Governance can be a vague and squishy term that is thrown around often but not very well defined or understood. It helps to start at the beginning with the basics of governance. This definition from the Corporate Finance Institute is a good starting point:

  • Corporate governance includes principles of transparency, accountability, and security.
  • Poor corporate governance, at best, leads to a company failing to achieve its stated goals, and, at worst, can lead to the collapse of the company and significant financial losses for shareholders.

Because AI is impacting every industry, there are justifiable concerns emerging about how these automated decision-making systems will impact people, the economy, and the environment. Because of this, multiple organizations have taken on
the task of defining “AI Governance”  as a subset of governance. With that in mind, here are a couple of definitions of AI Governance:

“Accelerating the societal benefits of artificial intelligence and machine learning while ensuring equity, privacy, transparency, accountability, and social impact.”
World Economic Forum

“AI Governance is a framework and process that guides the design, development, and deployment of AI in a way that is explainable, transparent, and ethical. It comprises guidelines for actions, as well as systems and processes.”
Basis AI

Some examples of ways in which AI and ML systems could ensure good governance are listed here, borrowed from an excellent article by Scott Zoldi, Chief Analytics Officer at FICO titled “
Establish AI Governance, Not Best Intentions, to Keep Companies Honest

  • Accountability:
    • ensuring that everything is tracked and immutable so it can be audited;
    • Having a consistent approach to designing, developing, training, and evaluating models;
  • Fairness:
    • Looking for, identifying, and mitigating bias that is built into the data;
  • Transparency:
    • Having a way to understand how models were built, what features were used, and being able to assess whether they are still relevant as the situation and environment (and laws and regulations and social norms) shift;
  • Responsibility:
    • Assigning responsibility to humans who ultimately must be held accountable for the systems that they build.

Going back to the beginning, if we assume that AI will impact every part of our lives and we want to gain all the benefits while mitigating the risks, and building systems that are transparent, accountable, secure, fair, and good for society, we need to make governance a key priority and demand it from the vendors. Our organizations and our relationships with our stakeholders depend on it.

SAS Visual Data Mining and Machine Learning
Solve the most complex analytical problems with a single, integrated, collaborative solution
Take control of your R code
The enterprise feature store
Weights & Biases
With a few lines of code, save everything you need to debug, compare and reproduce your models
AI and machine learning model management and operations for enterprise data science teams
Power your machine learning lifecycle
Oracle – Data Science Platform
Data science platform
Creating data science
Fiddler Labs
Know the why and the how behind your AI solutions
Make your data science, engineering, devops, and governance teams work more efficiently