Establish Responsible AI Values

March, 2023 • Alex Serban, Koen van der Blom, Joost Visser

39 / 46 • Governance •

This practice was ranked as advanced.
Click to read more. •

This practice addresses requirements
from the EU guidelines for trustworthy ML.
Click to read more.

Intent

Explicitly align all stakeholders on the ethical values and constraints of your machine learning application

Motivation

Machine learning applications can severely impact human lives. Avoiding negative impacts, even without malicious intent, requires all stakeholders to operate according to the same ethical values.

Applicability

Values for Responsible AI should be established for any ML application.

Description

A good starting point for sharing ethical values across organisations is to subscribe to a code of conduct. You can create one specific for your situation, or you can refer to a general governance framework.

General governance frameworks that you may want to refer to include:

These frameworks set high-level objectives for concerns such as security, privacy, and fairness. They can be complemented or refined for your specific situation.

Defining or subscribing to a code of conduct helps to build trust with users and enhances the auditability of your development process and your applications.

The values set by a code of conduct can be refined into a set of concrete governance objectives. The governance objectives appropriate for a given machine learning application depend on factors such as:

the type of data being processed: for example, when personal information is processed, privacy objectives are relevant
the usage context: for example, when used for autonomous driving, safety objectives are relevant
the organisational context: for example, a government agency may require usage of open data and non-proprietary machine learning algorithms

Adoption