Machine Learning, Dynamical Systems and Control

Our Approach to Ethics in AI

Recognizing the significant impact of artificial intelligence (AI) on society, we are committed to upholding high standards in our efforts. As technology and its regulation advance, we anticipate making updates to this approach:

 

1. Ethical Awareness and Education
  • We provide free, accessible education on the ethics of artificial intelligence. Our annual series covers strategies for navigating ethical challenges. By equipping individuals with knowledge to handle these, we help cultivate a community that is informed and mindful.
  • Our workshops include discussions of ethics in AI, e.g., “How do we ask the right questions of a data set to not influence the outcome that your methods work the best? How closely do you track the data provenance pipeline?” These tough questions lead to productive conversations and foster a community approach to ethical decision-making in AI.
  • Our Institute’s Code of Conduct policy encourages and upholds ethical behavior.
  • Our internal resources point to an ethical AI checklist for researchers. (Available in Slack.)
2. Interdisciplinary Collaboration
  • Through cross-institutional partnerships, we include diverse perspectives to ensure ethical AI practices align with varied values and priorities.
  • We strive to offer widely adopted, industry-leading open-source code, ensuring reproducible and transparent research to further democratize AI.
3. Ethical Implications of Our Tools
  • The Sensing & Optimization Thrust emphasizes active learning to achieve modeling objectives, enhance model trustworthiness, and reduce biased sampling in training. It also emphasizes sensor redundancy and the study of failure events (in sensors, models, and control) and ensures AI models do not introduce bias.
  • Both the Modeling and Control Thrusts incorporate physics, explainable AI, and uncertainty quantification to enhance trust and streamline licensing and deployment in critical applications.
  • Our Common Task Framework (CTF) underpins the development of interpretable, ethical, and inclusive tools in the following ways:
    • Supports sustainable, open-source challenge datasets.
    • The CTF’s platform, Synapse, is hosted by Sage Bionetworks, a non-profit organization known for collaborating with notable organizations on various challenges.
    • Their platform is equipped with strict data privacy governance and controls. With environmental impact and equitable resources in mind, we are designing the CTF as less resource-intensive. This approach promotes inclusivity, allowing users with limited resources to participate. Our goal is not just technological advancement, but also sustainable practices for the future.