Conversational AI (Technology that Talks) and Speculative Design Activism - Women Reclaiming AI
Conversational AI is a research project spanning three design activist projects, ‘The Infinite Guide’, ‘Women Reclaiming AI’, and ‘Looking for The Cloud’. These projects investigate how conversational AI is affecting human relationships with technology, the limitations of poor diversity in AI development teams, the lack of public understanding of AI and its environmental impact, and the ethical implications of the commercial pursuit of humanising AI.
Women Reclaiming AI (WRAI) is a feminist voice assistant and expanding piece of design activism authored entirely by women. The voice assistant is programmed through participatory workshops by a growing community of 100+ women. Through creating a platform for collective writing and editing the project co-creates an AI personality that challenges gender roles.
The voice assistant/chatbot documented in this item (through the ARS electronica exhibition documentation, website and video) forms one of the components of the output for this project. The rest of the documentation in this item is contextualising information.
The visual representation of WRAI uses GAN frameworks to generate purposefully disorderly images, made of a dataset of images of women, from our community and women the community admires. The GAN is made in protest of the averaging out of women’s experiences and the non-inclusive depiction of women by unaccountable development teams who are reinforcing negative gender stereotypes.
A GAN is a Generative Adversarial Network. It is an AI algorithmic architecture that uses two neural networks operating against each other to generate a new, synthetic instance of the data fed to them. In this case we fed the GAN a dataset of images of women, and the GAN then created new instances of the images that we then arranged in a grid and animated.±
Overview document courtesy of Coral Manton and Birgitte Aga.
Photos Courtesy of Jürgen Grünwald, Coral Manton and Birgitte Aga.
Used with permission.
The work is under copyright and may not be used without permission. Use of this repository acknowledges cooperation with its policies and relevant copyright law.
