Want better online recommendations? It takes better AI and a human touch

Spending hours on endless side-scrolling rows of Netflix movies or hunting through the perpetually long lists of identically rated restaurants on Yelp — this can’t be the way it should work. Part of the whole promise of the internet is that platforms and services would take the web’s infinite supply of everything – the things to watch, read, watch, play with, buy, eat, invest in, respond to, listen to. to or have feelings about – combine it with a deep understanding of who you are and what you like, and give back to you an endless supply of all your favorite things.

When it works, it can feel magical, like the TikTok algorithm that seems to know you better than you know yourself. But that is quite rare. More often, you’re haunted the web by Amazon ads for products you’ve already purchased, or you’re stuck browsing hundreds of 3.5-star Yelp listings or hundreds of similar-sounding true-crime podcasts on Spotify to find something. that you like. Or are you just going to watch The office† Again.

Good recommendations seem like a simple enough problem, right? The companies and platforms working on these personalization machines say it’s a more difficult problem than it looks. Mainly because people, you see, are hard to track down. But they also say there is a way to do it better. And a way you can help.

When the content recommendations app team Likewise first started building its platform, it thought the best way to make recommendations was by building a social network. “What happens in real life,” said Ian Morris, Idea’s CEO, “is you out for lunch or dinner, and the first thing after the ‘how are you, how are the kids? talking about things you’ve read or that great new show you’ve seen or a podcast you really need to listen to. That’s life!” Online, he found, those human connections and recommendations had been replaced by poor algorithms optimized for engagement and growth rather than real quality content. He thought Evenzo could be a resource for finding movies, shows, books and podcasts all in one place.

Morris is still convinced that this was the right approach. However, things didn’t go as quickly as he’d hoped – building a social network from scratch is seriously hard work – and so Equally started thinking about how to make the platform more useful, even for those who don’t have a large group of like-using friends. It hired an editorial team to scour the internet for the best and most interesting new stuff while also building a machine learning system that could make automated recommendations.

In the same way, you collect all the things you want to watch and all the things you think you should watch.
Image: Likewise

Now, when you start using the Evenzo app for the first time, you have to tell it about things you like. If you want movie recommendations, you’ll need to pick some genres first — comedy, drama, western — then pick some of your favorites from a curated array of titles. You won’t be able to access the rest of the app until you’ve chosen at least 20 of them. “The payoff is huge,” said Salim Hemdani, CTO of Alsos. “The more you tell us, the better it will be.” He says people never stop at 20 because it’s just fun to pick things you like. And by doing that, you’re telling Alsos’ algorithm who you really are.

Also uses that information to put you in a “cluster,” which refers to a group of people with similar tastes to yours. These clusters are constantly changing based on what else you watch and rate, and they inform you about everything else you recommend as well. “It gives us a starting point to say, how many people are like you in the world and how many clusters can we create?” says Hemdani. The more detailed and specific those clusters, the more accurate they can be. Know you like it succession is somewhat useful; knowing you like it succession, novels by Michael Crichton, the podcast the adventure zone, and anything with Marvel in the title is much more useful.

The simplest and most pervasive recommendation system, on Also and Elsewhere, is known as collaborative filtering. It works by assuming that if you like something, and someone else likes that thing and also a second thing, you probably like the second thing too. That is it! It usually involves more data and more people, but that’s the core idea: if you want severance pay and other people who liked it severance pay are really digging The old manyou probably will too.

One of Morris’s theories is that Evenzo can make better recommendations, not just by getting to know users better, but simply by giving them more stuff. Netflix, HBO, and Disney will never recommend each other’s catalogs, but Likewise (along with apps like Just look and reel good) can index them all. “We’re not aware of any recommendation engine that looks at things like the social chart or looks through books, podcasts, TV shows, movies,” Morris says, “and lets your preferences and other things influence each other in those categories.”

The easiest way to get better recommendations, told almost everyone in this space, is to make the apps and platforms work more. Multiple executives described the ideal personalization process as a collaborative exercise in which you and the AI ​​work together to paint an accurate picture of what you really like. Everything you do on Netflix helps the app put you in the right clusters; every filter you check on Yelp makes the restaurant recommendations more useful. Downvotes and dislikes are just as helpful. Clicks, likes, and even engagement can mean many things, but an explicit approval sends a much stronger signal.

Screenshot of a Pinterest search for

Pinterest has embraced personalization as a collaborative process with users.
Image: Pinterest

Oddly enough, though, many platforms have gone the other way, opting to infer what you like based on what you click on or linger on while scrolling or engaging in some way. It’s based on a desire for a completely frictionless user experience, but from Facebook to YouTube to TikTok, we’ve seen what that can lead to: misinformation, rabbit holes, echo chambers, all kinds of problems. It also requires collecting astonishing amounts of data, collecting all possible information about you and your habits just in case some of it is useful.

Naveen Gavini, the SVP of product at Pinterest, says he understands the tendency towards frictionlessness. “If you opened up your favorite streaming content platform and you went to watch a movie,” he says, “I don’t think you’d want to answer a 30-question quiz first: Hey, what are all your favorite movies? you judge them? Who are your favorite actors? I don’t think anyone wants to do that job.” Instead, he says, it’s important to find just the right times to ask questions. “I have a hairdresser that I’ve been going to for 10 years who cuts my hair,” says Gavini as an example. “And if you think about that experience every time, it’s a personal experience, and I don’t have to tell him when I come in how I want my haircut, because he knows me. But it started with that first conversation: it was an explicit conversation, like, ‘Hey, how do you like your hair in general?’” Making that same kind of dialogue explicit, without overusing it, is a key goal for Pinterest .

A side effect of that collaborative process is that it can also provide users with greater transparency about what is recommended to them and why. Nearly everyone I spoke to for this story said it’s important to help people have great experiences online and to instill confidence in the things that are recommended. “More and more”, says Gavini, “I think we want to know: what are the decisions? What are the things that inform some of these algorithms that actually deliver content to us?”

Trust is really everything. There’s a hypothetical version of the Yelp app — and the Netflix app, Spotify app, Kindle app, and dozens of others — that’s nothing more than a big button. You sit down to watch something, press the button and Netflix knows exactly what you’re looking for. Spotify puts on just the right song. Yelp orders exactly the dish you crave. Everything is personalized and automated and delivers the One True Recommendation every time. But would you believe it enough to just push the button? Akhil Ramesh, head of consumer products at Yelp, doesn’t think so. “I often joke that if God landed in front of me and said, ‘This is the person you’re going to marry, and you never have to waste a second,’ I wouldn’t believe a second of it,” he says. “I was going to do my reconnaissance.”

The one true recommendation isn’t just impossible – it isn’t even really worth pursuing. But that doesn’t mean it couldn’t be better. As the services we use get to know us better — and, just as importantly, get better at asking about ourselves — they may be able to narrow the world down to a handful of options rather than an endlessly scrolling list. All you have to do is pick your favorite and go. Because really, there is not a good answer. There is only the one you have chosen.

Leave a Comment

Your email address will not be published. Required fields are marked *