CONTENT

1. Ethical Foundations

  • Definitions of ethics in data science
  • Transparency: explainable models and open pipelines
  • Accountability: auditability, reproducibility, and traceability
  • Bias mitigation principles (procedural, distributive, and intersectional fairness)
  • Real-world harms: predictive policing, credit scoring, and health algorithm bias
  • Responsible Stack
    • Data Layer: provenance, consent, inclusion in data collection (feminist principles)
    • Storage Layer: energy cost, cloud jurisdiction, green infrastructure
    • Processing Layer: low-energy compute, scheduling, decentralization (the cost of connection)
    • Modeling Layer: fairness-aware modeling, robustness, explainability
    • Deployment Layer: edge vs. cloud tradeoffs, latency vs. sovereignty (accessibility and digital sovereignty)
    • Governance Layer: logging, access control, human-in-the-loop

2. AI & Environment

  • Environmental cost of training large AI models (GPT, BERT, etc.)
  • Cloud computing and water/electricity consumption
  • Carbon emissions tracking (tools: CodeCarbon, Green Algorithms)
  • Carbon-aware scheduling and green load balancing
  • Sustainable DS pipeline design strategies

3. Data Sovereignty

  • Digital colonialism and extractivist data practices
  • Legal and regulatory frameworks (GDPR, Indigenous Data Sovereignty protocols)
    • GDPR: data minimization, portability, consent
    • EU AI Act: risk-based classification and governance
    • UNESCO AI Ethics: proportionality, inclusiveness, sustainability
    • OECD AI Principles: transparency, robustness, accountability
  • Rights-based frameworks and Indigenous protocols (e.g., OCAP)
  • Community ownership and participatory governance
  • Hybrid sovereign clouds and data localization policies
  • Culturally-sensitive and respectful data collection protocols

4. Fairness & Bias

  • Sources of bias: historical, representational, measurement
  • Fairness metrics:
    • Demographic Parity
    • Equal Opportunity / Equalized Odds
    • Predictive Parity
  • Algorithmic auditing tools and techniques
  • Inclusive dataset design and rebalancing strategies
  • Bias-variance-fairness trade-offs in model evaluation

5. Feminist & Decolonial Data Science

  • Feminist critiques of objectivity and “neutrality” in tech
  • Technopolitical critiques of extractive data infrastructures
  • Data feminism principles (e.g., make power visible, consider context)
  • Epistemic violence and knowledge representation gaps
  • Feminist infrastructure and makerspaces as alternatives
    • Trade-offs in DS design: accuracy vs. fairness vs. energy use
    • Pareto efficiency and fairness frontiers
    • Composite scoring metrics (e.g., weighted sum models)
    • Fair resource dispatching in cloud systems (Dominant Resource Fairness)
    • Scheduling for diversity, sustainability, and transparency
  • Decolonial methodologies and care-based design

6. Conclusions and Open Challenges