How Can Chatbots Help Us Respond to Humanitarian Crisis?

 

WFP/Rein Skullerud

Kaukuma refugee camp in Kenya (WFP/Rein Skullerud)

At the moment, The World Food Programme (WFP) and the wider humanitarian system are #FightingFamine in four countries. In Somalia, Yemen, North-Eastern Nigeria and South Sudan 20 million people are on the brink of starvation.

Our recent study “At the Root of Exodus” found that high levels of food insecurity lead to higher levels of migration across borders; UNHCR estimates that there are 65.6 million people forcibly displaced worldwide. The stakes are high, we need all the information we can get.

Who Needs a Humanitarian Chatbot?

Based on our previous experiences and secondary sources, we knew that some displaced people could potentially access our humanitarian chatbot. But, of course, the reality on the ground is incredibly complex.

They could be refugees who have travelled to a different continent or Internally Displaced People (IDPs) who’ve moved within their own country. Some may have just arrived in a camp or settlement a few months ago, others may be adults who were born in the camp and spent their whole lives there.

Within the same camp there might be different nationalities, languages, family demographics and education levels- the list goes on. These huge disparities mean that extensive in-country, context-specific research is needed to confirm any initial design assumptions. Only then can we really design the best tool to reach our target population.

This was very clear when we spoke to IDPs in Maiduguri, but also during our visit to Kakuma Refugee camp and Kalobeyei refugee settlements in northern Kenya where we conducted some focus group discussions about the possibility of using a humanitarian chatbot.

There are currently more than 200,000 people living in the two refugee settlements we visited in Kenya, including Somali and South Sudanese refugees, so we spoke to various groups in each of the settlements to get a broad overview of what different people felt could or couldn’t work.

Our research was therefore constructed around four fronts: phone usage, Internet usage, information needs, and a product review of the chatbot.

Refugee Phone and Internet Usage in Kakuma

First, phone and Internet usage. For people fleeing their homes phones can be their lifelines. They’re the way that they communicate with friends and family, get updates on the situations at home and even receive remittances or money from WFP.

It therefore wasn’t a surprise that a lot of the households, particularly in the older and more established camps, had access to a phone, and many also to the Internet. Kakuma refugee camp was established in 1992 so many of the young people we spoke to had lived their whole lives in the camp.

Here people told us that nine out of ten households had a phone (either basic or Internet-enabled) and they had good network coverage- encouraging news if we want to contact them with this channel! They confirmed our assumption that young people were most likely to have access to the Internet, either owning their own smartphone or by borrowing one from friends and used social messaging apps. But of course this wasn’t the only story.

The enormous increase in refugees means there are actually four different ‘Kakuma’ camps all of which are now full. To cope with ever-increasing numbers the Kalobeyei settlement was set up a few kilometres away to host new arrivals in an integrated settlement with the host community. The differences in people’s responses were stark.

Here people reported that only one or two out of ten households had handsets and it was very difficult and expensive to charge them. The young people also said they had Facebook profiles and knew how to use the Internet but they had no way to access it.

They pointed out that it’s now nearly impossible to get a Kenyan SIM card because as refugees they don’t have the necessary ID and aren’t eligible to legally work. Clearly whilst our assumption that a chatbot could be a good way of communicating with young refugees or IDPs was correct – there were some serious limitations.

Would Refugees Actually Use a Chatbot?

During our visit, we did not only want to check the feasibility of the chatbot and identify potential users, we also wanted to find out how to make it a) useful and b) interesting enough to hold users’ attention.

From our focus groups we learnt that refugees wanted to get more information about WFP programs. WFP already uses many different media like posters, hotlines, desks and even loudspeakers on cars to speak to people in the camps but a chatbot could act as a complement that sends outs updates immediately.

Refugees also really liked that they could have a direct feedback channel – they didn’t have to wait in line to send a message and WFP staff were happy they would have an automated way of handling these messages. Nevertheless there are limitations. Both staff and refugees reiterated that the chatbot isn’t the right communication channel to ask sensitive questions about security issues.

Two Surprising Findings

These interviews also gave us insight into some basic communication designs. Initially we thought that to reach the most people we needed to translate the bot into Kiswahili, Somali and Dinka (the most commonly spoken language by South Sudanese refugees). However, after talking to the refugees we soon realised there was no need as young people, our primary target audience, spoke English or Kiswahili.

In fact, the biggest takeaway was that the chatbot could be a springboard for other channels. By its nature a refugee camp is a very close-knit community. Even if not everyone has access to the Internet, young people said that they would still be able to share the messages they receive verbally with their friends and family.

Our trip to Kakuma therefore reiterated that context-specific research is the only way that we can produce a great communication tool for our target audience. There are challenges that we’ll explore in the next phases of our rollout, particularly as we explore using our chatbot in other countries and camps. Nevertheless, the future looks bright for the chatbot!

This blog was originally published on ICT Works 


Also published on Medium.

Leave a Reply

Your email address will not be published. Required fields are marked *