Thoughts About the Need for More Diversity and Inclusion in AI

AI has a diversity and inclusion problem, but the good news is that AI itself can give us a hand.

I’m black, and I am a foreigner living in Poland, a country that still has a long road to cover regarding diversity and inclusion of minorities and I’ve been very active in volunteering on a local ONG, designing and delivering workshops that help hundreds of teenagers better understand the value of tolerance and inclusion in our society.

I am also a father of three young girls. I’ve been very committed to encouraging girls and women in computing because I’m sure that diversity — not only of gender but racial and social — is fundamental to the area in which I work: Artificial Intelligence.

Because I believe that without the right amount of diversity, AI will not be able to reasonably contribute to solving the most relevant problems and deepen our world’s inequality, maybe you are asking yourself why it matters…

Well… let me tell you simply that AI is already impacting your life, every day, and everything you do… so probably you would prefer that these ubiquitous and almighty algorithms should be fair, ethical, and inclusive; just life should be, right? If you are still not convinced, I will try to give you something to think about…

Why does the lack of diversity in Artificial Intelligence can be an issue?

We can count several cases of prejudice in AI systems, including improper classification of minorities, chatbots who have learned to disseminate hate speech (Microsoft), black people being classified as gorillas in search engines (Google), etc.

An interesting report called Discriminating Systems — Gender, Race, and Power in AI by the University of New York draws attention to facts that deserve discussion: many “flaws” related to cases of discrimination widespread in Artificial Intelligence (AI) systems would be associated with the lack of diversity in the teams that work these technologies.

Unfortunately, this lack of diversity in the field of AI negatively affects systems, as many works with the issue of automatic data learning, and there is a bias (ideological, gender, race, etc.).

To get an idea of ​​this lack of diversity in AI teams, here are some data from the report:

• More than 80% of AI teachers are men;
• Only 15% of AI researchers on Facebook and 10% of AI researchers at Google are women;
• Men currently represent around 71% of the group of candidates for IA US jobs, as shown in the 2018 AI Index report.

As Rob Doyle wrote in an interesting article about Sexism in Tech: An Inconvenient Truth:

To tackle sexism in the tech industry, first, the status quo must change. Then, it must be consistently monitored at all levels. This not only involves solving problems in the current workplace but also getting things right at an educational level. More females need to be encouraged to learn STEM subjects to create an equilibrium in the tech industry — Rob Doyle

This lack of diversity can be considered as a reflection of the field of Computer Science itself, considering that it is estimated that only 24% of computer professionals are women. I wrote an article about it:

It’s clear that one of the biggest problems here is not the data or the algorithms: it is the blind spots created by the lack of diversity — of experience, education, and thinking — in the teams that develop AI, which makes it challenging to anticipate biases and their potential impact.

Machine learning plays a crucial role in most of the artificial intelligence solutions used today. It is the process by which large amounts of data are used to train AI systems to, for example, extract meaning from text or audio files to answer questions or make recommendations.

Flawed data and biased models can easily lead AI to erroneous conclusions that affect people’s credit scores, employment options, school admissions, and their level of risk in criminal court cases.

Artificial intelligence & diversity

Despite being a recent technology, Artificial Intelligence has already shown its value in the field of health, education, and urban mobility, among others.

Considering the pace at AI, Machine Learning, and Deep Learning has been developing in recent years, it is expected that Artificial Intelligence (AI) will profoundly modify the way we live and work.

One of the areas in which this paradigm may stand out in the future is related to Diversity & Inclusion within companies.

AI can detect potential bias and prejudice in decision-making by simulating intelligent behavior, potentially reducing trends and prejudices that could hinder organizations’ ability to recruit diversely and inclusively.

Machines do not have an intrinsic propensity to let their personal experiences, opinions, and beliefs influence their decisions. Still, we need to consider that computers are based on data and algorithms created by the people who developed them.

Consequently, if designed and applied ethically, through multidisciplinary and diverse teams, AI can detect situations of potential bias and prejudice in decision making, particularly the one that, being unintentional, becomes more difficult to see.

Can AI itself help us to solve the lack of diversity in AI?

Artificial Intelligence has several ways to impact Diversity & Inclusion within organizations, including promoting equal access to work opportunities and supporting HR departments in making more ethical decisions.

AI can support the development of more diverse and inclusive workplaces allowing access to employment opportunities, supporting the design of more inclusive job ads, managing an ethical recruitment and selection process, and of course, minimizing prejudice during the worker’s life cycle.

Despite the AI hype, we must not forget that this technology depends on humans’ data collected and selected.

Given that AI solution development depends on human decisions and by design, human beings have several unconscious biases; data and machine learning models must be tested and monitored continuously to ensure that they fit their objective in a very ethical way.

By the way, if you want to find out which are your biases, you can take the Implicit Association Test (IAT), carried out by Harvard University.

Conclusion

As important as the improvements and optimizations that AI can bring to society, it must also be put in the condition to be ethical, diverse, and inclusive.

As we saw, the potential benefits of inclusion and diversity are irrefutable; therefore, if automation is a welcome move for companies that invest in technology, the diversity of teams is essential for this to be a path of no return.

Further content

If you want to read more about AI ethics, Diversity and Inclusion, and also about volunteering, here are some other articles that I’ve written about it:

References:

1. Annual IBM List Celebrates Global Women Leaders Shaping …. https://newsroom.ibm.com/2020-05-06-Annual-IBM-List-Celebrates-Global-Women-Leaders-Shaping-the-Future-of-Artificial-Intelligence
2. Como a falta de diversidade nas equipes de Inteligência Artificial (IA) tem afetado as tecnologias — Patrick Pedreira — https://www.linkedin.com/pulse/como-falta-de-diversidade-nas-equipes-intelig%C3%AAncia-ia-pedreira
3. ‘Disastrous’ lack of diversity in AI industry perpetuates …. https://www.theguardian.com/technology/2019/apr/16/artificial-intelligence-lack-diversity-new-york-university-study
4. Corp! salutes diversity award winners — Corp! Magazine. https://www.corpmagazine.com/features/cover-stories/corp-salutes-diversity-winners/
5. Paving the Way for Diversity in the Decade of Ubiquitous AI — https://www.ibm.com/blogs/think/2020/05/paving-the-way-for-diversity-in-the-decade-of-ubiquitous-ai/
6. Racial discrimination persists at Facebook and Google …. https://www.usatoday.com/story/tech/2020/02/10/racial-discrimination-persists-facebook-google-employees-say/4307591002/
7. Charter School Governance: An Exploration of Autonomy and …. https://digitalcommons.georgiasouthern.edu/cgi/viewcontent.cgi?article=2604&context=etd
8. IBM BrandVoice: For AI That Works For Everyone, We Need …. https://www.forbes.com/sites/ibm/2019/12/09/for-ai-that-works-for-everyone-we-need-everyone-to-help-design-it/
9. IIDA’s Leadership Discusses the Importance of Diversity in …. https://www.interiorsandsources.com/article-details/articleid/22798/title/iida-diversity-design-industry
10. Discriminating Systems — Gender, Race, and Power in AI — University of New York — https://ainowinstitute.org/discriminatingsystems.pdf
11. Eye-Opening Statistics on Diversity Every Recruiter Needs …. https://www.censia.com/blog/diversity-statistics/