Dance and Creative Computing as Spaces for Critically Engaging and Learning About Artificial Intelligence/Machine Learning (Poster 10)

• In Event: Perspectives on Promoting Algorithmic Justice in Artificial Intelligence and Machine Learning Education in K–12

Abstract

Objectives: AIML systems are ubiquitous across many fields and the continued rise of these systems have also resulted in a rise in the ethical challenges resulting from how these systems are designed and implemented. Most of these systems have been shown to propagate and amplify inequities such as racism and sexism, among many others (Noble, 2018; Raji et al., 2020). Through creative dance computing, we attend to the aforementioned dimensions of inequity by developing our understanding of how we can equip learners to recognize and rectify issues of ML systems within culturally sustaining learning experiences (Castro et al., 2021).
Theoretical framework: We focus on enabling learners to engage with their identities and community practices—key components of culturally responsive computing learning experiences (Scott et al., 2015)—as they learn and reason about ML algorithms through dance. We draw on co-design (Roschelle et al., 2006) to facilitate and center learners’ and teachers’ values, ownership, and authentic contexts in co-creating curricula that connect dance with AI and creating space for discourse around how ML algorithms that process body position data can be shaped by human design decisions and their underlying biases.
Methods & data sources: We present two case studies from two co-design spaces where high school learners and dance educators used danceON (Payne et al., 2021), a web-based, free creative coding environment that uses pose detection algorithms and enables users to draw animations over user-produced videos. The first co-design space draws from sessions of a 15-week educational internship with six students where they used danceON to explore interactions between dance and pose detection algorithms as they produced expressive dance performances. The second co-design space draws from a series of design sessions with three dance educators where they used danceON and Google Teachable Machine to explore the behaviors, limits, and opportunities of ML models and algorithms in relation to dance poses and body movements and used their insights and their own teaching contexts to design learning activities for students across dance and ML.
Results: Students and teachers encountered various limitations with danceON’s pose detection algorithm as they created dance artifacts and performances with the system. Encountering and experiencing the algorithm’s limits served as opportunities to investigate how human design decisions and sociopolitical factors could impact users of danceON and similar AIML systems, and the various ways in which biases within these systems could manifest and cause harm and injustice across different communities. Students used their ideas around dance to explore how and why ML algorithms break and connected their ideas on algorithmic harm to their own experiences of how everyday ML-powered systems exhibited biases. Teachers also identified opportunities around creative ideation and production with ML as they designed learning activities to teach ML and computing through dance.
Significance: Our work demonstrates how we can meaningfully engage with learners and educators within dance computing spaces in ways that draw on their situated knowledge, practices, and experiences to co-create dance artifacts and educational resources that interrogate, critique, and teach about ML, computing, and dance.

Authors

• Francisco Castro, New York University

• Kayla DesPortes, New York University