Embracing Responsible AI

Social well-being and Planet

Robustness & Stability



Privacy & Safety

Fairness & Equity

Models with uncertainty estimates to improve robustness

Set standards and assign responsibility for AI-generated content

Employ diverse and representative training datasets

Develop standards for transparency and disclosure

Carbon aware computing

Set data privacy policies

Document data sources, model architecture, and training procedures

Interpretability techniques such as feature importance etc.

Techniques like Reinforcement Learning from Human Feedback

Adversarial training to protect against attacks

Regularly audit AI models for fairness and equity

Sustainability reporting

Data augmentation, data compression, and selective data sampling

Data augmentation, reweighting, or

Watermarking, metadata embedding for attribution

Guidelines to generate synthetic data

Error Analysis and Validation

algorithmic adjustments

Minimize PII data collection and ensure secure handling

Strategic roadmap to scale


Ethics Committee

Privacy by Design

AI Engineers

The Dynamic Duo: Synergy of human behavior and AI systems After laying out the fundamental principles, attention should be directed toward the actions of humans and AI systems. When considering behaviors, areas for consideration include the following:


Contestability is the ability for individuals and stakeholders to challenge or contest the decisions, processes, and outcomes generated by AI systems. For instance, if you applied for a loan and your application got rejected, it is important to have the right to contest that decision. The RAI framework should promote contestability as a desirable behavior. Users should be free to raise concerns or contest decisions related to ethical challenges or sensitive topics. Empowering users to contest AI outcomes is crucial for maintaining transparency and fairness.


When it comes to enterprise adoption of AI systems, the goal should not be to implement them for the sake of doing so. Instead, careful consideration and strategic planning are essential to ensure that implementing AI benefits your business. In today's tech-savvy world, users have become overly reliant on technological artifacts, making decisions primarily influenced by technological developments. Henceforth, it becomes crucial to prioritize human-centric AI systems, catering to their needs as the mainstay, rather than expecting individuals to put AI at the center.

© 2023 Fractal Analytics Inc. All rights reserved


Made with FlippingBook - PDF hosting