Blame Machine Learning for Your Echo Chamber

One of the media’s trending topics is how echo chambers led to the almost unanimously inaccurate election predictions. Media outlets started by blaming themselves, but have lately shifted the blame to social media. They have a point, to a degree.

Echo chambers are created when people surround themselves with others who validate their own opinions, and mass media outlets tend to create large chambers by leaning one way or another. However, social media doesn’t necessarily lean left or right… it leans you. It runs machine learning algorithms on massive data sets to deliver all your favorite aspects of your friends and news in a conveniently digestible feed. As South Park phrased it, it creates your own personal safe space.

Let’s dive a little deeper into how this works. Machine learning is actually based on how our brains work, and we naturally tend to categorize and label things. One particular algorithm called k-means clustering is quite good at categorizing. The graph below is a visualization of k-means in action. The algorithm identifies clusters and neatly segregates them into colors.

Imagine this is a graph of how social media accounts fall across two quantifiable attribute scales: love for cats vs love for the NY Knicks (I totally made those up… I’m not actually sure what this graph is plotting). Go ahead and look for your account– the dot that’s closest to your preference. Your account will most likely be surrounded by other dots of the same color. Social media calibrates to your tastes by showing you the cats/Knicks opinions of the color that you align with. Because social media gets you to come back by overfeeding you content that you love to like and share.

Going back to the graph, each color becomes its own echo chamber, segregated from other echo chambers. As a result, you’re not going to see the opinions of the other colored dots very often. Of course it’s not this two-dimensional, though, because it’s personalized across millions of attributes. This means we’re not only clustered on cats and Knicks. We’re clustered on countless other fronts.

Through clustering, we’re blinding ourselves from other opinions and reinforcing our own without being fully informed of other perspectives. Machine learning has enabled us to unknowingly ignore the diversity around us. It’s bad enough being uninformed about a topic you’re passionate about. It’s far worse to falsely believe you’re fully informed on that topic.

The Importance of Diversity

Diversity reaches far beyond race. In NY Real Estate, Fair Housing Laws protect from discrimination based on race, creed, color, national origin, sex, age, disability, marital status, military status, family status, and sexual orientation. And the concept behind affordable and inclusionary housing is to encourage mixed income neighborhoods.

From this past election season, we’ve seen that the country is fractured on many more issues. But it’s awesome that there are so many opinions across so many topics. Much like how biological evolution arises from genetic diversity, diversity of thought accelerates ideation and innovation.

I believe one of the reasons that America has such a strong innovation culture is that we have pockets of extreme heterogeneity in our major cities. Below is Dustin Cable’s racial dot map, which plots a colored dot indicating the race for every person living in the US (This map also made my list of NYC’s 10 Coolest Public Real Estate Data Sets and Visualizations.) Note that the more densely populated areas also tend to be more colorful.

Many other cities around the world are more culturally homogeneous, creating natural echo chambers that make it harder to solve problems through different perspectives.

It’s Not All Your Fault but You’re not Off the Hook

We shouldn’t just sit back and wait while social media companies adjust their news feed algorithms (in fact, Facebook published a study that said their feeds don’t create echo chambers at all- read and decide for yourself). And the news will always have some degree of subjectivity in the reporting. There’s a lot that we can do to counteract echo chambers and media bias.

At the city scale, there are a lot of policies in place to encourage diversity and inclusion. Infrastructure also plays a huge part by facilitating multi-modal transportation and information dissemination. Cities should be designed to safely encourage cross-cultural pollination.

Businesses can also create an innovative culture by designing work spaces to maximize idea-sharing. Apple designed their new spaceship campus to enable chance encounters.

At the individual level, take advantage of these opportunities to make a conscious effort to be more informed about the opinions you hold. There’s a lot of freely accessible information out there, but parse through what you consume and understand the biases (a good litmus test for bias is how emotional an article makes you). And instead of passively discovering content, actively research topics you hear about. Leave your safe place. Listen to other opinions. Learn from one another.

Wait a Minute, this Article is Biased

Of course this article is biased. It’s my opinion, formed from my own experiences and research and topped off by a controversial title. If you’re interested in the topic, here’s some more reading to get you started. Conduct some research of your own, and share your thoughts in the comments!