Informing about fake information and media bias in automated news
Developer Student Clubs are university communities of students from all backgrounds who are passionate about Google technologies. We work on local community problems to solve them using technology through innovation.
Collaborators Tanvi Karwa
The Problem Fake news and media bias in automated digital content in news outlets
01
Robo-journalism (automated articles) is increasing fake news online
02
Automated articles tend to be based on foundational biases of reporters
Social media plays a significant role in online news consumption and spread of fake information
Through automated journalism, workflows are being automated and through crunched data, bots are able to generate news articles.
More than 50% of the global population uses social media and it is often the first place to find daily news rather than specific news outlet websites. Social media is also the easiest destination for ads and marketing, making it a prominent place for fake news.
Understanding Users
Trust and usage of online news content varies across age, gender, culture, and internet speed
67% Gen Z users depend on social media for news, as compared to 47% matures (above 55 age). There are contrasting differences in behaviors between generations and I classified my personas based on user behaviors.
The five ways people consume content
Focused consumption is consuming one form of content on one device. Dual consumption is when users consume content from more than one device. Information snacking is using wasted time to consume content. Time shifted content is postponing consumption of content. Content binging is when users consume multiple parts in a single session.
New mediums of content search
There is a shift in trend of the medium of content. Younger generations are using visual platforms more. Overall 62% internet users prefer consuming video content every day. In the last one year, there has been significant increase in use of voice and image recognition for search.
The Solution
Browser-based plugin to detect and classify potential fake information based on logistic regressions
Any text based information is broken down to paragraphs, sentences, and words to create tokens which are analyzed in comparison with similar keywords from trusted sources to create an immediate report to identify bias or potential of fake information.
Automated articles generated by bots lack author attributions, back-links, and sources, which are retraced to find similar articles and confirm the veracity of the facts.
Beyond detection
Classifying and organizing bias information
Any text based information is broken down to paragraphs, sentences, and words to create tokens which are analyzed in comparison with similar keywords from trusted sources to create an immediate report to identify bias or potential of fake information.
Automated articles generated by bots lack author attributions, back-links, and sources, which are retraced to find similar articles and confirm the veracity of the facts.
Mayen Dashboard
The full report presented in a dashboard includes a bias meter to describe instances of different biases that may be due to political inclinations, terminology problems (like phrasing of the headlines), negativity bias, etc.
FINAL DESIGN
Conveying our transparent process
I created a landing page for the process and marketing of the plugin. The aim for the website is more than a place to know about features and install the plugin because for a brand that promises true information, transparency of our process is a major concern.
Reflection
Takeaways and next steps
With the rise of automation, there will be clear increase of fake information. I have always been interested in ethics in technology and this was a very insightful process of learning in how human perceptions can transform to fake information through machine bias or limited algorithms. I enjoyed how automation can solve problems in automation and I believe there is greater need of similar implementations to detect erroneous information transmissions.
While this is impactful to detect fake content on browsers, a majority of consumers use apps and websites on smartphones and I believe the next step is to create solutions for the mobile.
