Your Personal Female STEM AI Coach

According to the « Why So Few? Women in Science, Technology, Engineering & Mathematics » study conducted by the American Association of University Women, one of the factors explaining why women are less likely to pursue STEM careers is the lack of female role models in the field. The lack of representation and inspirational figures that women can recognize themselves in makes it hard for women to imagine themselves as successful in those fields. Which in turn reinforces gender bias against women embracing STEM early on in their education and the cultural beliefs about women’s capacity to embrace STEM careers. And so and so forth. How do we break free from this vicious cycle that keeps on perpetuating the gender gap in STEM ?

 Fortunately, exposing girls to successful female role models can help counter negative stereotypes because girls see that people like them can be successful and stereotype threat can be managed and overcome. (McIntyre, R. B., Lord, C. G., Gresky, D. M., Ten Eyck, L. L., Frye, G. D. J., & Bond, C. F., Jr.  (2005). A social impact trend in the effects of role models on alleviating women’s  mathematics stereotype threat. Current Research in Social Psychology, 10(9), 116–36.)

Research has shown that exposing students to role models who can help students see their struggles as a normal part of the learning process rather than as a signal of low ability, and in turn can boost the test scores of girls.

However, having girls read biographies of a few successful female and diverse STEM leaders before standardized tests fails to provide personalized and personable mentorship that caters to the specific needs of the female students involved. Between Marie Curie’s 2 Nobel Prizes and a female middle schooler’s level of confidence when about to take a STEM test – there is a significant gap that might not be closed by imagination alone. How can we make female role models more accessible, personable and relatable to the girls we wish to inspire at scale? 

We are building a diverse crew of AI female STEM coaches that provide personalized on-demand mentorship to K-12 girls and are modeled after real-world inspirational female STEM figures. Imagine being coached by the most trail-blazing female figures in STEM on the eve of your STEM standardized test. They are no longer just Nobel Prize winners, or black and white profile pictures on Wikipedia articles. Rosalind Franklin, Dr. Mae Jemison and Marie Curie are available for guidance at your fingertips, and most importantly, they are all rooting in for you. 

The user experience is simple. This is a web app you can access on any device. The journey starts with a short quiz to get to know more about you, your interests, your specific challenges, so we can match you with an AI coach that best suits you based on their life story and field of expertise.  Once you are matched with an AI coach, the interaction can begin. Ask any question you wish, share your struggles and your wins, ask for guidance.   

At this point, whomever you end up being matched with, these extraordinary women are not only Nobel Prize winners, they are no longer intimidating figures that seem unattainable: they are easy to talk to, to confide with, to relate to. Just like you would chat with a real person, a mentor and even a friend. You don’t have to go to the library to pick up their biographies, or scroll through lengthy Wikipedia articles to get to know them: in other words, it never feels like homework. Just take your smartphone, open up the app and meet your newest supporters in a few clicks. And they are not here to talk about their successes, they are here to listen to you, to what you are going through and how you feel. They can share their own struggles and how they overcame them and how it applies to your specific situation. Winning against all (gender) odds has never felt so close. 

What’s even better is that you can customize your AI coach’s age so that it looks like you, and over time, your AI friend ‘grows up’ with you. In so doing, we aim to strike the optimal balance between inspiration and identification. You are inspired by the female STEM leader, while you can still see yourself in her.

Alongside Miami-Dade County Public Schools and the French National Department of Labor, we are working with girls and young female adults between the ages of 12 and 25 years old both in the United States and France, to enable them to visualize themselves as successful professionals in counter-stereotypical career pathways, notably STEM career pathways. 

From the onset, in our design workflow, we have organized multiple focus groups with representative cohorts of the youth we are serving through this tech solution in order to guide and inform our design strategy and roadmap. They have influenced anything from features prioritization to the choice of careers to simulate first, and even the color schemes used in the 3D computer-generated scenes. 

We’ve had direct user feedback from girls confirming that it’s crucial for them to be  exposed to inspiring female role-models in STEM in order for them to be able to visualize the possibility of success in these fields. Although we proactively seek out female professionals to build the career simulations with them, sometimes for some STEM careers, it’s just hard to find female pros. So based on this user feedback, we started working on a specific add-on to the career sim to address this specific challenge: to harness generative AI to assemble a crew of diverse embodied conversational agents that act as female AI mentors. 

We work with girls of all backgrounds and cultural identities. Miami-Dade is an extremely diverse county. It’s very important for us to assemble a diverse team of AI coaches so each girl will find the AI mentor that she most resonates with, and she can see herself in. 

This solution's secondary beneficiaries are teachers. In the “What (and Who) Is Holding Women Back in Tech?” Report, Delphine Donné, general manager of personal workspace solutions at Logitech explains: “The problem often is that bias can be both conscious and unconscious. Whether it is a parent, a teacher, a coworker or a leader, we need to help them become aware and overcome their unconscious bias, for example, through training. This is important for both men and women. » This is why our solution is also meant to be used by teachers to help raise awareness about their own unconscious biases. Embarking teachers in this process will help create a favorable environment for our girl users where the girls will be accompanied by increasingly self-aware authority figures who will be better positioned to provide them with bias-free career orientation guidance. 

We are a female-led majority-female interdisciplinary team who has been working on bridging the gender gap in STEM careers for the last 2 years. 

TEAM + LEADER REPRESENTATION 

We are a majority-female team led by a female creative technologist. In many ways, we embody the issue of female empowerment in STEM. This is a challenge that hits close to home, as we have been confronted first-hand by the difficulty of being taken seriously by venture capitalists as females in tech. 

Not being able to raise capital in the current configuration where only 2% of VC funds go to female-led teams in the US (and 1% in Europe), we have chosen to bootstrap our social enterprise, and in the process, we have won multiple multi-year 6-figure contracts with governments and an innovation award by the National Academy of Sciences and Medicine. Knowing exactly how it feels to be discounted because of our gender, we are supercharged with motivation for this project. Along the way, it’s been crucial for us to keep feeling inspired by pioneering women who have come before us and paved the way. Building a tech that makes female empowerment and mentorship in STEM ubiquitous is one of our lives’ missions.   

EXTENSIVE EXPERIENCE 

We are also perfectly positioned to address this challenge since our interdisciplinary team has extensive experience working on the gender gap issue for STEM careers. We have been developing a career simulator to empower underserved youth to test-drive different professional careers on 2 continents for the last 2 years. Among other features (such as virtually visiting workplaces and virtually job shadowing), you can also virtually embody counter-stereotypical figures of success, and notably women in STEM. As mentioned in the response to the previous question, we have designed, developed and tested this career simulator in partnership with Miami-Dade County Public Schools and the French department of Labor.

For the past 2 years, we have experienced the difficulty of finding female role models in STEM careers to build the career simulation in the first place. We have witnessed the importance of representation in the career simulator as a lot of the feedback coming from female students comments on the presence of counter-stereotypical figures of success in STEM. However, it is sometimes very challenging to find female STEM professionals in certain career fields to begin with. Hence the idea of memorializing female feats in STEM through generative AI and embodied conversational agents, with this  "League of Extraordinary STEM Ladies". 

We have successfully tested the Career Simulator with a little over 30,000 users in the US and Europe, alongside Miami-Dade County Public Schools and the French National Department of Labor, with a 95% satisfaction rate. 

The specific add-on to the career sim, which is the solution to this challenge, has not been distributed to the target populations yet.  

We would be thrilled to become part of the SOLVE community and more particularly have access to a diverse group of resource partners, benefit from the personalized coaching opportunities and the peer-to-peer network. 

For having met individuals from the SOLVE network, there is a huge overlap in terms of ethical design and impact-focused methodology. 

It would be hugely beneficial for us to follow the development modules to make sure our theory of change and strategy for scaling are optimal, as we roll out this specific add-on to our Career sim. 

Dr. Alexandra Ivanovitch as a female entrepreneur in tech has faced numerous challenges to raise funds, and to just even be given a chance. Making it easier for women to succeed in STEM career pathways is her life mission. 

From a cultural perspective, as a US citizen of French and Montenegrin descent, she has resonated deeply with the diverse and vibrant immigrant communities in Miami-Dade County, where she resides. More than half of Miami-Dade's residents were born outside the U.S. This immigrant haven has felt like home from the start. 

PERSONALIZED MENTORSHIP FROM THE PAST TO THE FUTURE 

The solution immortalizes the most brilliant and pioneering female minds in STEM and makes them readily available to inspire future generations of female groundbreakers, and so on and so forth. According to Descartes, “reading all good books is like having a conversation with the most honest people of past centuries who wrote them, and even a studied conversation, in which they reveal only the best of their thoughts.” (René Descartes, “Discourse on the Method of Rightly Conducting One's Reason and of Seeking Truth in the Sciences”). The solution harnesses the latest advances in generative AI to transform inspirational role models into realistic and relatable embodied conversational agents. So that girls across the country have the impression that they are truly conversing with and befriending the greatest female STEM minds of all times. 

The solution is positioned to catalyze further edtech innovation by spurring an AI tutorship / mentorship / coaching revolution in the nascent field of personalized education. 

ALGORITHMIC EQUITY FOR GENDER ISSUES 

It is now all too well documented that one of the challenges to AI safety is algorithmic bias.  On the specific issue of the STEM gender gap, despite explicit intentions to be gender neutral, a Facebook ad for STEM careers was shown more often to men than women, potentially due to economic forces and competition among advertisers. With algorithmic bias, society runs the risk of encoding and indefinitely perpetuating human biases. The solution is truly innovative because we turn algorithmic bias upside down : we trained inspirational STEM figures on female examples only. This is the “affirmative action” moment of algorithmic equity for gender issues. 

The solution could very well catalyze a wider algorithmic equity movement on other key dimensions of diversity. 

Our overarching goal is to ensure STEM education continuity and decrease drop off rates at various key moments in girls’ educational journey (grades 8 and 12).  

We aim to decrease drop off rates by 25% in the first year with the pilot-test, and ultimately divide drop off rates by 2. 

Here is your impact roadmap in terms of scale and number of students we can reach through the solution: 

• In one year from now : pilot-test with culturally diverse cohort of 2,000 female students (from grade 8 and grade 12) and teachers, in partnership with Miami-Dade County Public Schools, and refine prototype based on initial user feedback;

• In 2 years : extend to 10,000 students in 3 different states to make sure the solution can adapt to different cultural contexts with finalized product;

• In 3 years: impact 100,000 students in 10 states; 

• In 5 years : impact 1M students in all states. 

Below are the key measurable indicators we plan to measure our progress on a variety of dimensions: 

• Number of monthly active users and distribution across subsets of targeted populations (female students + teachers); 

• Churn rate to assess the solution’s stickiness;

• Net promoter score; 

• Our users’ retention rates for STEM tracks, as compared to the school district average retention rate observed in female students. 

Below you shall find our actor-oriented theory of change framework: 

For our female students users: 

• Activities: repeated exposure to, and relationship building with female inspirational STEM leaders who can coach K-12 female students as personable embodied conversational agents, 

• Immediate outputs: increased confidence in STEM skills, increased identification with inspirational figures of success, which makes it easier for girls to visualize themselves as being successful in STEM, and helps them counter stereotype, 

• Outcomes: decreased drop off rates and increased retention rates at various key moments in educational trajectory from K-12 to undergraduate years. 

For our users who are teachers: 

• Activities: repeated exposure to, and relationship building with female inspirational STEM leaders, 

• Immediate outputs: decreased stereotypical negative association between girls and STEM, 

• Outcomes: increased ability to coach girls to keep on studying STEM at various key moments in educational trajectory from K-12 to undergraduate years. 

Our theory of change relies on the following key assumptions: 

• Exposure to female figures of success helps girls manage and overcome stereotype threat, as evidenced by peer-reviewed research; (McIntyre, R. B., Lord, C. G., Gresky, D. M., Ten Eyck, L. L., Frye, G. D. J., & Bond, C. F., Jr.  (2005). A social impact trend in the effects of role models on alleviating women’s  mathematics stereotype threat. Current Research in Social Psychology, 10(9), 116–36.)

• Conversational agents are sticky, easy to use and adopt, including by our 2 target populations, as evidenced by the rapid adoption of chatGPT. 

• It is important to target both girls and teachers, since one of the most influential factors that determines whether girls will pursue a career in the technology industry is having a parent or teacher who encouraged them to study computer science, according to the « What (and Who) is Holding Women Back in Tech? » research report by Logitech and Girls Who Code published in 2022.

In its prototype phase, the solution is powered by a combination of multiple generative AI technologies: 

• GPT-4 for now, and in the future, subsequent iterations, to produce conversational agents modeled on real-world STEM female figures through prompt engineering, 

• Text-to-image AI Midjourney to produce evolving avatars of real-world STEM female figures that match the user’ preference, so the AI mentors grow together with the users, 

• D-ID generative AI to match the LLMs with the evolving avatars. 

The user experience for female students is straight-forward: 

• The user goes through a quick quiz to get matched with a mentor that is specifically adapted to their preferences, interests, challenges of the moment,  

• The user and the mentor interact through an easy-to-use web app, 

• The user can change mentors at any point in their journey, the initial match is merely a suggestion;

• Because we are ultimately targeting populations of a wide variety of ages, the user can select different iterations of the same mentor based on different ages (eg. Rosalind Franklin at age 10, 15 or 22), so that the mentor can ‘grow up’ alongside the female student. 

Below is a high-level presentation of our roadmap in the months to come: 

• Test the relevance of our first selection of diverse AI mentors with our pilot cohort and add more inspirational figures based on direct user feedback, 

• Test the relevance of our embodied AI approach, the underlying hypothesis being that it is more impactful for girls to interact with an AI that has a face they can easily identify with, 

• Test our pilot-test engineering hypothesis, according to which we are targeting not only girls and also teachers, in order to transform teachers' perception as well, given the influence teachers exercise over girls and their career orientation. 

We are a lean interdisciplinary team composed of 4 full-time staff: 

• 1 prompt engineer,

• 1 3D artist,

• 1 UX designer,   

• 1 project manager.

We also have 1 outside consultant. 

We have been working on the career simulator for the last 2 years, and ever since we started, it was clear to us that one of our biggest challenges was figuring out how to  best expose girls to counter-stereotypical figures of STEM success. 

We tried to proactively seek female professionals to film in 360° and create 3D animated avatars based on their appearance. But there are still too few.

 That’s why we had the idea of creating this gender STEM project within the wider career simulator project 3 months ago. 

We are a female-led, majority female tech team. Diversity is not an afterthought, it’s our origin story. Apart from gender diversity, our team has 4 continents on board: America, Africa, Asia and Europe, with 4 nationalities on board: US, Algeria, Sri Lanka and France. It’s an inside joke of ours that we are not just developing transformational experiences for people ; in our own microcosm, we are also prototyping a micro-world government. 

We have had the chance to work on the career simulator in extremely diverse settings, such as Miami-Dade County Public Schools for the US portion of it. Miami-Dade County is considered one of the most diverse counties in the United States. It is known for its multicultural and multiethnic population, with residents representing a wide range of backgrounds and nationalities:  a large Hispanic and Latino population, with individuals of Cuban, Puerto Rican, Colombian, Venezuelan, and other Latin American origins, as well as significant populations of African Americans, Haitians, Jamaicans, Dominicans, and people from various Caribbean and African countries.

From the onset, we have pledged to build diverse experiences and services for our beneficiaries. We do not want the career simulator to reproduce gender and racial gaps in career orientation. We proactively sought counter-stereotypical figures to represent success in career pathways that offer the perspective of financial freedom for communities who might not even consider those as attainable. 

We are adopting the exact same approach for the diverse crew of AI coaches. We want to develop a service that matches each girl in America with the STEM mentor that will make it click for her, and her specifically. Because of our commitment towards diversity and inclusion, we opted for the tech framework of offering embodied conversational agents, so girls could actually look at the inspirational figures of STEM achievement and see females that look like them. 

Our mission is to enable each girl in the nation to meet the inspirational female mathematician or scientist that paved the way for her to be able to break gender boundaries today. 

For this project, we adopt a subscription-based business model where customers will pay an annual fee based on the number of users. 

• Our key beneficiaries are the K-12 female students and STEM teachers. 

• Our customers are school districts and private schools. 

Our initial plan is to keep the service free for key beneficiaries, and have school leadership pay for it. 

Our services include both pure SaaS and consulting services: 

• Access to the diverse crew of AI mentors for key beneficiaries, 

• Upskilling for teachers to get them up to speed with generative AI and how it can help counter real-world stereotypes,

• An evaluation plan to help school leadership assess the impact of the solution. 

Our plan to fund the development and expansion of this project is to combine multiple funding mechanisms: 

• Sell this service to the school district we have already sold the Career Simulator to, as an add-on (the nation’s 4th largest school district, Miami-Dade County Public Schools) 

• Raise investment capital (application in progress with several accelerator programs such as OnRamp Education and Workforce Innovation Accelerator)

• Keep on applying for grants and challenges. 

So far we have managed to achieve financial sustainability by combining a variety of funding instruments: 

• Primarily revenue from B2G contracts: 148k USD with Miami-Dade County Public Schools, 111k USD with the US Institute of Peace (VR training for conflict resolution), 240k USD from the French Government (for the Career simulator and a Paris 2024 Olympic Games-focused Career Sim);  

• Innovation award from the US National Academy of Science, Engineering and Medicine: 50k USD,

• Investment capital from Techstars: 120k USD.  

As a female-led, majority-female team, we overcame the gender gap in VC funding by proactively seeking to become financially sustainable by generating revenue. Especially since we are operating with a tech stack (extended reality : virtual reality and augmented reality) where there is quite a bit of noise and a rocky hype cycle, it was important for us to prove that we bring value to customers for real, and historically conservative customers like governments, and that we can get paid today. Hence the current distribution of our funding streams.