Over the course of the past decade, social media has become a primary source for spreading information. Due to this, misinformation has become easy to spread and social biases can become amplified in a never-ending chamber of shouted opinions.
There are two primary biases at work here: cognitive and social biases. The former happens when people read something, misinterpret or misremember it, then share it to more people as they remembered it incorrectly to be. The latter is a result of years of social interactions and nurture, which causes people to act differently towards different groups of people.
The problem of misinformation and biases extends not only to people, but likewise to AI that try to learn from groups of people online -- they're more vulnerable because they often have no way of checking against what they're learning!
To present a solution to this ever-growing problem, I made VerifyMe, a mobile app that tests your ability to detect fake news, helps identify your social biases by analyzing your responses to common situations, and provides a search utility for news articles while informing you if that news is fictional, an unverified claim, or factually checked.
What it does
The app user is presented with three options: Test their ability to identify misinformation, test their social biases, and perform simple keyword searches to see what the most relevant and trending results are.
How I built it
TidyUp was created using Xamarin, Hoaxy, and the Microsoft Azure Text Analysis service. Xamarin provided the framework for creating the application for mobile devices, permitting us to write the app in C# using .Net packages. Hoaxy provided a curated database of articles online that could be used to test whether people could identify fake news or not. Microsoft Azure Text Analysis services allowed the social bias test to be based purely on natural responses to commonplace situations; depending on the positivity or negativity of the user's response, social biases could be identified.
Challenges I ran into
The biggest challenge was an unexpected loss of data on the morning the project needed to be submitted. While some of the data was recovered, the app will have to be almost entirely rewritten. The only remaining element was the demo on one of my test devices.
Accomplishments that I'm proud of
I was proud to see how well the app provided news for the user to classify and how elegantly anyone could be tested for common social biases. Designing and implementing a system that could teach people two-fold how to overcome the torrent of complex media in today's society and maintain an open mind was an excellent accomplishment.
What I learned
This was a fantastic learning experience for me; I had never worked with sentiment analysis or automatically curating massive datasets of information before. Despite the last-minute issues, it was a lot of fun to make this project.
What's next for VerifyMe
The first step will be to redesign the app from the ground up, using everything I learned before the data loss to streamline the project. From there, major improvements can begin: Using AI to generate new scenarios for the user to explore based upon detected biases, visualization of the dissemination of news by tracking its source and showing the user where it comes from, and Facebook/Twitter integration to analyze how biased the user's news consumption is.