Research

Generative AI on messaging apps and the digital good

This project aims to understand what people want when using generative AI on messaging apps such as WhatsApp or Telegram. It will pay special attention to the perspectives of communities traditionally marginalized from technological design processes.

After ChatGPT’s launch set records for the fastest-growing user base (Hu, 2023), companies rushed to develop their own AI products and integrations. Messaging apps, as key players in global communication, were no exception: Meta AI, for example, is already available on WhatsApp in most countries. Considering that WhatsApp is estimated to have nearly 3 billion monthly users, such integrations are set to have far-reaching consequences.

However, little is known about the actual preferences of users regarding specific implementations of Gen AI. Furthermore, AI is often fine-tuned by small, non-representative groups of individuals (Kirk et al., 2023). When more people “participate”, it is often unknowingly – that is, when their data is harvested from social media sites. This may result in technologies that cater to a small, wealthy subset of the global population.

This research is leveraging previous feedback on generative AI from the PRISM dataset, which involved 1,500 participants from more than 75 countries. It will employ a range of methods in order to gather in-depth information about what good uses of GenAI on messaging apps look like from users’ perspectives. It may also prototype new applications.

Team

Dr Scott Hale, Digital Good Network: Emerging Methodologies Lead, Associate Professor and Senior Research Fellow at the Oxford Internet Institute

Julia Sepulveda, Research Assistant, Digital Good Network. Oxford Internet Institute, University of Oxford.