How can we collect authentic chat data from young people in an ethical way? Can AI learn how adolescents communicate and use that knowledge to make health messages more engaging?
To understand how adolescents communicate, we need good data. At the same time, asking people to share chat conversations often raises privacy concerns.
Responsible ánd fun
To tackle this challenge, as part of the SocialMovez project and in cocreation with young people, we developed BetweenUs: a chat game that allows adolescents to donate chat conversations for research in a safe and anonymous way. Inspired by games such as Among Us, participants engage in one-to-one and group conversations, enabling authentic language data to be collected in a playful environment where privacy is paramount.
The game was developed in collaboration with adolescents. They contributed ideas on the visual design, the avatars, conversation topics and how the game works.
AI-assisted personalisation
Using this chat data, we trained AI to adapt health messages to the language style of adolescents. The model learns to recognise patterns in word choice, tone, sentence structure, and the use of emojis. It then rewrites messages without altering the substantive information.
This enables us to investigate whether health communication becomes more engaging and relevant when it sounds like the language of the target audience for which it is intended.
Try BetweenUs for yourself by clicking the link below (Replace ‘yourname’ with a name of your choice). The game starts automatically as soon as four players have logged in.
