Inspiration
Our research shows that people are never just bullies or victims, but it's a spectrum of behavior characterized by levels of empathy. The lack of nuance enables the dehumanization of victims in online interactions; that dehumanization directs the passive observation and that causes bystander apathy online, amplifying bullying.
We learned that improving the presence of emotion through visualization in the cold text of social media made it easier to asses people in conversations. Increasing contrast in social media through visualization can increase the efficacy of self-assessment.
What it does
The Rehumanize Project is a chrome extension that uses sentiment analysis of social media to visualize the emotion in online interactions and reduce harassment. Using the Watson sentiment analysis, we analyze social media messages past and present through the user's browser to determine the tone; we elaborate on both the emotional tone through mood-hue and emoji & stickers as the user creates or consumes messages. The Rehumanize Project gives a metaphorical Jiminy Cricket conscience to the user to better understand and react to what they read and write.
How I built it
The Rehumanize Project comprises a couple AWS Lambda functions to fetch and analyze data, a Chrome extension built with Javascript to extend Facebook, and a static website to host our resources and Twitter analysis tool. We used the IBM Watson Tone Analyzer API and the Twitter API to assist us.
We used the Serverless framework to deploy our Lambda functions. The Serverless framework makes it easy to tie together Lambda, API Gateway, and other AWS services, so you can focus on writing your application code, and not worrying about infrastructure. The two endpoints we created are for analyzing the tone of text, and for analyzing the tone of a person's tweet history.
The Chrome extension is made with vanilla Javascript and extends facebook.com with attention-directing sentiment indicators. We want to draw attention to good content, and make users aware of potentially negative content. The Inside Out emotions accompany posts that embody those emotions. The goal is to increase awareness of online message's effects and connotations.
Our static website is a clean, modern Bootstrap website. It is built with HTML, LESS/CSS, Javascript, and Gulp. We are hosting it on AWS S3 and hooked up our domain name to our bucket. The website serves as the primary way for users to find relevant resources for their cyberbullying experiences and to access our Twitter historical report card tool.
Challenges I ran into
We tried to translate the visualization into generatively created typography messages, but the toolsets were to limited, so we substituted emotions and parametric-determinative color relationships to the tone analysis findings. Facebook has an interesting
Accomplishments that I'm proud of
We finished! We didn't give into feature creep! We worked as a team with a variety of skill-sets and levels. Most importantly, we made something that didn't forget the humanity of bullies or censor to hide the problem.
What I learned
Bystander Apathy is expressed in unique ways online, and machine learning is hard. Facebook's code is difficult.
What's next for Rehumanize
Build applications native to the social networks to seamlessly offer. Build generative typography visualization tools to better convey online messages. Involving and enabling schools and institutions that can use our tools to help the efforts they're already working on. Prove the institutional tie in and scale to other institutions while letting the project scale naturally digitally.
Log in or sign up for Devpost to join the conversation.