“I Would Like to Design”: Black Girls Analyzing and Ideating Fair and Accountable AI

Title of the Paper

"I Would Like to Design": Black Girls Analyzing and Ideating Fair and Accountable AI

Bibliographic Information

• Subject Area: Artificial Intelligence (AI) Education, Fairness and Ethics, Culturally Responsive Pedagogy
• Keywords: AI literacy, AI ethics, artificial intelligence, Black girls, design, fairness, accountability, STEM education

Research Background and Issues

• Problems and Challenges:

  • Algorithmic bias in automated systems exacerbates social inequalities, disproportionately affecting marginalized groups such as Black communities.
  • The U.S. K-12 education system lacks emphasis on the societal impacts and ethics of AI, particularly for underrepresented groups like Black girls.
  • Black girls are significantly underrepresented in the design of AI technologies, despite being frequently impacted by algorithmic biases.

• Significance:

  • AI systems are increasingly integral to the lives of young people, from social media to educational applications.
  • Identifying algorithmic bias and promoting fairness are critical for social justice; the participation of Black girls can drive the design of more equitable technologies.

• Research Motivation:

  • To address the research gap on how Black girls understand AI fairness and ethics.
  • To empower Black girls through culturally responsive computing education, fostering their STEM identity and positioning them as "agents of technological social change."

Solutions

• Research Methods:

  • Conducted three summer camp workshops and one after-school session with 10 Black girls in fifth and sixth grades at a school on the U.S. East Coast with a 99% Black student population.
  • Used activities and discussion prompts to explore participants' definitions of fairness and AI, encouraging them to creatively ideate AI solutions.
  • Data collection involved a case study approach, including interviews, activity outcomes, and material analysis.

• Innovative Aspects:

  • Focused on how Black girls conceptualize fair AI from their unique perspectives, addressing the specific gap in prior research on the "race-gender intersection" in algorithmic fairness.
  • Leveraged culturally responsive education theory and learning science methodologies to design workshops that consider the cultural and social contexts of the participants.

• Implementation Steps:

  1. Guided Black girls in defining and analyzing "fairness" through storytelling and discussion activities.
  2. Presented specific AI algorithm examples and discussed potential unfair scenarios.
  3. Encouraged participants to design the AI systems they envision for the future, discussing potential fairness challenges in their designs.

Research Outcomes

• Key Findings:

  • Most participants defined fairness as "equality" and "kindness," but had limited understanding of the more complex concept of "fairness as justice."
  • For certain AI cases (e.g., family subsidy recommendation algorithms), participants leaned toward equality in resource allocation rather than fairness accounting for socioeconomic disparities.
  • Most girls trusted algorithmic decisions over human ones, fundamentally believing algorithms could eliminate racial bias.
  • However, they also critiqued algorithms and expressed a strong interest in participating in the design process.

• Advantages Over Existing Solutions:

  • Unlike traditional technology education, this study emphasized the "social participation" and leadership roles of girls in AI design rather than merely teaching coding skills.
  • Offered a caring and culturally responsive approach, making the learning experience an extension of their personal growth, identity, and community culture.

• Experimental and Evaluation Results:

  • Even without a full understanding of AI complexity, participants creatively designed AI technologies to address specific problems.
  • For example, one participant designed an AI called "Hypnotism Place of Kindness" to address gender inequality in social or extracurricular activities.

• Limitations and Future Directions:

  • The small sample size and incomplete data due to the COVID-19 pandemic may limit the generalizability of the findings to all Black girls in the U.S.
  • Future research should optimize the timeframe of educational activities and provide additional technical guidance to support participants in transitioning from designers to implementers of technology.

Conclusion and Significance

This study demonstrates how innovative community workshops can guide Black girls to critically engage with the design of fair AI and explores how cultural factors shape their creativity. The findings provide valuable insights for developing socially oriented AI education tailored to underrepresented groups such as Black girls.