While working at OpenLab I have been involved in the WhatFutures project with the International Federation for the Red Cross (IFRC), exploring how large-scale conversational data generated via a large-scale game, played entirely over WhatsApp, can be interpreted and visualised.
IFRC have volunteers spread across many countries, often in rural areas where it is hard to reach and engage with them. However, the IFRC are key to explore new ways to engage with their volunteers to that their experiences on the ground can feed into the overarching strategies of the IFRC.
Daniel Lambton-Howard devised a playful and innovative and large-scale multiplayer future forecasting game, played entirely over WhatsApp. It is designed to engage IFRC’s youth volunteers in sharing their hopes and fears for the future, and to include these voices in shaping IFRC’s Strategy 2030.
The game was piloted in June 2017, where over 400 young volunteers from 5 National Societies played the game. Collectively, they sent over 30,000 messages and produced over 90 video, audio and written future forecasts to help inform Strategy 2030. Read more about the results and media content produced.
The game was successful at attracting a diverse audience directly and maintaining a substantive dialogue. The topics discussed ranged from more practically and locally focused concerns to wider societal issues and reflections on environmental and technological issues.
I processed the data in Python. To summarise the key topics the volunteers talked about, I used:
Topic modeling (LDA)
n-grams, keyword and key phrase extraction
I navigated the topics and n-grams/key phrases in context to find related quotes and used yEd graph editor to create the diagrams.
The pink nodes are quotes associated with that theme.
The orange nodes are key themes that were talked about by the country groups.
The size of the orange node provides an indicator of how talked about that theme was (larger node = more discussion on this topic).
Size of orange nodes or `talked about score’ based on two factors automatic topic modelling & most frequently mentioned words/concepts.
Green nodes highlight country specific concern: local issues and opportunities and challenges.
Multi-lingual dataset: players used English predominantly but some groups switched to their native language in the discussions, leading to a mixture of Bulgarian, Finnish and
Social chat: the game also provided a unique social forum for volunteers to get to know one another and share experiences. While this data is interesting in its own right, it was not helpful for honing in on the substantive topics discussed.
Game co-ordination: as the players had time limited challenges to complete a considerable portion of the data relates to co-ordinating the process of remotely collaborating. Again, while interesting, this added a level of noise to the data for accessing the more topical content.
Locating country specific content: Volunteers from different countries were brought together in the discussion spaces, so we needed to extract the contributions from the different country contexts, whilst not sacrificing the sequential coherence.
WhatFutures is a collaboration between Open Lab, Newcastle University, UK and The International Federation of Red Cross and Red Crescent Societies (IFRC). The game concept was designed by Daniel Lambton-Howard.