Chatbot builder

1 /5

Client

AIDA was developed by Manas.Tech and InSTEDD, in collaboration with mVAM, an innovative team at the World Food Programme (WFP), the food aid arm of the United Nations, and thanks to a grant from the Cisco Foundation and additional support from the Korea International Cooperation Agency (KOICA).

AIDA (Artificial Intelligence Development Assistant) is a chatbot builder that allows you to create interactive chatbots that engage populations on popular chat apps like Facebook messenger or Telegram. Chatbots built by AIDA can send surveys, respond to specific intents, run periodic outgoing check-ins with users and improve their interaction over time by learning from them.

2 /5

Requirements & Context

The WFP regularly deals with the challenge of effectively delivering information to families that are in hard-to-reach areas. Maintaining fluid communication with their beneficiaries ensures WFP can best serve its mission of fighting hunger by improving access to sufficient quantities of affordable and nutritious food. The mVAM team uses mobile technology to enable WFP to share critical information with its beneficiaries.

Hunger map

3 /5

Approach

After multiple visits to affected communities over 2016 and 2017, we confirmed that using a chatbot had the potential to address the communication challenges. However, each humanitarian context where WFP works is unique, and thus each situation requires a chatbot with specific skills and abilities.

We spent some time with our WFP colleagues in the Kakuma and Nairobi Offices, brainstorming ways in which the chatbot could complement existing activities and provide useful information about their work. So, an ambitious goal was set for AIDA –to allow humanitarian staff in diverse on-the-ground situations to create their own chatbots, so that they would meet local demands. The first pilot in the series of chatbots was the Food Bot, designed to communicate with families in the Kakuma Refugee Camp in north-west Kenya.

AIDA’s backend is built on Elixir, its frontend is built on Ruby on Rails and React.JS, and it uses a PostgreSQL database.

Testing with real users at Kakuma Refugee Camp, in Western Kenya (WFP/Lucia Casarin)

4 /5

User research

KENYA

To get to know our users better and start defining the design of the Food Bot, we travelled to the Kakuma Refugee Camp, where we spent a few days collaborating with WFP staff and youth leaders in the refugee camp to understand how to create a user-friendly chatbot to meet their needs.

We first worked with a small group of people, to understand how they used the technology. We employed a popular prototype technique called ‘Wizard of Oz’. This type of human-centered approach allowed us to quickly learn what types of information the Kakuma refugees were interested in receiving, as well as how they asked questions.

During the field test, we also confirmed our hypothesis that chatbot conversations need to be as light as possible (not using too many pictures, menus, or emoticons) in order to minimize data charges and make conversations possible for low-bandwidth network. To reach as many people as possible, we decided to create a bot that would operate on a popular messaging app, like Facebook Messenger, so people could take our surveys on a platform they already use.

5 /5

Results

The kenyan-pilot chatbot shares information on food assistance availability and simultaneously collects beneficiaries’ feedback and complaints. It complements existing WFP non-digital channels such as hotlines, help desks, posters and loudspeakers in the camp.

Furthermore, chatbots built by AIDA work on both the regular and lite versions of messenger apps, which are often used in low-bandwidth settings. It also allows camp managers to schedule messages and surveys, with customizable message reminders, and it allows for surveys to be filled directly without leaving the chat screen.

AIDA is being designed by humanitarians, for use by humanitarians. This means it will have unique capabilities not found in other commercial chatbot builder tools. For example, AIDA can train chatbots to follow and respond to conversations without the need for existing natural language processing technology, which requires a chatbot to be built in widely spoken languages like English and French. This ability is critical because the languages spoken in places with humanitarian crises are often not supported by mainstream AI packages.

Analytics

To keep the tool evolving and improving, it has an analytics module built-in, that displays key performance indicators, which can also be used for reporting.