Mpathy

A Facebook bot that recognizes harassment and offers solutions

Comment

Inspiration

Mpathy was inspired by the growing problem of online harassment in the United States over the last few years. While our software capabilities have continued to grow and evolve, we have not yet seen a solution that is both effective, but also discrete.

What it does

Mpathy emulates what Facebook's M is capable of as far as event planning and conversations go, and uses Facebook's AI innovation as a means to tackle online harassment. Mpathy listens to your conversations at your discretion and determines if you are the victim on cyber-bulling or harassment. Once it has detected a situation that involves harassment, it offers the victim a set of quickly available options to deal with the aggressor.

How we built it

Mpathy was built using the facebook-chat-api, which is a third party tool that allows developers to interact with Facebook's chat system, as well as the PerspectiveAPI, which was used to determine whether or not situations of harassment were present.

Challenges we ran into

Implementing proprietary APIs was difficult! Before turning to a third party API for Facebook chat manipulation, we struggled and failed to find a way to either emulate Facebook's M AI, or manipulate the chat system. Obtaining API keys caused us a lot of stress as it was a game of simply waiting and watching a lot of our very short time go to third parties.

Accomplishments that we're proud of

Mpathy is responsive to a very colorful amount of commands, and offers a great range of solutions. One solution we though we wouldn't be able to demo was the ability for Mpathy to selectively filter messages from users. We were both proud and excited to be able to integrate this capability.

What we learned

We all learned how to use a brand new set of software tools. All of us came from different backgrounds in our hacking capabilities, and we all managed to advance our capabilities here.

What's next for Mpathy

Full integration with Perspective API, as well as multi-user interaction. Facebook comment thread parsing, and moving on to different mediums as well.

Share this project:

Updates