Addressing Algorithmic Justice in an AI/ML Curricular Module for High School Students (Poster 6)

• In Event: Opportunities and Challenges in Teaching Algorithmic Justice in K–12 Computing Education

Abstract

Objective: The Computer Science Frontiers (CSF) project introduces teachers to emergent CS topics including AIML through curricular modules that are designed to engage minoritized students, especially females. The pedagogy uses project-based learning, pair programming, societal implication discussions, and programming activities that connect real-world issues to relevant cutting-edge technologies. Here we describe the teacher PD and pilot of the AIML module designed with the goal of raising awareness about AIML among high school students, especially issues of algorithmic justice and bias
Theoretical Framework: The AIML curriculum draws on the AI4K12 framework (Touretzsky et al, 2019). Early in the curriculum, students are introduced to interrogations of bias and ethics in AIML which is then a throughline in all the activities. In the AI & Drawing activity, students explore Google’s Quick, Draw! app. They examine the datasets and watch a video that explains how gathering millions of drawings helped Google train its ML models. Students examine how gendered biases crept in the training of Quick, Draw! through watching a video and discussing how such bias could be mitigated or avoided. Students also interrogate the ethics of creating an app like Quick, Draw! that collects data from unsuspecting users. They also discuss watch and discuss The Truth About Algorithms. Activities in the ‘Ethics and Real-world Applications’ unit students examine specific real-life examples of algorithmic (in)justice in facial recognition, targeted advertising, and the role of AI in the criminal justice system.
Methods & Data Sources: 6 female (2 Black, 3 White, 1 Asian) and 2 male (1 Black, 1 White) high school teachers participated in this study. Our team facilitated a week of PD based on the Teacher-Learner-Observer (TLO) model (Goode, Margolis, and Chapman, 2014). In this peer-teaching-centered model, participants spend time playing the roles of teacher and learner while the PD facilitators observe and guide a structured reflection after each TLO session (Catete ́ et al., 2020). The goal of this PD was also for the educators to co-design and refine the curriculum (Grover et al., 2020). For the AIML activities on ethics and bias, teachers researched and shared additional examples of algorithmic injustice from news stories. At the end of each day during the PD, participants completed a debrief survey where they answered questions about how their TLO sessions went overall and suggested changes to the activities. Teachers then co-taught a one-week-long summer camp in pairs. Camp participants were recruited from high schools in regions across Tennessee and North Carolina (56% female and self-identified as Asian (65%), White (15%), Black (9%), and Other (11%)).
Results & Significance: During the TLO activities, teachers found the materials to be eye-opening, and engaging (especially the Sentiment analysis activity). Their pilot camp with students yielded positive results as reflected in students’ “takeaways” from the camp. The results suggest that presenting AIML curricula with a socially-relevant focus helped students engage in issues of algorithmic justice in AIML.

Authors

• Shuchi Grover, Looking Glass Ventures

• Isabella Gransbury, North Carolina State University

• Veronica Catete, North Carolina State University

• Tiffany Barnes, North Carolina State University