Considerations for a More Ethical Approach to Data in AI: On Data Representation and Infrastructure

Area of Science:

• Artificial Intelligence (AI)
• Data Science
• Computer Ethics

Background:

• Centralized data infrastructures have historically dominated AI development, leading to issues with data representation.
• Increasing AI integration into daily life necessitates improvements in data sourcing and usage.
• Ethical concerns, such as selection bias, arise from current centralized data approaches.

Purpose of the Study:

• To survey data representation and infrastructures in AI.
• To explore democratic data management approaches in AI research.
• To identify key considerations for ethical AI development.

Main Methods:

• Mini-survey on data representation and infrastructures in AI.
• Outline of four key considerations: auditing, benchmarking, confidence and trust, explainability and interpretability.
• Analysis of decentralized networking for AI data management.

Main Results:

• Decentralized networking offers greater transparency in AI data infrastructures.
• Transparency can alleviate core ethical concerns, including selection bias.
• Key considerations for data-driven AI were identified.

Conclusions:

• Reflection on auditing, benchmarking, confidence and trust, and explainability/interpretability is crucial for data-driven AI.
• Improved interdisciplinary discussion can mitigate data-based AI ethical concerns.
• Addressing these issues can ultimately improve individual wellbeing when interacting with AI.