Companion Chat 


Can AI remove human bias on victims while
reporting sexual harassment and rape?






AI enabled chatbot for victims of sexual assault



In an attempt to eliminate human bias for millions of sexual assault victims, Companion Chat is a chatbot that asks victims of their conditions to guide them to the necessary trauma-coping and law enforcement resources. The system asks users a series of questions and uses natural language processing to perform sentiment analysis and keyword extraction on these responses. Based on the aggregation of these results, a variable known as the threat constant is developed which is based on the z-score distance of individual cases from the aggregated distribution of sentiment scores and keywords, and it is used to guide victims to adequate resources. The entire system is built using the Messenger SDK for Facebook, Node JS, and MongoDB. We are open-sourcing the data sets derived from the results for their extensive research potential in social sciences and law enforcement.

Read more about the project here.︎︎︎


CONTEXT
AI For Good Summit 2018
CATEGORY
Strategy, Design Research, Digital Product Design,
Storytelling, Illustration, UX Research, UI
COLLABORATOR
Adi Sidapara

TIMELINE
15 weeks




THE PROBLEM

We determined that there are two main scenarios in which such events go unreported:

1. Victims are aware of their assault and scared to report it as they fear human-bias and re-victimization

2. Victims aren’t aware of the severity of their situation and aren’t sure about the kind of action they could be taking in these scenarios. 




Ilustrations are the cradle of human imagination and intellect. The Imaginarium of Technology(iMT) believes that engineering starts from dreams, dreams that are made of pictures each worth a thousand words. This illustration was my first attempt to conceive, capture and refine Companion Chat, in collaboration with Adi Sidapara, through pictures in order to communicate the idea, solution and the application of artificial intelligence for social impact.
The “A.I. for Good” idea illustration challenge held by iMT and XPRIZE was an opportunity to expand the possibilities of AI and provide a platform to communicate this vision to an influential audience at A.I. for Good Global Summit︎︎︎ held in Geneva ini June 2017.

GOALS 

  • An easily accessible chatbot that asks victims of their conditions to guide them to the necessary trauma-coping and law enforcement resources.
  • Attempt to eliminate human bias and re-victimization of potential victims of sexual assault.
  • Creates conversations that provide trust, comfort and confidentiality
  • Seen as a portal for accurate and certified information.
  • Distinguishes itself as not replacing humans, law enforcement professionals or medical professionals.
  • Create datasets for exploration of rape culture among different demographics.
  • Allows for better utilization and deployment of existing resources



MARKET RESEARCH
Currently the National Sexual Assault Telephone Hotline | RAINN takes an average of 30 minutes to get connected with a hotline volunteer which includes initial conversations and additional holds to be connected to support specialists.

Screenshot from RAIN website indicating wait times to speak to an advocate

CONVERSATION RUNDOWN   



Mark