The Human Touch: How Tech Titans Are Fixing Their AI Recommendations

September 30, 2019 - 7 minutes read

After a long, hard day at work, you’re finally home! You plop down on your couch, turn on your TV, and tune in to your favorite streaming service. Looking for something new to watch, you navigate to your “Recommendations” section. Much to your displeasure, all of the options listed don’t interest you one bit.

Does this sound familiar? Well, you’re not alone.

For the past few years, big tech companies have been investing in recommendation engines powered by artificial intelligence (AI) algorithms. But all of this AI development and research has led to one indisputable insight: the human touch is irreplaceable.

AI Algorithms Aren’t Infallible

HBO recently launched “Recommended by Humans,” a website that helps viewers figure out which HBO show they should watch next. But rather than rely on AI to sift through its extensive catalog, the premium channel has instead hired humans to make recommendations via video testimonials. A few weeks after the site’s debut, Netflix started testing a human-curated section in its own app.

These two companies are far from the only examples of organizations rediscovering the human touch’s value. In recent years, both Apple and Google have revamped their app marketplaces to emphasize picks and recommendations from human curators.

For its Apple News app, Apple employs roughly 12 journalists to decide which stories are front page-worthy. Similarly, Facebook also recently began hiring journalists again to select top stories for News Tab, its upcoming news app. Last but not least, Google just rehired the creator of Google News, Krishna Bharat. While absent from the company, Bharat became one of the service’s sharpest critics by saying its lack of vetting story sources was “shameful and irresponsible.”

The human touch is nothing new when it comes to AI development; most tech companies still turn to humans to train their machine learning applications and moderate content. But it seems that many organizations have recently come to the conclusion that handing over the curation reins to humans still has some merit.

The Comeback of Human Curation

For some tech pioneers, this revolution seems inevitable. During the 2015 launch of Apple Music, Jimmy Iovine, the former leader of the service, stated that algorithms alone couldn’t take on the “emotional task” of curating the right music at the right time. But while this trend back towards humans seems like a long time coming to some, it has taken on greater urgency due to the negative effects of AI-fueled tech products.

Both Google and Facebook have admitted that their algorithm-fueled recommendations played unfortunate roles in showing young users inappropriate content and spreading fake news. And on a much lighter note, many critics of Netflix have pondered if the service’s algorithms are to blame for the failure and non-renewal of some of its highly-rated shows.

AI algorithms definitely have their place in recommendations. But in order to compensate for their shortcomings, the human touch is sorely needed. Case in point: Facebook’s News Tab team has instructed its human editors to avoid promoting polarizing stories. For all the “intelligence” they offer, algorithms still have a hard time doing this because they do not truly understand the meaning of the source material.

“It’s going to be a long time before machine learning—or whatever you want to call these algorithms—can understand the meaning of a statement,” explains former Apple executive Jean-Louis Gassée.

Michael Bhaskar, author of Curation: The Power of Selection in a World of Excess, shares a similar sentiment: “I think you need to have machine-driven stuff, just because the sets of information and media are so large. But then, people like people.”

Can We Scale the Human Touch?

San Francisco developers like Apple and Facebook embraced algorithm-powered recommendation engines for numerous pertinent reasons. Perhaps the biggest ones are that not only does it enables them to process immense amounts of information cost-effectively, but it also allows for infinite personalization.

Unfortunately, human efforts can’t scale to this magnitude. As a result, the following question begs to be asked: is this investment in the human touch a permanent one, or is it just a stopgap until algorithms can improve?

YouTube has already made it known that the human-curated “Collections” section in its YouTube Kids app is helping the company’s algorithms improve their discernment of quality content. And while the future of News Tab seems bright for human editors, it’s actually Facebook’s second attempt at human-curated stories. After accusations of liberal bias, the company fired the human editors behind its previous “trending news” section and gave algorithms the reins until the segment’s complete shutdown in 2018.

But some companies are showing that it is possible to find a fine balance between human and AI curation. For instance, the news app Flipboard uses AI for general personalization and humans for more fine-grained tweaks.

On the other side of the equation, some companies are doubling down on human curation, regardless of its potential (or lack thereof) to scale. Apple’s Beats 1 station, for example, has renowned musicians and DJs curate songs and act as on-air personalities. This affords them the ability to speak about and explain their song selections. Bhaskar believes this ingredient of human curation is irreplaceable: “The thing you can never get from an algorithm is that it doesn’t have a story behind it.”

For now, this story itself is still an ongoing one; it remains to be seen whether companies will turn back towards algorithms, hire more humans, or rely on a blend of the two for recommendations. What do you believe leads to the best curation of media? Let us know your thoughts in the comments!

Tags: , , , , , , , , , , , , , , , , , ,