This was a challenge given by Itaú's Hackathon "Batalha de dados" (Data Battle), December 2018, where we were put together in a group of 5 people, being 1 UX Designer, 2 Data Scientists and 2 Social Techs, and had the challenge to solve a problem using data in 30h.

As a group, we choose to solve the problem of violence towards women, as Brazil is one of the most violent countries for women.        
My role
This project was made with fellow colleagues: Henrique Marques (Data Scientist), Karin Colantonio (Social Tech), Mariana Valle (Social Tech) and Rafael Monteiro (Data Scientist).
I assumed the role of the UX Designer, and my contributed to ideas and benchmarking, making personas, user flow, prototyping, writing, user interface and user experience design.

Analysis & User Research

We first started to research what was the most dangerous place for women, so we could focus our project. Before studies, we thought most dangerous places would be open ones, like dark streets or public transportation. But what we found out was that the most dangerous place for a Brazilian woman is their own home.
With astonished data we discovered that:

of the aggressions are committed by their husband 

of the aggressions are committed by an ex-partner 

of femicide in 2017 were committed by the partner or a family member 
(UNODC survey);

of serious aggressions happened at home 
(DataFolha Survey).
So, after facing these data, we decided to focus our project in solving the violence inside home.

With the benchmarking we saw there is already great solutions to help women and support them in an emergency moment or after an aggression has occurred.

However, we realised there is no solution to prevent cases of violence, and help women to recognise the first signals of abuse and know what to do. So that was the case we decided to focus: preventing violence against women

Based on the researches gathered by the Brazilian Federal Senate, we set up two personas so we could focus on them before start ideating concepts.

After analysis, problem statement and benchmarking we started to trace possible solutions. We thought about landing pages with pages in social media, apps and a messenger bot. We made a quick analysis to understand what would be a better approach to our solution.
After the analysis, we decided to go for a messenger bot, because it is easy to interact with, since most people in Brazil use Facebook frequently. So it requires less learning processes, does not need to install any further application, and it works well in desktop and mobile devices.

Valentina - The Bot

So, we designed started to design a Messenger Bot to be presented as a friend to the user. We gave her a female name, Valentina,  that means "valent" and is a current trendy name in Brazil. That way it makes women feel safe when chatting to the bot, and does not drives attention if the partner has access to the women's phone.

The Bot's Purpose

Valentina's purpose is to prevent either way of violence, being psychological, physical, domestic, to happen; and to give support to women to get away of relationships that is turning abusive.   

How a bot will prevent violence?

Using data from the base of the Brazilian Federal Senate, we created a probabilistic model based on random forest, which uses python, and machine learning, to identifies characteristics of women who have already experienced domestic violence, and predicts whether a woman may be at risk.
Valentina will deliver what we called "content pills" to woman chatting to her. This content will be wither text, images, videos and links to websites, mainly about relationships. Presented to make the woman aware and recognize possible signs of an abusive relationship. 

Valentina will also deliver information about what to do and where to find help if the situation escalates and also empowerment content.

The conscious use of Data 

Beyond using data to identify women to predicts woman at risk, the Bot will collect some data by sending quick quizzes and simple questions, based on lean data survey techniques. 

In that way, the women's privacy is secured, and they'll be only answering specific questions formulated to personalize the experience, and to generate more data to answer gender violence questions, and understand better this issue, so more public politics can be asked.    

First Prototype

In the hackathon we build a high fidelity prototype to be presented, using Telegram as a base for our bot. The bot was trained with demographic information, such as age, location, income, that we collected from a survey from Brazilian Federal Senate, that has has rich data about domestic violence.

Bellow there is a video recorded showing an user interacting with the bot. This video was presented as a deliverable together with the codes created.  

User Flow

After the Hackathon, I decided to keep going with the project. So I started by writing an user flow based on the first prototype and feedback gather in the presentation. This user flow would allow me to understand the steps of an user interacting with the bot.  

Second Prototype - Testing User Scenarios

Following the processes, I decided to make a second prototype, exploring the possible interactions offered my Messenger, and also exploring ways to write and the bot's responses.

In this prototype I wanted to test the following scenarios: first interaction with the bot, the user receiving the messages, the user receiving a survey question, the user and the bot chatting, the user receiving contents and the user asking for help.

The Bot’s looks and Visual Style

For the "final version" I adjust the text messages and content based on the second prototype and user scenarios. In addiction I designed a brief style guide for the bot, based mainly in images and text tone, once the Messenger interface is not much manipulable when it comes to user interface.
I choose to give a face for the bot, in a way to make it more next to the experience of chatting to a friend, and it looks more like a usual chat with a human being. The model chosen was next to the average Brazilian woman, as most women identify themselves as mixed race.  
The logo was specially designed for us by the Hackathon Designers, following our briefing. From the logo I extracted the main color. I applied the purple as the accent color in the photos presented in the instructional cards and in the overall photos used for Valentina.
Valentina has a friendly personality, using informal language and being positive and helper with the messages exchange.

Final Prototype


Our intention in this Hackathon was to create a product that could collaborate with a real problem faced by most brazilian women. We hoped to build a project that could work in preventing violence and also use the collected data in a way to complement the existing data and contribute to feed the databases about Public Security.

Being in this Hackathon gave me the opportunity to For the future, I would like to test this prototype with real women to collect feedback and push the project further to a real application.

Back to Top