AI for Society – Inclusion in AI Development and Deployment


Draft – For discussion purposes only; does not represent the views of the G7 or its members. This discussion paper was drafted to guide the discussion during the breakout sessions at the December 6, 2018, G7 Multistakeholder Conference on Artificial Intelligence in Montreal, Canada.

Executive Summary

Artificial intelligence (AI) opens up wonderful opportunities, but democratic societies will need to address major challenges to ensure that the positive effects of algorithm development and use are shared equitably. In particular, we must ensure that AI growth does not amplify existing prejudice, or lead to the exclusion of vulnerable populations.

This short paper emphasizes the benefits of inclusion and diversity. It is also key to try to raise the numeracy and digital literacy levels of all citizens, and to improve access to key resources (including venture capital, infrastructure, databases, training).

There may be disparities in how AI-generated productivity gains are distributed (e.g. only rich countries or hospitals may have access to the latest tools for early detection of disease). As such, initiatives will need to be undertaken to promote the delivery of applications that benefit excluded groups (e.g. by promoting broad strategic engagement across government, developing technical solutions to social and ethical problems).

Even well-designed AI tools might have negative effects on the long-term economic and social inclusion of the world's population, e.g. automation-led job displacement. Countries will therefore be led to invest in the development of new industries, devise reskilling schemes, consider a broader social safety net, and modernize national labour policies.

There are also risks that the development of AI technologies may inadvertently lead to the exclusion of some groups. Researchers and private sector players have started to develop tools to mitigate such risks, but more will be needed to safeguard populations (e.g. closer involvement of end users during the design and implementation phases, ensuring adequate governance mechanisms). Finally, it is important to bear in mind that AI can and indeed should be part of the solution by revealing and disarming biases in existing systems and processes.

To conclude, countries should explore the following questions to ensure the safe and responsible development of AI technologies:

  • How do we make sure gender, social and cultural diversity are incorporated into AI design and development?
  • How can we together learn and share good practice for inclusion and diversity?
  • What can government, academia, and industry do to promote an inclusive use of AI technology that will benefit a diverse society?

Discussion Paper: AI for Society (PDF, 310 KB)