Human-driven question answering systems have been deployed by digital service providers for many years. Take for example, text-based approaches such as online forums as well as audio-based services such as Google Neighbourly or IVR systems like the Spoken Web. While machine powered smart-speakers from the likes of Google and Amazon are rapidly increasing in popularity and reach, it is the value of combining the richness of human-curated question responses and the efficiency of machine powered responses that inspired us to develop TalkBack; a human centered AI based question and answer system for use in public emergent user settings. There are two elements to the TalkBack design: the local installation with which the users interact, and the cloud-based back-end (which provides machine responses and crowd-sourced answers).
Studio Hasi facilitated deployment of TalkBack prototypes in 9 shops in Dharavi during the month of March and August of 2019, We recruited a team of 10 answerers from the nearby communities to answer to the questions forwarded using the TalkBack mobile app. We collaborated with a team of senior HCI researchers at Computational foundry, Swansea University to conduct user interviews and amalgamate the findings into a research paper.
Hardware specifications, code and accompanying app are available on GitHub.
Published in Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. Read research article