I used to be a member of a small lending library called ‘A to Z’ in Saibaba Colony in Coimbatore, while in school in the 90s. The owner used to observe the kind of books I pick and make contextual recommendations on what I could pick up next. I used to really appreciate that gesture since it is showcasing his personal interest in his customers, and also his expansive knowledge of books.
When Amazon makes so-called contextual recommendations calling it, “People who bought this also bought…”, I really have nothing to appreciate it for. It’s simply an algorithm designed that way, mining troves of data to add that context when I’m on the site.
But, when Amazon connects the dots between my picking 3 different soaps with a citrus fragrance and make a recommendation for a 4th citrus fragrance based soap, I would find that both useful and interesting, though not be appreciative of it yet 🙂 Again, it’s an algorithm, after all.
Netflix does precisely this. Netflix’s use of data is legendary – it knows the kind of scenes we like, the kind of endings we like, the kind of plots we like and uses all that to make decisions on new shows and plots. So, the question arises – is the dog wagging the tail or the tail wagging the dog? That is, are we dictating, based on our behavior on Netflix’s platform, what kind of shows they make, or is Netflix dictating what we should be watching based on aggregating our own behavior?
The same question arises from a children’s book platform called Epic, as reported in The Wall Street Journal recently.
Epic, as an online subscription book platform, reaches 50 million children around the world. During the pandemic last year, they opened a limited version of their site at no fee, to the public (I signed up for my daughter, who can read 2 hours free, per week).
Some of the kinds of data Epic gathers are extremely interesting in terms of children’s collective behavior. For instance,
Epic’s team knows that children prefer owls to chickens and chickens to hedgehogs. Kids hunt for unicorns almost twice as often as they look for mermaids. Volcanoes are more popular than tsunamis, which are more popular than earthquakes. The Titanic is bigger than cowboys, pizza is bigger than cake, science is bigger than art and “poop” is bigger than all of them.
Snapshots of recent activity on the site offer a fascinating glimpse into the child hive mind. Searches are by kids up to 12 years old, the company said, with most of them likely by children 6 to 10.
On its rankings of animals searched most often by kids, owls score higher than koalas, worms rack up more hits than kangaroos and ferocious beats have a loyal following, as does the grumpy-faced blobfish. An entire category of search-term data is called “gross”.
Last fall, when the seventh “Cat Ninja” book hit the Epic platform with a cover picture of an owl sidekick—the Hoot character, a burgeoning superhero who is mostly two big eyeballs and wings—traffic shot up to 350,000 reads in one week, more than twice the average for previous books in the series. All those clicks led to plans for a five-book Hoot series.Source: The Wall Street Journal
Now, this is not new at all. Google has a treasure trove of our search data to customize our search results. But when Epic uses data such as ‘owls are the most popular amongst our readers’ to create a new owl character, there is no surprise in why that title become so popular.
The question, however, is once again, “Is the tail wagging the dog, or the dog wagging the tail?”. That may be difficult to conclude, but consider the possibility that kids on Epic may never ever stumble upon an artistic hedgehog (for whatever it is worth).
Consider another excerpt from the same article:
At a recent series of Zoom brainstorms, led by director of original content Eric Wight, staffers improvised ideas for future Hoot plots, peppered in random facts (guinea pigs travel in herds, who knew?), worked to clarify story lines (“Is it confusing if the chicken is named Mona Lisa?”) and grappled with identity politics in ancient Egypt (“With the cat as the villain, are we getting into any cultural appropriation stuff?”).
Most of the process was more art than science—the writers and editors are veterans of major publishing houses with many print titles between them—and their imaginations factor into decisions more than any machine learning, company executives say. Still the data can be a springboard for new ideas, as it was when it came time to find future Hoot characters.
Out came the spreadsheet with animal rankings. “Bug is higher on the list than crocodile,” said one Epic writer, scanning a column. “Do kids ever check for hippos?”. No. “Snake is pretty high,” a colleague countered, imagining a villain cobra in striking position with its hood flaring like a cape.
“If we create a book about a dinosaur riding a shark,” Mr. Wight said, “we could just retire.”Source: The Wall Street Journal
While I understand using such data to dictate new choices in terms of story lines and characters, something I’m sure Netflix factors in for viewers too, where does that leave the children (and us, when it comes to Netflix) in terms of feeding our own biases and reiterating our predisposed interests? At its most basic, this is, “A LOT of you like X, so here’s more of X”. As the kids grow up, would they be incentivized to like things outside of their cycle of content contours? As a direction to respond to that question, consider your own preference of TV shows on Netflix!