Confirmation Bias and the Feedback Loop of Algorithms

Algorithms are bad for media (and society)

I read a post from Boston University this week on how data is transforming what we read and watch – and I assume listen – throughout media. The changing landscape of media isn’t anything new, but this does seem to be an unintended consequence of having terabytes of data at our fingertips.

In the early days of the internet, as revenue streams started to decrease in traditional media, newsrooms had to cut staff in order for organizations to remain profitable. Inevitably this meant that newsrooms were creating less content. But it wasn’t just that they were creating less content.

They had to make hard decisions around the content they did create. The content they created had to reach the largest audience possible in order to sell advertising and subscriptions.

When I got my first job at a newspaper in 2001 one of the long-time employees would frequently bemoan that we no longer covered family reunions. That kind of coverage went away, and in some way its loss likely played a role in the rise of social media.

Flash forward to today, as Google and Facebook continue to eat up most of the incremental digital advertising dollars, revenue losses are still impacting traditional media. But now, to make things more difficult, another threat has been identified.


The backbone of search and social media, is changing what, and how content, is being created. And they change. A lot.

Already in 2019 Google has tweaked its algorithm seven times. That’s in addition to 15 times in 2018. Facebook has only made four changes so far in 2019.

No matter how much they claim it’s not true or insist they are immune to it, every newsroom is impacted by these changes. In order for a newsroom to survive in 2019 they have to be. Content is no longer created in a vacuum with the tools of distribution only belonging to a few.

The type of content written, the way it’s written, where it’s distributed – all of it is affected by an algorithm. Some newsrooms, as the Boston University article points out, are even using algorithms to write content.

Algorithms have become a necessary evil. The shear volume content created every day makes it so. Without them the chance of your content getting found on the internet is slim-to-none. With them newsrooms are reluctantly giving over more control to outside forces and, by extension, run the risk of allowing for more confirmation bias.

Personally, I think confirmation bias will have the most devastating long-term impact on media and society at large. When asked I know most news consumers will say they want a center-of-the-road, fact-based, well-sourced story. Unfortunately what users want and what they actually engage with are frequently two different things. It’s why your newsfeed is made up of stories from the political edges of news sites and not Associated Press and Reuters.

media bias chart - ad fontes media
Souce: Ad Fontes Media Bias Chart Vision 4.0

In all but the most rare cases, in order for a piece of content to get found it has to produce some kind of engagement. In media engagement is one of those terms that isn’t anything, but can refer to whatever you want it to in order prove your point of view.

Engagement can be page views, shares, likes, tweets, follows, time on page, anything. It’s whatever you want it to be so you can say, “Hey, someone interacted with this content is some way, which means it’s great!”

In turn, the more engagement you get, the more your content gets found. This creates a self-sustaining feedback loop.

The hard part is getting that feedback loop started. You have to find that initial engagement. Unfortunately, thanks to algorithms, the easiest way to get the feedback loop started is to push content away from the center to either extreme of the political landscape.

The algorithms reward this engagement by pushing the content higher up, even to those users who consider themselves moderate. All that we end up seeing is content from the edges, giving us the impression that the world must be this way, and pushing us away from the other side.

This week’s links

I’ve grown tired of hearing Silicon Valley tech/media CEOs talk about how there are changing the world, while at the same time insisting nothing is there fault, but this week on the Recode Decode podcast Kara Swisher spoke with Reddit CEO Steve Huffman, which is an interesting conversation. He discussed the idea around quarantining subreddits instead of banning them and Reddit as a social media/media/technology company.

You can listen here.

Remember when Mic was the next great media brand that would rule the world? Mashable has a good article on its fall that you can read here.

Finally, Vulture did a story around Apple’s podcast strategy.

Weekend Media Threads, June 15, 2019

Digging into Mary Meeker’s 2019 Internet Trends Report.

As any media nerd can tell you the best report of the year is Mary Meeker’s Internet Trends report, which was released this week at Recode Media’s annual Code Conference. After last year’s report I wrote about print media’s continued revenue decline as tied to its loss in time spent (as many others did and continue to do). Here is this year’s slide, which includes a nice comparison to 2010:

Media Time vs. Advertising Spending. Slide 22 from Mary Meeker’s 2019 Internet Trends Report.

And here is last year’s slide as a reminder:

Media Time vs. Advertising Spending. Slide 96 from Mary Meeker’s 2018 Internet Trends Report.

It’s difficult to gauge the true loss year-over-year because we don’t know the true percentages. This is particularly true with print because the numbers are small. For example, if last year’s time spent was 4.4% (rounded to 4%) and this year’s is 2.5% (rounded to 3%), the YoY loss is 43%. But if last year was 3.5% and this year is 3.4% it’s just a 3% loss.

For arguments sake let’s use the round numbers in the report.

From the 2018 to 2019 report print saw a loss of 25% in advertising revenue share and a 22% loss in time spent. Which would make sense. If true, and it continues to track this way, the loss in time is likely to fall another 43% before it matches today’s print advertising revenue output.

Which means print advertising revenue will likely fall another 40-50% to somewhere between one and two percent of the ad revenue market. It’s difficult to imagine how legacy publishers survive this kind of loss.

Another interesting item from this year’s slide is that three of the five categories – TV, desktop, and mobile – all have the same percentage advertising spend as they do time spent. TV is at 35%, desktop 18% and mobile 33%. However, there is movement within each category.

In TV this is the second year in a row where the two columns were equal. Last year they were at 36%, now they are at 33%. A loss of 6% from each column.

On desktop time spent remained the same, but advertising spend fell from 20% to 18% to match the time being spent.

The big mover was mobile. Time spent increased from 29% to 33%, a 14% gain. While revenue increased from 26% to 33%, a 27% gain.

This leaves us with radio. Similar to print, radio if off kilter. Except in the opposite columns, which is likely not awful news. Time spent with radio has a 12% share, down from 13% last year. Advertising spend has an 8% share, down from 9%.

The question now is, will advertising spend continue to align with time spent, or will marketers move more dollars into mobile until it has a disproportionate spend?

This chart has been moving towards this equilibrium since it has existed, so my best guess is that the shares will move in sync with each other. The exceptions being print and radio.

Print’s revenue will continue to fall, maybe at a faster rate that it has been. To keep profits steady, print will slash more content, which will suppress time even more. Eventually both will settle in at 1 – 2%, but what will be left?

And radio? I haven’t followed as closely as print, but if I’m correct in saying that time spent and advertising spend are finding an equilibrium, I could see revenues stabilize as time continues to fall.

And time will continue to fall in traditional media.


Adults are constantly online …

… or ‘almost constantly,’ as this slide illustrates:

Percentage of adults online almost constantly. Slide 162 from Mary Meeker’s 2019 Internet Trends Report.

Overall adults spend 26% of their time online. Up 24% from 2015. More than a quarter of time spent online is a big number, but look at the groups 18-29 and 30-49. They spend 39% and 36% of their time respectively online.

The biggest jump from 2015 to 2018 are those folks in the 30-49 age range. They jumped from 28% in 2015 to 36% in 2018. That’s an almost 30% increase in just three years. I’m sure a lot of this is driven by folks in the 18-29 range aging into their 30’s, but it also speaks to how connected we’ve become to our technology.

To help combat the amount of time spent with our technology, technology companies (Apple, Google, Facebook, YouTube) are now offering wellness and time tracking tools.

Which is the definition of irony. Especially in the case of Facebook and YouTube. They’ve built technology and algorithms specifically designed to get you to spend more time on their platforms, not less.

But maybe it’s starting to work.

Percentage of adults online almost constantly. Slide 164 from Mary Meeker’s 2019 Internet Trends Report.

Time spent on social media is starting to decelerate. It’s still growing, just not as fast as it was. I’m sure after Facebook and Alphabet shareholders get a look at this slide those time watching tools will get tweaked to match market expectations, not user health.

And there is so much more

Seriously, this report has so much in it.

I know it’s 333 slides, but if you have any interest in the internet I highly recommend you take some time and page through it. It’s not just about media either. It gets into how the internet amplifies our own interests as well as bad actors, privacy concerns and regulation around the world, content moderation, the economy and on-demand jobs …

I could go on, but you need to look for yourself. Here’s the link, go …

Media Threads March 23, 2019

In the world of digital media there are a few companies that stand above the rest. Facebook and YouTube instantly come to mind, and get most of the headlines. It seems like each week at least one of these three is involved in another controversy. And this week is no different.

This week Facebook and YouTube came under heavy scrutiny after allowing video of the terror attack in Christchurch, New Zealand to be broadcast on their platforms. I’ll never understand how either company continues to allow things like this to happen. Now that advertisers are responding by boycotting, it will be interesting to see how quickly things get done.

Given the ongoing content distribution, as well as privacy, issues that most social media sites struggle with, I’m still surprised when I see publishers and brands that continue to spend more time and money on these platforms. Not only are these platforms no longer a safe environment for publishers and brands – in some cases they are downright hostile – they are a time suck.

I continue to push for publishers to move away from organic posting on social and into SEO and performance marketing. Instead of spending money on a team of people posting to social media, chasing algorithms, hire marketers and put money into performance marketing.

As users, at least in the United States, begin to shift from mass social sites into niche platforms, publishers of all sizes need to look at new methods of reaching audiences. Post and pray just won’t cut it anymore.

Finding Younger Audiences

Speaking of finding new audiences, here’s an interesting story from Digiday on how Nordic publisher Schibsted is attracting more young readers.

News publishers are scrambling to get younger audiences acquainted with their brands. Nordic publishing giant Schibsted, which pinned its hopes on the development of a new app two years ago, is now finding that it’s helping: 60 percent of the app users are under 25, and retention — measured by those using the app for the fifth consecutive week — has grown from 25 percent to 35 percent.

After seeing a decline in younger readers on Schibsted’s most widely read newspaper, Norwegian tabloid Verdens Gang, a team of 15 people — across editorial product and research — started working on a news app, Peil, targeted at readers between 18 and 25 years old in 2016, and launched it at the end of 2017.

It’s impressive that any publisher can find a solid audience in an app environment. People outside of publishing – and many of those inside – underestimate how difficult it is to build a brand via an app. It’s an extraordinary difficult task.

ISO: Publishers

Another story I’ve been following comes from another in the top tier of digital media – Apple. Apple continues to court publishers to try its new Apple News subscription service. Of course it’s being touted by Apple as the (next) savior of publishing. Reminding publishers about its success with Apple Music as proof.

During Apple’s meetings with news publishers, the company is reportedly pointing to its success with Apple Music to showcase evidence of its previous subscription success and convince partners to join. One source said that the company is pitching itself as a savior to the publishing industry, but some publishing executives have said that Apple’s logic is flawed.

“Based on our experiences with Apple Music, we’re very good at running a subscription business,” said one publishing exec, describing how Apple pitched the service. “We know how to build a subscription business, and we’re going to do that for news.” 

And while Apple is looking for sacrificial lambs partners, New York Times’ CEO Mark Thompson thinks publishers should be cautious.

“We tend to be quite leery about the idea of almost habituating people to find our journalism somewhere else. We’re also generically worried about our journalism being scrambled in a kind of Magimix (blender) with everyone else’s journalism.”

I’m in agreement with Mark Thompson here. I don’t necessarily think it’s a bad idea. As a news consumer I would love to be able to purchase a subscription, or even a single edition of my favorite local newspaper without going through the hassle that comes with the process.

As a publisher, I just think that enough digital companies have built their business off the backs of content publishers. We’ve seen this movie before, we know how it ends.