Pulling Back the Curtain on Personalization Algorithms

Algorithms are everywhere. They power social media sites like Facebook, search engines like Google and increasingly digital news outlets that want to show users more of what they like and less of what they don’t.

Algorithms rarely get attention, except for when they are in the news for allegedly having a left-leaning bias or other (usually negative) reasons. But they should be a regular topic of conversation in college classrooms. As an article I wrote for Digital Journalism argues:

Personalization algorithms, widely used by digital media sources, filter and prioritize news in ways that may be unapparent to users. Savvy media consumers should be aware of how this technology is used to tailor news to their tastes. This two-part study examines the extent to which U.S. college students are aware of news personalization, and the actions and criteria that affect news selection and prioritization. Interviews with one set of students (n=37) focus on the news sources they use most often to begin a news search. A subsequent survey given to a second set of students (n=147) focuses on Google and Facebook, two influential gatekeepers. Results show that students are largely unaware of whether and how news sources track user data and apply editorial judgments to deliver personalized results. These studies identify aspects of news personalization that warrant greater attention in college curricula.

I presented my research, boiled down big time in poster form below, at the 2016 AEJMC conference in Minneapolis.

Screen Shot 2016-06-21 at 1.29.58 PM

Screen Shot 2016-06-21 at 1.50.40 PM

Here’s my elevator pitch (OK, it’s a long ride — maybe the Empire State Building?) for why this is such an important topic:

Endless streams of information have become so central to the ways in which news is consumed that it is tempting to assume that content is selected and ranked based upon some universal standard of editorial importance – or to ignore how these decisions get made altogether. To be savvy media consumers in the digital age, students need to understand the implications of news personalization and how the sites they most commonly rely upon for news and information use their data – and data from millions of other users – to filter the news they consume. While past research and news reports have focused on identifying some of the ways in which news is personalized on sites such as Facebook and Google, little attention has been paid to the extent to which users are aware of the concept of personalization and the specific types of user data that are tracked by personalization algorithms.

This exploratory, two-part study attempts to fill that void. Taken together, these studies shed light on what students know and do not know about news personalization and how educators can tailor lessons about this topic to areas about which students are least informed.

The headline from the study is that students know very little about news personalization. While this is troubling, it’s also an opening for journalism educators to find creative ways to cover this topic in class.

This was also the topic of discussion during a four-hour pre-conference workshop that I organized at the conference. Here’s the official program information:

1 p.m. to 5 p.m.
(w6) Teaching Algorithmic Transparency
$10 Fee
Teaching college students about the power of algorithms should be a central component to mass communication. This panel session will provide educators with an overview of algorithmic transparency and advice on how to teach these concepts. For additional information contact Jennifer Kowalewski, Georgia Southern University, (912) 478-0126 or at jkowalewski@georgiasouthern.edu (MCSD)


Studying Student Civic Engagement in an Election Year

With the 2016 presidential election in full swing, journalism educators may soon begin thinking about how they can talk to students about their engagement with politics, and use of new and traditional media to track election results. A study I conducted, split into two parts and published separately, examines youth civic engagement during an election year. One part, “Political Engagement During a Presidential Election Year: A Case Study of Media Literacy Students,” was recently published in the Journal of Media Literacy Education. From the abstract:

This exploratory, mixed-methods study uses data gathered during the previous U.S. presidential election in 2012 to evaluate student political engagement and digital culture. Survey results and media diary entries revealed that college students enrolled in a media literacy course during Super Tuesday or Election Day gravitated toward low-barrier political actions and expressive modes of citizenship, and they were most engaged when there was a social component to following election news. These results, coupled with recent data on political engagement and media consumption, present an opportunity to consider the role of digital platforms and online communities in the 2016 election.


To quote our research questions:

RQ1: To what extent are media literacy students engaged in politics during a presidential election year?
RQ2: How do students demonstrate dutiful and expressive forms of political engagement?
RQ3:How do students use traditional and new media to track and share political news?

We found that students were far more engaged during the general election than during the primary and were interested in expressive modes of political engagement, largely thanks to opportunities to connect with people in their personal networks through social media, digital technologies and election watch parties. As the paper found:

Data suggest that one of the predictors of youth political engagement is whether a political issue or candidate is perceived to be of personal interest and

benefit…This study suggests that another key predictor of civic engagement during a presidential election year is whether a political event or campaign has a social component and has interest for students’ peer groups…

According to this study, college students today are not engaging in the kind of classic grass-roots community and political action as generations past. Politics for students today is first about reaching out to engage with their own personal networks. This study suggests that community has become an anachronistic word in an era of social media. Students do not use the term. In its place the term “network” has emerged. That shift in language suggests not only the technological underpinnings of today’s communications but the geographical deracination of a sense of community. Students do not think about “building” a community as much as they are concerned with “maintaining” their network.

Back in 2012, while I was lead teaching assistant in a media literacy course at the University of Maryland, I led a team that conducted this study. We asked a range of questions: How do students demonstrate political engagement? What mediums and news sources do they prefer to follow election coverage? What topics most interest them?


The focus of part one of the study appeared in the December 2014 edition of the Journal of Digital and Media Literacy. To quote our research questions:

RQ1: To what extent do students rely upon information push and pull to track political news in a presidential election year?

RQ2: How do students come across political news in an ambient news environment?

The short answer to RQ1  is that students during a major political event still rely primarily on information pull — that is, actively seeking out information rather than expecting that it will find them through push notifications or word of mouth. A major reason why: Watching live election coverage while discussing results with friends remains extremely popular.

The short answer to RQ2 is that students are awash in news. It is all around them (thus the ambient part). Facebook and Twitter are first alert systems that help students follow political news without having to pay full attention.

My takeaways? (Again, quoting the article):

This study shows that at a time of information abundance, when technology and social connectedness make it easier than ever to adopt a news-will-find-me mindset, students tracking a major political news event still seek out news as it is breaking. The increasing popularity of news apps and the ubiquity of cell phones is one explanation for the prevalence of students “pulling” news. Around Election Day, however, a significant number of students went beyond quick news searches on their smartphones in between classes. Across all time blocks, they investigated claims made on social media, researched political issues or candidates that came up in conversation, and played with electoral college maps.

The main storyline emerging from Super Tuesday was that students took little interest in an election in which few had a voting or a partisan interest. The vast majority of participants did not self-identify as Republicans. They commonly referenced feeling disinterested in news about the primaries because they did not see any personal relevance. The predominance of information pull was more a product of the lack of word-of-mouth conversations or buzz on students’ social media feeds than a sign of widespread active engagement. Put another way, Super Tuesday news did not come to—or find—students because people in their personal networks generally did not push news toward them or create an ambient news environment filled with election chatter. The few students who were motivated enough to follow Super Tuesday news found it on their own.

The ways in which fall students tracked political news in the thirty-plus hours before polls closed and the day after the election—tracking live blogs, scrolling through news feeds, texting with friends—is indicative of the modern media landscape. Students used social media and, to a much lesser extent, push notifications from news outlets to monitor trending stories and track commentary, particularly first thing in the morning as a daily tip sheet. When news found students, it was more commonly through social channels (Facebook, Twitter, word-of-mouth communication) than through official channels (news blasts from media outlets). Fall students sought out news of interest and relied on their social networks to remain peripherally aware of most political news. With news already surrounding them and with some complaining of information overload during the peak hours of general election coverage, students showed little interest in yet another way of receiving news: signing up for push notifications (i.e., opting in).

The ways in which fall students followed political news in the hours immediately after polls closed was strikingly traditional—gathering in front of a television to track election results. In the fall, 46% of students watched television in Tuesday primetime and 43.5% of these viewers only watched television to get news during this time slot. These findings show that Election Night is one of few national events in which appointment viewing is widespread and viewers give their full attention to one screen. Even with the explosion of digital media technologies, television still plays a central role in connecting audiences during a major political event. Results speak not only to the primacy of television but also to the limitations of social media, the use of which slightly decreased as television viewership rose. While users can monitor a flood of political news and commentary with little effort on social media, they cannot easily witness a live event of significant importance. Social media regained its primacy in the wee hours of Tuesday and early Wednesday, as students turned to (and often were turned off by) analysis from those in their social networks.

The long answer is in the study. Take a read — it’s online and free.

Teaching Analytics in the Classroom and Newsroom

Last summer, I wrote about the rise of engagement editors at news outlets across the country. These hybrid positions (part editorial, part business) typically include social media and SEO strategy, examining analytics and strategizing with top editors about other ways to engage audiences.

I learned, not surprisingly, that when you write about engagement editors — people who are very good at generating buzz — your posts get plenty of attention. But this was more than just a conversation happening in an echo chamber. Readers, including people with no discernible experience in newsrooms, seemed very engaged in the topic. I began to wonder what role universities can play in helping to train future engagement editors.

That’s the primary question I address in a recent MediaShift piece, “J-Schools Try Different Approaches to Training Aspiring Engagement Editors.” I focused on journalism schools that adhere to the teaching hospital model in which learning happens in the classroom and newsroom. Many university-operated, student-powered newsrooms are teaching students about analytics dashboards and social media management tools. The University of Maryland is creating a new bureau for its Capital News Service focusing exclusively on social journalism and audience engagement. City University of New York has launched a master’s program in social journalism. Students at the University of Florida and Arizona State University track audience data for affiliates of NPR, PBS, etc. that are based on their campuses.

When university newsrooms teach metrics and engagement, what is the real goal? That’s the main question I answer in part two of my MediaShift series on university news services and teaching engagement, “Practicing What They Preach: J-School News Services Track Engagement and Impact.”

From the article:

Using analytics can help students evaluate and improve their journalism. But it can also throw up roadblocks through low readership numbers, chasing metrics before grasping the basics of reporting or making broad assumptions from limited data. The result is that college newsrooms — training centers, outlets of record and content distributors — are still debating what audience engagement metrics and other measures of impact to use.

Professors who run these newsrooms argue that page views are not the most important audience metric. They often prefer advanced measures of engagement such as share of returning visitors, time spent on site and social media traction — and they want students to think critically about what metrics matter most to them. Some stress that quantitative metrics are not the only way for student journalists to measure audience engagement. “If we declare victory, it’s going to be where our stories brought action,” said a CUNY journalism professor.

There are limitations when measuring offsite metrics. When stories are republished, how do audiences engage with that content and what kind of impact does it have? It’s an important question for university-operated news services that regularly feed student work to news outlets with a much broader reach, but also a question no one can answer.

However, professors say they try to focus on the metrics they have, not the ones they don’t have.

As an Arizona State professor noted:

“I want [students] to get a really clear picture of how content works on one site. Adding someone else’s site introduces so many other variables into the equation. Would it be great to have that data for every single story? Absolutely. But I just chalk it up to, well, if they can’t get the entire picture then at least they can get a really clear picture of the story’s performance on our site.”