Using Quantitative Ethnography to Model How Black Girls Develop Sociocritical AI Literacies

• In Event: Overcoming Master Narratives: Unforgetting, Reimagining, and Reconstructing Through Quantitative Ethnography

Abstract

Objectives
This study investigates how Black middle school girls develop sociocritical AI literacies--the capacity to interrogate, critique, and reimagine artificial intelligence systems (Arastoopour Irgens et al., 2022a). Amid growing concerns about bias and exclusion in AI, this research centers Black girls’ creative agency in technology learning environments that often marginalize them. The core objective is to examine how scaffolded activities involving critique and design can foster both technical understanding and sociopolitical awareness.
Theoretical Frameworks
The study is grounded in Critical Race Technology Theory (CRTT; Tanksley, 2023), which challenges assumptions of technological neutrality by revealing how anti-Blackness is encoded into digital systems. This framework encompasses recognizing algorithmic racism, centering Black lived experiences, and advocating for abolitionist technologies. This study also draws on “homeplaces” (hooks, 1990)—emotionally affirming spaces that support resistance, healing, and Black joy.
Methods
This QE (Shaffer, 2017) study was conducted in a STEAM-focused middle school for girls in the U.S. Fifteen students engaged in the 10-hour critical machine learning (CML) curriculum (Arastoopour Irgens et al., 2022b), which encourages students to build AI tools while interrogating their social impact. Researchers collected and coded 2,607 lines of discourse across classroom sessions, interviews, and student artifacts using a CRTT- and CML-informed codebook. These codes captured students’ technical understanding (algorithmic bias, dataset quality, model functionality) and sociopolitical reasoning (systemic racism, representation, human values in design). Epistemic Network Analysis (ENA; Shaffer et al., 2016) was used to measure and model student data.
Findings
ENA modeled how students connected technical and sociopolitical dimensions of their developing AI literacies and revealed three learning phases. In “Interact and Critique,” students linked algorithmic limitations to biased training data. In “Critique and Reimagine,” discourse expanded to include historical and systemic oppression, critiquing dominant ideologies embedded in AI technologies. Students connected exclusionary AI systems to lived experiences, including biased image searches and misclassifications in medical algorithms. In “Reimagine and Build,” students developed classification models and robots rooted in cultural identity and diverse personal interests, such as Black hairstyles and ballet.
Significance
The benefits of using QE were multifold. First, the two-dimensional mathematical space enabled temporal analysis across the curriculum, revealing three overlapping phases of learning supported by statistically significant comparisons (Figure 1a). Second, the weighted discourse networks visualized how technical critical consciousness was not a linear acquisition but a
dynamic interplay of emotion, identities, knowledge, and design (Figure 1b). Third, ENA allowed for both broader claims about students’ developing sociocritical AI literacies and for investigating variations within the group. For example, the mean networks revealed strong connections between technical and sociopolitical content. However, when viewing individual girls’ discourse networks, some girls connected between technical aspects of algorithmic bias and a commitment to social justice, while some girls connected more between technical aspects of algorithmic limitations and acknowledging technological racism.
Through the affordances of QE, the findings advocate for curricula that blend sociopolitical critique, emotional expression, and joyful creation. By showcasing Black girls as critical technologists and imaginative designers, this research redefines AI literacy as technically rich, critically engaged, and identity affirming.

Author

• Golnaz Arastoopour Irgens, Vanderbilt University