Mar 13, 2014
Trend ID Algorithms: What Communicators Need to Know
You may not realize it, but much of what you see online is determined by the algorithms that power search engines and social networks. Designed to surface the information that is most compelling, and likely to get you to read the article/view the video/take the survey – and then share it with your friends – algorithms are doing more than serving information. They are shaping journalism and arguably, having a negative impact on democracy, according to Heidi McBride, a senior member of the faculty of the Poynter Institute and Gilad Lotan, chief data scientist for Betaworks, during the discussion they lead at South by Southwest.
“Trend identification algorithms are all over the web,” Lotan stated. “We have to think about the power they encode, and that is the power to draw attention.”
Learn more about how the digital evolution has impacted newsrooms and journalism by viewing our on-demand webinar, “The Evolution of Media: How Newsrooms are Adapting to the Ever-Changing Digital Environment.”
On web sites everywhere, data scientists are using algorithms to find and display content that is likely to draw readers and inspire social sharing (thus drawing more readers.) These numbers have an economic impact – after all, Google, Facebook, CNN and the New York Times are all add-supported, and more visitors to their web sites (and visitors who stay longer) equal more ad impressions, and thus, more dollars.
In building the trend ID algorithms, data scientists are looking for trends away from the norm.
“We look for spikes, things out of the ordinary, outliers,” noted Lotan. “And activities around celebs spike much more dramatically than other conversations.”
There’s a self-reinforcing effect as the journalism companies respond to the algorithms, as the algorithms have an economic effect on the journalistic companies, effectively steering news coverage.
Lotan reminded us that algorithms are created by humans, and thus may reflect their creators’ biases or preferences. Additionally, he noted that algorithms can be selectively manipulated, citing as a case in point changes Twitter made when Justin Beiber was constantly trending, causing user complaints. The team changed the algo, making it more difficult for Beiber to trend. Another example of selective algorithm manipulation happens on the search engine side of the house, such as when Google penalized JC Penney for poor SEO practices by dropping the Penney website to the bottom of the rankings heap.
What’s needed, McBride and Lotan posited, is more public understanding of how these algorithms work, and more transparency from the companies employing them.
“The companies that control our attention to so without any transparency,” McBride stated. “We build our understanding of ourselves and the world around us through the stories we tell, and if algorithms only reinforce certain types of stories, it reduces our understanding of ourselves and our communities.”
The session did offer one important tactical take-away for brands. Stories take hold fast and algorithms reinforce this. If a problematic story is gathering steam, swift response is absolutely essential. The more quickly you can correct information, the more quickly the entire news democracy can reference that information as the topic trends. But if you miss the gap, your message will be left by the wayside.
Click to register for the our free on-demand webinar
You can also read more from this session on The Guardian’s extensive recap.