Context: UNHCR Northern Europe have phone lines staffed by interns with the legal unit, 3 days per week. The team do an excellent job fielding questions about family reunification and individual asylum applications, requests for information about finding family members, and referring urgent cases to UNHCR colleagues in the field, and around the world. Callers are often redirected to other, national agencies or legal partners that can deal with cases in a more personalised manner.
Process: Our first step was to get an understanding of the calls to UNHCR Northern Europe — the profiles of callers, the kinds of questions they ask, and the information they receive. We interviewed staff, and phone operators/lawyers, and together we mapped out the ‘red-tape’ — the things they cannot say, the advice they cannot give.
“When I’m on the phone, I am the voice of UNHCR, so I need to be particularly careful that I don’t say something that the organisation hasn’t approved” — UNHCR legal intern
Designing for trust is paramount in this context. Designers must ensure that users trust the information they are given, and understand that it is particular to them and their situation. However, in order for the service to gain an understanding of the users’ circumstances, users would have to trust the service enough to provide it with some personal information in the first instance.
“How might we design a conversational interface that promotes feelings of trust in difficult circumstances?”
At this point, we started examining the information UNHCR Northern Europe wanted to provide users, and prototyped typical conversations based on our research so far. After a few long nights sifting through legal documents, and creating maps of the chatbot’s typical conversations, we had a basic template that we could work with.
We tested it with the legal staff at UNHCR and iterated and tweaked it based on their feedback. By the middle of week two, we had a finalised ‘map’ of the information that users could navigate conversationally.
We explored a variety of different bot builders to deploy the bot online and decided to go with Motion.ai for the ease-of-integration they offered for websites. They have a nice interface for building bots, although we still prefer to use Textit.in for larger projects, particularly on mobile devices.
Result: UNHCR Northern Europe launched their first chatbot on their website on November 14th, 2016. The bot handled between 20–25 conversations each day.
We also used Motion.ai’s data analytics to provide UNHCR Northern Europe with feedback on their chatbot, who was using it and what information was most requested.
“Social Projects gave us the tools to quickly open a new channel for talking with refugees and asylum seekers. In over just two weeks they demonstrated that there are many innovative ways of making social impact and that the necessary technical expertise is available for institutions to harness” — Zoran Stevanovic, Snr Regional Strategic Communications and Advocacy Offfcer, UNHCR Northern Europe.
It was great to work with a major humanitarian actor such as UNHCR and give them some insight into how this technology can make their work more efficient. We look forward to continuing to explore avenues for future collaboration with UNHCR.