Live by the Algorithm, Die by the Algorithm
YouTube has had a really bad couple of weeks. Large advertisers have been abandoning the platform in droves following reports that the platform runs pre-roll ads for major consumer brands on racist, hate speech videos. In response, YouTube has gone into full-on crisis mode, issuing abject apologies and promising a complete overhaul of its advertising policies.
What in the world just happened, and what does it mean for the future of TV? Three thoughts….
1. Algorithms Do Not Get Offended, But People Do
MIT is conducting new research on how to teach Baxter robots to be embarrassed. Maybe these robots should apply for a job at YouTube. The online video platform giant has built its entire business on the basis of software automation and tremendous scale. YouTube has over a billion users generating more than a billion viewing hours per day. More than 400 hours of video get uploaded to the platform every minute.
As a result, YouTube (the business) is completely dependent on YouTube (the algorithms) to function. The Company’s defense of its approach (including Eric Schmidt’s interview on Fox News) was both well meaning and nonsensical. On the one hand, Dr. Schmidt claimed the company’s algorithms are great at matching ads and content, but then immediately admitted that the algorithms make mistakes and can be beaten. So Dr. Schmidt (the engineer) promised that the algorithms would get smarter and magically solve the problem even as the company promised advertisers that YouTube had posted thousands of humans to manually review the machines’ work. For a billion hours of video? Give me a break.
The bottom line is that an open video publishing platform like YouTube is never going to be able to identify in advance all the content that an advertiser might find offensive. The reality is that the best they can do is react quickly and take things down once an issue has been identified, 98% of which the company says are resolved within 24 hours. But the real problem is that, in the new world of screen grabs and Twitter journalism, takedowns may come too late and the damage to advertisers may already be done.
2. Free Speech And Advertising Don’t Get Along Nearly As Well As We’d Like To Believe
The real elephant-in-the-room here is the relationship between freedom of expression and advertising. YouTube has gone to great pains in all of its public communications on this topic to defend the importance of creative freedom. Basically, the only thing YouTube fears as much as the wrath of advertisers is being accused of content censorship.
Of course, YouTube already censors loads of content every day (and regardless of whether the content is technically legal or not). Pornography has long been banned from the site, even though such content is perfectly legal for adults. Videos showing drug use are equally verboten (although watching a video of someone using drugs is certainly not illegal). YouTube bans educational videos if the lesson being taught is how to make a bomb. More recently, YouTube started censoring violent videos after multiple incidents of beheading videos by ISIL and other terrorist groups appeared on the site. These are actually the easy cases (at least for humans) because the content in question seems to clearly invoke the harm principle.
The more difficult question is where to draw the line regarding content that is not harmful but merely offensive. As YouTube just found out last month with PewDiePie, the line between entertaining and offensive is vanishingly small, and often in the eye of the beholder. A whole lot of comedy, for example, is based largely on the capacity to offend. This was true long before YouTube came along. Eddie Murphy’s Raw (released in 1987) is still the #1 standup comedy film of all time at the box office. Thirty years later, most mainstream brands still would not want to run pre-roll ads in front of clips from that show.
Why are advertisers so conservative when it comes to content? The crux of the problem is that, in the minds of both consumers and advertisers, video advertising equals sponsorship equals endorsement. When a viewer is offended by the content, the offense is transferred to the advertiser that sponsored it. Not good. This brings us to our last point.
3. What Does ‘Advertising-Supported Content’ Actually Mean?
In the early days of television, advertisers would sponsor an entire show. The TV viewer clearly understood that the sponsoring brand (Chesterfield cigarettes, kids!) was supporting and underwriting the cost of the show on an ongoing basis on behalf of the audience. In the 1960s, this model faded away and was replaced by ad breaks. The advertiser bought ‘time’ on a particular show to be aired at a particular hour. The linkage was still there but in a weaker form. The show was produced independently of any particular advertiser and made no mention of the ads that would accompany it, meaning ads could be run during a variety of different shows. At the same time, however, the advertiser retained total control over the show (or shows) they wanted to advertise against. In the case of some shows and advertisers (e.g., Ford or Chevy pickups and the NFL), the relationship became well known to viewers via repetition and could continue for decades.
Programmatic, personalized advertising to quantum viewers completely changes this relationship between the advertiser and the content in three key ways.
First, viewers of the same video can (in theory though not yet all that frequently in practice) each be served a different ad. A middle-aged male interested in motorcycles might see a Harley Davidson advertisement, while a new mom might get an ad for Pampers diapers. In each case, the choice of ad is entirely about the viewer and has nothing to do with the content of the video being viewed. This could even be done within the same household by identifying who’s watching each screen (gotta love those facial recognition algorithms).
Second, an algorithm, not a human, is making the match between content, viewer, and advertiser. The only ‘relationship’ between the content provider and the advertiser in this case is a mathematical one, as an algorithm inside a YouTube data center determines (in .01 seconds or less) the best match.
Third, the advertiser and content provider no longer have any direct financial relationship. Most content providers on YouTube get paid monthly by YouTube itself, based on a revenue share of all ads served against their content. This could comprise a theoretically limitless number of different advertisers spread across hundreds of different videos. Attributing that revenue to a particular ad (or advertiser) is practically impossible.
Once the algorithms are in charge, there’s no longer any factual basis for imputing responsibility to advertisers for video content. Unfortunately, brand marketing is built on perception and emotion as much (or more) than on reality. Human viewers experience the advertisement and content together and instinctively see an endorsement relationship. The emotional reactions from all parties (including both advertisers and YouTube itself) suggest that managing these perceptions will remain a big problem going forward.
Stick with TDG and stay ahead of the curve.
Joel Espelien is a Senior Advisor for TDG and serves as an advisor and Board Member to technology start ups. He lives near Seattle, WA.