WHAT IS INCLUSIVE AI? A PERSPECTIVE FROM THE WOMEN’S FORUM

Right now, discussion about the need to reduce bias in technology is everywhere. The question now should not be only about how we can reduce bias, but how we can actively make AI inclusive. 

We know that carrying on with business-as-usual would mean continuing to hardwire bias and discrimination into society. However, currently, there is not enough diversity in the rooms (and Zooms) where technology is designed to change the direction in which things are going. Only 15% of data scientists worldwide are women, there is not enough diversity in the datasets on which AI is based. We see that data can either be unrepresentative and skewed which can cause biases or be based on a biased society and hence further engraining these biases. It’s not just the data, but the way that algorithms are designed, for example, software which filters job applicants and removes those who don’t look like previously successful candidates. 

 Too often, organizations attempt to adopt policies which address these issues, and they do this in good faith, but in the end, they implement these policies ineffectively. 

The problems are clear. But what would genuinely inclusive AI look like?  

Defining ‘Inclusive AI’

As there is currently no commonly recognised definition of Inclusive AI, we want to propose a clear definition and position it amongst the other terms in the ecosystem. We think this is crucial to enable organisations to be able to truly implement Inclusive AI. This definition must go beyond bias and representation.  

In this post, we are sharing proposed definitions to make our mark on the AI ecosystem and to receive input and reflections to refine our definitions further. We define ‘inclusive AI,’ along with ‘responsible AI,’ as subfields of ‘ethical AI’. 

·      Ethical AI: we define this as a subfield of applied ethics which aims to elicit ethical issues and the standards of right and wrong required for a moral conduct in the development and use of AI technologies. 

·      Responsible AI: a subfield of ethical AI, this is AI that seeks to translate the vision of ethical AI into actions which guide the development and use of AI, in terms of embedding the right values within the AI lifecycle.

·      Inclusive AI: another subfield of ethical AI, this is AI which accounts for different needs, and therefore benefits all of society including minority, marginalised, and underrepresented groups. It does this through reducing bias and discrimination within and resulting from AI systems, as well as trying to reduce the inequality of access to these systems and the digital literacy required to use them. Inclusive AI must be non-discriminatory in its production, unbiased in its consequences, and accessible to all. That means that no matter how virtuous an AI project is, if it was conceived, designed, and built by teams which are unrepresentative of society, it isn’t ‘ethical’ or ‘inclusive’ after all. 

From promise to practice

Because we, at the Women’s Forum want to continuously engage with our community and share with you best practices, I am very happy to announce the launch of our Women4AI Daring Circle blog. The blog will feature a series of articles on LinkedIn written by a cross-disciplinary set of experts working on AI around the world. It will offer perspectives from people at the chalkface – those who are working to draw up effective strategies for inclusive AI. It will be a space to share insights, learn from each other’s experiences, and celebrate successes. Expect to come across stories about how badly designed algorithms are being challenged in the courts, how companies are encouraging the next generation of women to work in AI, and plenty more.

Building Inclusive AI

So how, in practice, should organisations move towards inclusive AI? How should progress towards it be measured? What about inclusive AI in procurement? 

Together with our committed partners and with the renowned experts working at our side, within our Women4AI Daring Circle, we recognise that there is a gap between theoretical commitment to Inclusive AI principles and the implementation of those commitments, so we have decided to work on an Inclusive AI toolkit. We have launched our toolkit during our Global Meeting. Our toolkit will be a unique instrument which will to support organisations who are at different stages of their AI journey by providing practical tools for different roles within organisations and showcasing case-studies of Inclusive AI. 

There’s never been a better time to spark a generational change.

 Through this work, we encourage business leaders to embrace AI’s transformative potential and to all of us to take a responsible approach to it. In the end, it’s how we use the AI, not the technology itself, that ultimately enables us to not only solve problem but to build a better (working) world for all.