Algorithmic Diversity: Zero Exclusion, AI & Ethics (Women in AI)
Yonah Welker20:522021
A biased algorithm is never the culprit, the people behind it are
Only one in ten people with disabilities has access to inclusive technology, as the speaker cited, even as smart glasses, robots and emotion recognition tools multiply. Yonah Welker, a Women in AI board member and MIT fellow who evaluates European Commission projects, argues the fault lies less in code than in who builds it. So who should be held accountable when facial recognition fails women and people of color?
See where AI bias really starts and how representation and accountability can fix it
Yonah Welker argues that AI bias is a social problem rooted in who builds technology, not a flaw of algorithms alone. Drawing on work with the European Commission, MIT and Women in AI, Welker calls for representation of women, disabled and neurodiverse people, social science in tech teams, clear accountability, and ethics taught alongside coding.
“So, there's a phrase, nothing about us without us, so, how you plan to create technology for women without women in technology, and how you plan to create solutions for disabled people without disabled people.”
“So, you're not able to solve social problem with technical fixes.”
What you will learn
Explain why representation in data science and AI teams, not algorithms alone, drives biased technology.
Recognize why adding a single ethics hire to a team cannot replace an ethical ecosystem grounded in social science and human rights.
Distinguish who is accountable for AI bias, including responsibility for omissions as well as actions.
Identify gaps in ethical guidelines for children, disability and neurodiversity, and why ethics should be taught alongside coding.
Find ways to join inclusive technology efforts, such as remote hackathons open to participants worldwide.
Key takeaways from the speaker
ai
ethics
healthtech
edtech
neurodiversity
Details
Level
Beginner
Format
Talk
Technical depth
Introductory technical
For
All career stages
Length
20:52
Published
June 2021
Questions this talk answers
What is algorithmic diversity?
It covers both solutions to algorithmic bias and a diverse spectrum of technology solutions for accessibility, disability, gender and neurodiversity.
Why is AI technology still biased?
Welker argues the key problem is representation: only about 10% of data science and AI team members are women, as the speaker cited, and people with disabilities or neurodiversity are a tiny fraction.
Can hiring an ethics professional fix a team's ethical problems?
No. One person cannot build an ethical ecosystem; teams need shared grounding in bioethics, human rights, social science and an accessible moral vocabulary.
Who is accountable when AI systems are biased?
Not the algorithm or robot, but people, teams, corporations and companies, who are accountable both for their actions and for their omissions.
Why should ethics be taught alongside coding?
Most technology is created by startups, at hackathons and on university campuses, where ethics is not studied, so isolated ethics professionals cannot make up for it.
About this talk
I would love to participate in the next event in order to bring more attention to the problems of assistive technology, AI, ethics and human rights (including disability, women and children rights).
Ecosystems: MIT (Fellow), European Commission / ISDI (EIR), 500 Startups (EIR), Insead (EIR), Women in AI (EIR/board), All Tech Is Human (Fellow)
Recent appearances: AI Asia Conference, MozFest UK, Cognitive AI Switzerland, Spark Fest Australia, Horasis Summit AI US, AI Summit New York, CreAItivity, Digital AI Conclave,
Disability Tech UK, Way Davos etc (North America, Europe, APAC, Middle East, Australia)
Algorithmic diversity series (over 40 world destinations) - link
Recent projects: Women in AI US, Human-Centered AI & Ethics (EU Com), Human-Centered Factories (Eu Com), Future Of learning - ISDI, Decolonial AI, Algorithmic Diversity.
Upcoming papers:
-------------------
● AI Ethics & Algorithmic Diversity; (Germany)
● Ai Ethics, DEI & Postcolonial Technology (MIT, Global)
Description:
During the live session, we will dive into the meaning and application of algorithmic diversity in technology (including such cases as neurodiversity). Using the latest experiences, cases and research, we will analyze the current state and problems of inclusive innovation and technology, including the problems of representation and criteria, inclusive research and design-thinking, the building of inclusive products (AI-driven platforms, devices, apps, social and emotional robotics), ethical considerations and concerns (the "black-box" and "double-check" problems, transparency, explainability, fairness, surveillance), shortcomings of current technology ecosystems, policies and human rights frameworks.
Notions:
-------------------
● Neuro Representation in data science and AI teams
● Inclusive and neurodiverse solutions in hiring, learning, wellbeing and creative fields
● Collaborative AI, Data & NLP for open-source neuro solutions
● Ethics, policies and human rights
Overview
I've been working on the intersection of tech and society since 2005 year - when I've become a tech explorer and launched a hardware think tank. Over my journey, I founded and co-created tech startups and labs, helped to facilitate tech ecosystems through North America to APAC, MENA, Africa, Europe, screened over 2000 teams, contributed projects in ethics (data, AI, tech), deeptech, impact and sustainability (European Commission). I guided and screened the world youngest tech and AI inventors, Ashoka and United Nations fellows, talent from Google, Singularity University, 500 Startups, Techstars, Masschallenge, alumni of MIT, Harvard, Stanford, Oxford, womenpreneurs and leaders from Oprah Academy, WEDU, African Leadership Academy etc.
Links:
● yonah.ai - ai, technology, roboethics
● yonah.org - future of inclusive tech and research
You can't fix social problems with technical fixes (62 s)
Bias is a social and human rights problem (63 s)
I would love to welcome all the participants and there's a session called Algorithmic Diversity: Zero Exclusion, AI, and Ethics. My name is Yonah Welker and my work is focused on the future of ability, neurodiversity, accessibility, and gender. My work can be divided on several levels. On one hand, I try to connect the dots through technology companies, and I try to connect the dots through technology companies, and I try to find out the next level of these opportunities.
I serve as an evaluator and judge for projects funded by European Commission, but also collect portfolios of early-stage companies focused on the future of learning, well-being, accessibility, and gender. At the same time, I try to solve issues and challenges behind this technology, so I'm focused on guidelines, frameworks, ethics, and policies related to human rights, disability, women, and women.
At the same time, my goal is to democratize our movement and involve more people to this ecosystem, so I'm focused on hackathons driven by MIT, and this year I had the opportunity to collaborate with women in AI, and we launched and presented Zero Exclusion Hackathon, which involves people around the world, across the United States, Europe, Asia, and Africa.
So, as I mentioned, today's session called algorithmic diversity. So, this concept involves both solutions behind algorithmic bias, but also diverse spectrum of technology solutions related to accessibility, disability, gender, and neurodiversity. Unfortunately, unfortunately, until today, ecosystems were designed by one vision, one man, in most cases, one male, and afterward, we just tried to include others to this monopoly of design thinking, including women, neurodiverse people, disabled people, ethnic minorities.
And even though this system, and even though this system, these institutions, classroom, workplaces didn't fit them, we still tried to add them to this ecosystem, even though it was completely not for them in terms of a nature, logic, mechanism, and definitions. So, hopefully, hopefully in recent years, though, women are still underrepresented, women become a driver of innovation in accessibility movement, in disability movement, and many innovators help us to come up with new projects, new ventures.
And one of the reasons why I joined women in AI is to demonstrate is to how to create not only for women, but also by women. So, we try to demonstrate how girls and women could create sometimes much better than their fellow males, how they able to code, engineer, create robots, AI solutions, platforms, and apps. So, today's sessions, I would love to share how women are able to join this movement, how we're able to not only create technology and make our world more accessible for everyone, but also how we're able to make our technology less biased, more diverse, how we're able to come up with more solutions, solve issues related to representation, and the lack of connection between social science, gender studies, African Asian studies, and technology.
So, first of all, when we talk about accessibility, very often we can share some good news, because Microsoft and Amazon, they introduced their accessibility programs and programs related to disability. So, Google and Google Glass became a platform and technology solution to solve issues in emotion recognition for artisan-focused startups. So, we have tech companies try neurodiverse hiring platform, which help to hire more people with a learning disability, dyslexia, or autism into technology companies.
At the same time, schools experiment with dyslexia, ADHD, and other training solutions, and recently, I had the opportunity to work with a search company from Denmark, and we tried to incorporate the solution across not only Nordic countries, but beyond. But, unfortunately, the statistics related to representation, related to employment and unemployment are still pretty sad. 90% of people in the United States and UK with autism are not employed full-time, and only 1 of 10 people with disabilities actually have access to with inclusive technology.
So, although we have so many wonderful robots, smart glasses, emotional recognition, computer vision, and technology, which can be used by nurses, doctors, educators, it's still not so accessible, it's still not so penetrated actual market. But, unfortunately, we have another problem, ethical, and problems of framework behind this technology, and it includes algorithmic bias, lack of transparency, lack of accountability, lack of double-check principle, lack of social science behind technology, so it leads to technical fixes, lack of privacy, and, finally, we still have policy behind autonomous agent.
So, how we are able to solve it, I would love to share our recent lessons with European Commission, with the MIT, with the women in AI, and our both technology and policy-making efforts. So, first of all, I believe the key challenge we have today, and we're not able to make technology less biased, is a representation, because very often we try to say, like, it's a problem of algorithms, it's a problem of technology, but the actual problem, it's a representation.
We have only 10% of women in data science and AI teams, and we have just some 0.5, 0.4, just a tiny percent of people with disabilities, neurodiversity, we have a stigma towards mental health disorders in so many conditions. So, there's a phrase, nothing about us without us, so, how you plan to create technology for women without women in technology, and how you plan to create solutions for disabled people without disabled people.
Honestly, it was said on the World Economic Forum, AI can be both a solution for any problem in the world, but at the same time, the source of the apocalipses, depending who and how ask these questions. So, until we have a people, all of the spectrum, all of the ethnicity, all of the race and gender, we're not able to ask correct questions to build actually diverse micro teams in research and other departments.
So, in the end, we just fail, we fail to ask questions, to build the MVP of products, and for sure, all of our efforts are just useless. Another problem, and I know, specifically, in my personal journey, my first education related to liberal arts and history, and also, I was always passionate about social science, though I spent most of my life in technology world, is it just dramatic disconnection between technology and social science?
Because, because in most cases, because in most cases, when we talk about technology team, we just think about designers, engineers, but guys, how you plan to solve the problems of gender studies, African studies, Asian studies, how you plan to come up with the criteria and definition, race, ethnicity, neurodiversity, if you have no professionals, focus on these studies.
So, in most cases, so in most cases, we created not only gender and ethnicity vacuums and filter bubbles in our teams, but also filter bubbles in terms of studies. So, in most cases, I mean, we have no social science professionals in technology teams, and someone who would help to come up with the correct criteria, for sure, we came up with some innovation in this field as an ethics professional, so as a kind of advocates of more ethical technology in residence.
So, in most cases, but in most cases, first of all, startups have no money to hire such people, and second, it doesn't build any kind of actual ecosystem, so just one person who actually tried to push other people, but it doesn't help. So, in most cases, actual solution led in much bigger spectrum, and it's building ethical ecosystems driven by bioethics, human rights, where you have an accessible moral vocabulary, understanding of gender studies, African studies, ethnicity studies, and so on, where anyone has awareness about this problem.
So, in most cases, you're just not able just to incorporate some ethical, ethics person into not ethical team, you're not able to solve the Facebook problems, just integrating some ethics professional, and you know what happened in Google, they just fire these people, because we have no, any kind of a moral ethics just in the core of a company.
So, you're not able to solve social problem with technical fixes. It can be solved through complete reshaping of the knowledge, the background, and representation of these people in teams. Another problem, which I would love to mention, very often when we talk about technology, about bias specifically, in such fields like nursing, medicine, healthcare, wellbeing, education, schools, classes, schools, classrooms, workplaces, which we typically try to consider AI algorithms or technology as a subject of the law.
So, it's a technology bias, it's a robot is biased, but the problem is that it is technically impossible, biased person, biased people, biased teams behind it. And actual people, actual subjects of the law is a citizens, as individuals, as a team, as a corporation, as a company. So, though such movies like coded bias, they made amazing work in terms of awareness, we should go forward in terms of accountability.
So, when we talk about accountability, who actually is accountable for facial recognition bias, when we have a 30% error for women or people of color, who is actually responsible for that, and who is accountable. And that's why we are not only building ethical ecosystems and agents of ethics and accessible moral vocabulary, but we also try to understand who is accountable both for actions and for omissions.
Omissions is not actions, so it's a condition when you need to act or do something, but you do absolutely nothing. So, we define all of these agents of responsibility for particular levels of technology, research, implementation, and both evaluation, both negative and positive impact. And until we are able not only to define understanding of accountability, but also putting it into the policies, because just recently, European Commission and European Union came up with an updated vision of policies related to computer vision, facial recognition, security cameras, and other stuff.
But it's still like a yes or no, like we are banned or not, but it's not about deeper understanding, why it fails, why it doesn't work properly, why these companies develop it in this way, who's responsible? It's more about CEO, it's more about designers, engineers, researchers, or it's about the whole complete biased ecosystem. So, this problem is much bigger.
And finally, and it's specifically related to my work, because recently, I came up with an educational program focused on human-centered AI and ethics. It was founded by European Commission and created in collaboration with ISDI Business School in Spain and published on Shop4CF platform. It's a lack of access to ethical guidelines. So, in most cases, we have many policymakers, experts, practitioners, and other people who are developed with frameworks related to ethical development, robot ethics, and many, many stuff like Alan Turing Institute, Montreal AI ethics Institute.
But unfortunately, most of the technology are created by startups, by founders. They created during hackathons or in university campuses. And unfortunately, in today's world, ethics is not studied in schools or university campuses. So, if we organize hackathons or if we teach engineering or coding in schools, we should teach ethics at the same time. And any type of, let's say, implementing ethics through some isolated professionals, not through democratizing, it just doesn't work.
And another problem is the lack of the niche ethical guidelines. My key focus is accessibility, neurodiversity, disability, and in most cases, the people who are the most vulnerable ones, who are mostly affected both by society, but also negative consequences of technology use, we use our women, children, young people, elder ones, and so on. And unfortunately, until today, we had no frameworks and guidelines related to this field.
So, for instance, only in 2020, UNICEF came up with a framework called AI and children. So, it was August 2020. And we had no frameworks for this field before it. So, we built smart toys. Teenagers and young people were the main consumers of the content on the platform like YouTube, Facebook, and many, many, many types of solutions.
But we had no any kind of frameworks and guidelines which actually regulate how we use technology in this field. And something similar we have in general for disability. And that's why artistic people become victims of cyber bullying, some negative ways to use technology, and so on. So, that's why, for instance, when I work on the human-centered AI and ethics, in the end, I put the beginning of a huge discussion related to human rights.
Because, in most cases, when we talk about bias, it's not a technology problem. It's a social problem. It's a huge social problem because we just stop to actually take into account human rights. We only take into account corporate interest, profits, and how some CEOs make companies unicorns. So, all of the news, all of the magazines just filled by founders, entrepreneurs, successful people.
But we don't talk about how many people actually become victims of this technology. How many women become victims of sexual harassment on workplace. Or they use harassed by technology or through technology. So, actually, until we have all of this discussion, nothing actually will change. And I really spent many efforts to change it for good because I'm a disabled person.
I'm a non-binary person. And in recent years, women were my key allies in this movement. Trying to fix representation, to reshape and dismantle colonial technology. A colonial vertical of power driven by, in most cases, white males who use it for profits, for their own greed, but not for society. So, that's why, until today, all of the ecosystems, as I mentioned before, were designed by one vision, one man, one type of design thinking, without any kind of a competition, without any kind of discussion.
Maybe the whole road is wrong. Maybe we need different type of agile, lean approach, or design thinking, or approach to research, or criteria behind technology. But hopefully, now we have more opportunity to create dialogue, collaboration. And that's why I use almost any opportunity during my talks to say, Hey, guys, you're able, even though we have a digital divide, I mean, currently we have a unique opportunity to participate.
I mean, I organize hackathons across Africa, Asia, Middle East. And in most cases, the most talented people come as immigrants, come from emerging countries, from India, from China, from Saudi Arabia, in many other countries. So, they participate in our hackathons completely remotely. And very often, they become winners. Just a recent example is our startup AI4NI focused on autoimmune disorders and Crohn disease, became a winner of our hackathon in the United States.
And before, my mentors became winners of an open AI hackathon in MIT hackathon. So, though digital divide exists in this objective thing, we have more opportunities. So, I welcome all of you to participate in this accessibility movement. Now we're able to make a difference. We're able to build AI robots, apps, platforms. We're able to leverage all of our voices of women, LGBT, people of color to this discussion to make our voices heard.
So, please, let's do it. And you are welcome to this movement. Thank you so much for your time. It was Yona Welker. I'm always happy to connect on social media. You can find me on yona.ai or yona.org. It's a foundation focused on the future of disability and neurodiversity movement and technology. So, thank you so much and have a wonderful day.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
Why technical fixes cannot solve AI bias
Yonah Welker argues that biased technology starts with who sits on the team, and that accountability must land on people, not machines.
In short
AI bias is a social problem before it is a technical one, according to Yonah Welker. Welker, who works with Women in AI, MIT and the European Commission, argues that biased technology comes from teams that lack women, disabled and neurodiverse people and social scientists. Fixing it means changing who builds technology, making people accountable for what they do and fail to do, and teaching ethics alongside coding.
Key takeaways
Explain why representation in data science and AI teams, not algorithms alone, drives biased technology.
Recognize why adding a single ethics hire to a team cannot replace an ethical ecosystem grounded in social science and human rights.
Distinguish who is accountable for AI bias, including responsibility for omissions as well as actions.
Identify gaps in ethical guidelines for children, disability and neurodiversity, and why ethics should be taught alongside coding.
Find ways to join inclusive technology efforts, such as remote hackathons open to participants worldwide.
What algorithmic diversity means
Welker opens by describing work that spans evaluating European Commission projects, collecting portfolios of early-stage companies in learning, wellbeing, accessibility and gender, and running hackathons, including the Zero Exclusion Hackathon launched with Women in AI. The idea at the center of the talk, algorithmic diversity, combines two things: solving algorithmic bias and building a diverse range of technology for accessibility, disability, gender and neurodiversity.
The starting point is a critique of how technology ecosystems were built: "Unfortunately, unfortunately, until today, ecosystems were designed by one vision, one man, in most cases, one male, and afterward, we just tried to include others to this monopoly of design thinking, including women, neurodiverse people, disabled people, ethnic minorities." (1:30)
There is progress. Welker points to accessibility and disability programs at Microsoft and Amazon, Google Glass being used for emotion recognition by startups focused on autism, neurodiverse hiring platforms, and schools experimenting with dyslexia and ADHD tools. But the figures Welker cites remain sobering: 90% of people with autism in the United States and UK are not employed full-time, and only 1 in 10 people with disabilities has access to inclusive technology.
Representation is the real problem
Welker's central claim is that bias gets blamed on algorithms when the deeper cause is who is in the room. As the speaker cited, only about 10% of data science and AI team members are women, and people with disabilities or neurodiversity make up a tiny fraction, with stigma around mental health conditions on top of that.
"So, there's a phrase, nothing about us without us, so, how you plan to create technology for women without women in technology, and how you plan to create solutions for disabled people without disabled people." (6:39)
Without people from across the spectrum of ethnicity, race and gender, Welker argues, teams cannot even ask the right questions, so products and their first versions are built on the wrong foundations.
Social science belongs in technology teams
Welker, whose first education was in liberal arts and history, describes a "dramatic disconnection" between technology and social science. Teams are imagined as designers and engineers, yet they are expected to define criteria around race, ethnicity, gender and neurodiversity without anyone trained in gender studies, African studies or Asian studies. The result is filter bubbles not only of gender and ethnicity but of disciplines.
Hiring a single ethicist does not solve this. Startups often cannot afford such a role, and one person pushing others does not create an ecosystem. Welker instead calls for ethical ecosystems driven by bioethics and human rights, with an accessible moral vocabulary that everyone on the team shares.
"So, you're not able to solve social problem with technical fixes." (10:27)
Accountability belongs to people
Welker challenges the habit of treating an algorithm or a robot as the biased party. The real subjects of the law, Welker says, are citizens, teams, corporations and companies. Awareness efforts such as the film Coded Bias are valuable, but the next step is accountability, for example for the 30% facial recognition error rate for women and people of color that the speaker cited.
That accountability covers what people fail to do as well as what they do: "Omissions is not actions, so it's a condition when you need to act or do something, but you do absolutely nothing." (12:13)
Welker notes that recent European policy on facial recognition and surveillance cameras still tends toward a yes or no ban, rather than asking why systems fail and who, from CEOs to engineers to the wider ecosystem, is responsible.
Ethics must be taught, especially for the most vulnerable
Drawing on a human-centered AI and ethics program funded by the European Commission and created with ISDI Business School, Welker highlights a gap in access to ethical guidelines. Frameworks exist from bodies such as the Alan Turing Institute and the Montreal AI Ethics Institute, but most technology is built by founders at hackathons and on campuses where ethics is not taught.
"So, if we organize hackathons or if we teach engineering or coding in schools, we should teach ethics at the same time." (14:22)
Niche guidelines are also scarce. Welker points out that UNICEF only published its AI and children framework in August 2020, long after smart toys and youth-focused platforms became common, and that disability still lacks similar guidance. For Welker, who identifies as disabled and non-binary, this ultimately becomes a human rights discussion: "Because, in most cases, when we talk about bias, it's not a technology problem. It's a social problem." (16:50)
An open invitation to build
Welker closes on opportunity. Hackathons organized across Africa, Asia and the Middle East draw talented participants, often immigrants or people from emerging countries, who join remotely and frequently win, such as the startup AI4NI, focused on autoimmune disorders and Crohn's disease. Despite the digital divide, Welker invites everyone into the accessibility movement: "We're able to leverage all of our voices of women, LGBT, people of color to this discussion to make our voices heard." (19:44)
Watch the talk
"Algorithmic Diversity: Zero Exclusion, AI & Ethics (Women in AI)" by Yonah Welker, 21 minutes: https://www.womentech.net/video/algorithmic-diversity-zero-exclusion-ai-ethics-women-in-ai
About the speaker
Yonah Welker is an Explorer and Board Member with Women In AI, MIT and the European Commission, and has worked at the intersection of technology and society since 2005, when they launched a hardware think tank. Welker has founded and co-created tech startups and labs, helped facilitate tech ecosystems from North America to APAC, MENA, Africa and Europe, and screened over 2,000 teams.
Questions this talk answers
What is algorithmic diversity?
Why is AI technology still biased?
Can hiring an ethics professional fix a team's ethical problems?
Who is accountable when AI systems are biased?
Why should ethics be taught alongside coding?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Nothing about us without us
Double-check principle
Design thinking
UNICEF AI and children framework
Agile
Lean
Resources mentioned
Coded Bias (film)
UNICEF AI and children framework
Human-centered AI and ethics program (European Commission, ISDI Business School, Shop4CF)
Alan Turing Institute ethics frameworks
Montreal AI Ethics Institute
yonah.ai
yonah.org
Cite this talk
Yonah Welker. "Algorithmic Diversity: Zero Exclusion, AI & Ethics (Women in AI)." WomenTech Network, 2021. https://www.womentech.net/video/algorithmic-diversity-zero-exclusion-ai-ethics-women-in-ai