At Ushahidi, our software is always built on the idea of “meeting people where they already are.” We build software for crowdsourcing, and while we have our own mobile app, we firmly believe that our mobile app or our website shouldn’t be the only way someone can submit a report to Ushahidi. The key is to reduce the barriers to get someone to report. As such, the Ushahidi software has a number of datasources integrated into our platform. A “datasource” is the way we refer to the channels from which you can crowdsource information, or the means by which people can submit to Ushahidi. The datasources we currently have are SMS, twitter, email, embeddable web forms, and an online/offline smartphone-app. For quite a while, we have been tempted to integrate Facebook as a datasource, but the privacy structures around personal pages made this difficult. Over this past few months, that changed. For the Kenyan 2017 election we ran the Uchaguzi election monitoring program, and working in partnership with Facebook, we built a Facebook Messenger chatbot that integrated into Ushahidi. There are over 7 million Facebook users in Kenya, so this was a huge opportunity to engage a big part of the electorate.

This chatbot connects to the Uchaguzi Facebook page, and it replies and starts a conversion when you start talking to it. Building a chatbot has amazing potential for Ushahidi users, because it structures the incoming data (the way you can make a survey have required fields that have to be filled out before you can submit), vs Twitter, which pulls via hashtag or keyword, and then has to go through a manual structuring workflow process. With the chatbot, we made the bot ask the user to choose what survey type they wish to report, ask them to add a description of what they want to report and, if they want to, share a photo and their location.

The response was fantastic! Users clearly wanted to use Facebook as a datasource. 3000 people interacted with the bot during the election week. Half of them had something they wanted to share with us and this resulted in 923 structured posts which we still are working on validating and publishing on the Uchaguzi-site. We learnt a lot from the massive user-response we got and will use this to further refine and improve how we can integrate Facebook as a datasource into the core Ushahidi platform in the future.

The main lesson we learned was that the users want to start chatting with the bot straight away. We had a user flow that required users pressing a button to start reporting and answer the questions asked. We didn’t make this button-requirement to start clear enough, and as such some users sme struggled to get their message through, causing the bot to answer with a response like “OH MY, I am not programmed to understand what you are saying!”. We are redesigning this process, and for those of you who struggled to get the bot to understand you but still reported in your testimonies: we have extracted your conversations and all the things you wanted to share with us and loaded your messages as posts into the Uchaguzi-site for further validation.

Another interesting thing we saw was the ability to follow reports and updates during a longer time. For example one user reported he and other officials were stranded with election-materials and struggled to get transportation, only to later come back to us and report that they finally got transportation and the issue was solved. This kind of information and ability to follow a report and updates on what is happening during a longer timespan is very valuable and help the moderators of a deployment to track the emergency and validity of reports.

We believe in reaching people where they already are. There are over 2 billion Facebook users, so this clearly is essential to achieve that goal. But not only is it incredible to integrate Facebook into Ushahidi, the experience of integrating a chatbot to dramatically improve the tedious manual process of structuring is a game-changer for Ushahidi deployments dealing with massive amounts of data in time-sensitive situations like elections or crisis response. This allows those precious human hours to be focused on tasks that humans do better, like calling to validate or escalating an issue to the appropriate response teams.

After building and running this experiment, you can certainly look forward to future features with Facebook Messenger and chat bots.

Sincerely, The Ushahidi team.