Post-Anthropocene Poetics [De]Fused

Anne Sullivan & Anastasia Salter

Artists' Statement

In this piece, we critically reflect on societal gender representations and biases that are embedded within the data sets and AI code that are presented as nuetral agents. We created a poetry generator which uses identified "gendered" terminology[1] to generate short poetic works, drawing on existing encoded binaries and underlying assumptions that the algorithm magnifies. These short texts have been curated by the artists and then used as input for a text-to-art AI to generate fantastical paintings. Supporting Safiya Noble's claims in Algorithms of Oppression[2] and the many critiques of AI bias that have raised similar concerns, the work invites viewers into a better understanding of machine learning through creative visualization, these pieces are able to create visualizations of the systemic gender representations and biases embedded within the data and code.

The piece consists of two elements: the poetic Tracery[3] generator, which runs in the browser live to create new poetics here; and the pre-generated illustrated poems, which cycle through demonstrating the visualization of the post-anthropocene poetics (as seen above) using Disco Diffusion[4], which builds on human creativity and thus inevitably "dreams" through a gendered lens. These generated imagines demonstrate how the visual rhetorics of existing work are embedded into every iteration of our machine learning co-creation, as choices as seemingly netural as landscape or portrait are loaded in their influence on the outcome.

1. Gaucher, Danielle, Justin Friesen, and Aaron C. Kay. "Evidence that gendered wording in job advertisements exists and sustains gender inequality." Journal of personality and social psychology 101.1 (2011): 109.
Noble, Safiya Umoja. "Algorithms of oppression." Algorithms of Oppression. New York University Press, 2018.
2. Compton, Kate, Ben Kybartas, and Michael Mateas. "Tracery: an author-focused generative text tool." International Conference on Interactive Digital Storytelling. Springer, Cham, 2015.
3. Rafael-Patino, Jonathan, et al. "The diffusion-simulated connectivity (DiSCo) dataset." Data in Brief 38 (2021): 107429.