Women in AI (WAI) is a nonprofit do-tank working towards gender-inclusive artificial intelligence (AI) that benefits global society with the mission of increasing female representation and participation in AI. WAI believes nurturing more AI-empowered future STEM decision-makers who are women and come from diverse backgrounds is critical for the non-biased AI future. It is a community-driven initiative bringing empowerment, knowledge, and active collaboration via education, research, and events.
On June 9th, 2020 — we launched the Denver CO chapter of WAI.
We were joined by Cheryl Ingstad, the director of the Department of Energy’s Office of AI and Technology (AITO); and Rui Yang, a senior researcher from the National Renewable Energy Laboratory (NREL). The two women are in the arena and are leading by example: Cheryl as the inaugural director/leader of AITO — the office that is serving as the central body responsible for the development, coordination, and application of AI in the Department of Energy and Rui as a hands-on engineer researching new ways of solving clean energy integration challenges using AI.
Why do we need more female and minority representation in the technology field, AI to be specific?
The use of digital technologies is increasingly leading to the use of algorithms (which are set of processing instructions coded by humans) that influence our everyday life — for instance in granting loans or college places, preselecting job applicants, or supporting medical diagnostic processes. Artificial Intelligence systems feature special algorithms and rules that can learn and handle uncertainty and incomplete information.
The AI systems deployed today are largely trained using supervised learning — which means the decision-making behavior of the algorithm heavily depends on the data being used to train them.
If the data is not sourced correctly or in other words is predominantly made of samples of male profiles, there is a problem — the algorithm trained with such a dataset will not be able to provide correct recommendations and decisions for female or minority users/profiles. One such example, from the recent past, is — Facial recognition systems not detecting women of color’s faces. What if this facial recognition system is being used to build an AI system for diagnosing skin cancer for which accurate detection of skin color and its variance matter crucially?
AI is rapidly seeping into various sectors of the society, enhancing the effectiveness and efficiency of the operations and services provided by industries such as healthcare, financial services, retail, and media. It is teaming with the supercomputers to help gain valuable insights into technical and medical conundrums of the time. That is why it is crucial to address the issues of biases to ensure the gender-equitable design and use of the new AI applications as we integrate it deeply into our modern world.
Theoretically, there is an agreement that removing biases in the training data, by providing ample data points for female and minorities can help close the gender gaps but if the “right” questions are not asked during the data collection process, gender gaps can actually widen when algorithms are misinformed by data.
Let’s think about the question we posed earlier one more time — Why do we need more female and minority representation in the AI field? How will it help?
Short answer — Because humans are responsible for building algorithms, collecting and curating data to train them.
Slightly Long answer — Bias in AI algorithms doesn’t only occur because of problems in training data. When you dig deeper, it becomes readily apparent that how an AI developer frames a scenario or problem, also contributes to the bias. In turn, this perception informs how and what type of data is collected and utilized.
Perhaps the AI developer forgot to include a subset of the population for their study. Or maybe the bias arises from the process of humans manually marking and formatting the data for AI system consumption. Either way, by this point, it’s likely that the bias will flourish with repeated training over time and eventually degrade the decision-making power of the AI system, especially for female and minority users, due to the amplification of the bias injected in the beginning stages of development.
At this point, we can clearly see that having a diverse talent pool (with considerable female participation) can help address these sources of biases.
Therefore, diversifying the participants in various roles — decision-makers, programmers, data managers, business leads does seem like a promising way of ensuring that historical gender bias does not get amplified and projected into the future. But according to the World Economic Forum, only 22% of AI professionals globally are female, compared to 78% who are male. These stats are not very different from the over-all STEM field participation of the female workforce.
That’s why there is a lot of work that needs to be done to fill this gap.
The talk, titled “Pivoting to a career in AI industry from a conventional STEM background”, shed light on the fact that there are multiple ways to get involved in the AI field, even if the traditional education is not in computer science. Speakers shared their authentic and inspiring personal anecdotes about their own STEM trajectories that led them to their current roles. The goal of the discussion was to encourage the interested audience — students as well as working professionals — to take action toward pivoting into this field.
The recording of the virtual launch webinar is available here.
About the author: Sakshi Mishra is currently a researcher at the National Renewable Energy Laboratory working on the nexus of energy and AI. In the past, she has worked as a grid planning engineer at American Electric Power and currently holds a Professional Engineer License in the state of California. Sakshi is a Carnegie Mellon University graduate. She is leading the Women in AI’s Denver chapter and hopes to raise awareness about the existing challenges and increase engagement of diverse talent in the field of AI.