By Naomi Chen
Ryerson computer science PhD graduate Shaina Raza published a study in July on her news recommender system (NRS) research, which proposes a new model that enhances content diversity and combats user boredom due to consistent information.
News recommender systems provide people with the most relevant information based on their individual interests and preferences.
Today’s news is shared faster than ever before. From real-time breaking stories online to trending topics on social media, people are loaded with an overwhelming amount of information every day. An article in the Journal of Information Science notes that 2.5 quintillion bytes of data are produced on the internet daily, and it would take the average user 200,000 years to read all the information produced online.
With limited time and shorter attention spans, readers are becoming highly selective about the content they’re consuming.
Raza said the problem is that current news recommender systems confine people to a narrow window of content.
Remaining in echo chambers can lead to issues across the spectrum; “political polarization is another one of the examples,” said Raza. During times like the federal election cycle, this can directly affect voting decisions and have tangible political influences.
With limited time and shorter attention spans, readers are becoming highly selective about the content they’re consuming
Another example is the anti-vaccination movement during the pandemic. According to Raza, algorithms prompt content that help the anti-vaxx community reinforce their beliefs against the vaccine. Since these individuals reject vaccinations, they will continue to see articles that align with their values.
Raza said this is why news access needs to be diversified. “If you listen to the news on only one website or TV channel, over time you will be taking in their viewpoints and agreeing with what they say,” she said. “It’s important to break this kind of filter bubble.”
Raza’s newly-proposed model demonstrates a framework that enhances news recommender systems. Search histories often trigger algorithms to suggest similar content to users. The new model addresses user boredom from consistent content by providing them with new topics and new information. Simultaneously, by broadening the range of articles shown to users, it expands viewpoints outside of their confined sets of values.
Under the supervision of Ryerson computer science professor Cherie Ding, Raza concluded that although people enjoy seeing content that appeals to their interests, “algorithmic accuracy does not guarantee a better user experience.”
The study illustrates that there is a negative correlation between accuracy and diversity for the user experience. Raza‘s proposed model strikes a balance between the two. Users will still receive relevant information based on their preferences, but will also see content reflecting additional perspectives.
According to Ding, the study recognizes that algorithmic accuracy remains important for users as they appreciate familiar information; hence, the new model’s combination of accuracy and diversity.
Some examples of additional diversity can include stories in the user’s geographic area, ethnic background or those that carry other political viewpoints, said Ding.
“If you listen to the news on only one website or TV channel, over time you will be taking in their viewpoints and agreeing with what they say”
Since big technology companies such as Google and Amazon rely on customizing content based on user algorithms, “there was a lot of discouragement when I first introduced the topic,” said Raza. “At one point I was even feeling depressed because no one would listen to me,” she said.
Since the paper was published two years after she introduced the topic, Raza has finally gained acknowledgement. “I got some invitations from networks to express my views on why I decided to work on this research and what are the findings,” she said.
For Raza, her recommended approach for NRS also helps increase cultural awareness.
As an immigrant to Canada, Raza said her proposed model of news recommender systems would help Canadians to educate themselves on different ethnic groups.
There’s often false misinformation circulating about certain minority groups, but many still form their opinions based on it, she said. “There’s no way to break that [cycle]—that’s what prompted me to do this research.”
A revitalized news recommender system could decrease stereotypes against certain cultural communities, Raza said.
She also said that news recommender systems should “nudge the readers, give them awareness at the individual level and be careful” when it comes to echo chambers.
Raza is currently working on research about COVID-19 information and misinformation. Given more time and collaboration opportunities in the future, she said she would love to continue her research on news recommender systems, specifically its impact on ethnic or religious groups.