, has any of the jobs related to either gastronomy or journalism. Generally they’re things like “Chief Operating Officer of a Petroleum company”, or “Executive Director of an Animal Rights Foundation”, or “Receptionist and Whipping Boy in the Fashion industry”.
Coursera seems to think the courses that I would like best are things involving teaching English as a foreign language. I’ve never taken a course from them remotely tied to that theme, they’ve all been either food or history related. Pocket keeps recommending articles for me, on a wide variety of topics. I don’t use Pocket.
I could probably keep going, finding other examples from daily life, but enough. We all know it’s happening around us, we can’t really do anything about it except make annoying blog posts. So be it.