In 2018, Amazon scrapped its in-house applicant tracking system because of one major flaw: it didn’t like resumes containing the word “women,” so if an applicant listed “mentoring women entrepreneurs” as one of their achievements, they were out of luck. The system had been trained to vet applicants by observing patterns in resumes submitted to the company over a 10-year period, the majority of which had come from men—a reflection of the tech industry’s increasingly glaring gender gap.
Some 56% of women leave tech jobs mid-career, double the turnover rate for men. This is in spite of headline-making parental leave policies offered by Google (up to 18 weeks of paid leave); Facebook (16 weeks); and other FAANGs, suggesting an industry plagued by lack of work/life balance and microaggressions towards women.
According to the U.N., the COVID-19 pandemic exacerbated gender inequalities over the past year, as women bore the brunt of childcare during school closures. Unfortunately, the gender imbalance in tech is self-perpetuating—referral hiring, hiring for “culture fit” and subjective evaluation criteria leads to hiring managers selecting candidates based on shared similarities.
Research shows that female role models are key in encouraging other women to excel in areas where they may be underrepresented. “Having women as mentors is just so different on an emotional level,” said Leanne Kawahigashi, a UX designer and Springboard mentor. “You can be a little bit more candid and build a bond faster because there’s a sense of trust when you can do that.”
In honor of this year’s International Women’s Day and Women’s History Month, we’re taking an inside look at what it’s like to work in the tech industry from the perspective of women within four major subdisciplines: UX design, data science, AI/machine learning, and software engineering.
UX design outperforms other subdisciplines in tech in terms of gender diversity. Exposure to the “human side of technology” and having empathy for the user were two main factors cited by students at the UX Design Institute for their decision to start a career in this field, which “tends to appeal to both sexes.”
As a relatively new field, UX design blends a variety of disciplines, and professionals tend to come from diverse backgrounds, from visual arts to graphic design to psychology and more—another possible reason for large numbers of women in UX who have transferred from female-dominated fields. Even so, the majority of women in UX tend to occupy research-based rather than technical roles and hold just 11% of design leadership positions.
These disproportional stats are in spite of the fact that there aren’t that many design leadership roles up for grabs in the first place. Design is often a small department and doesn’t have the same decision-making clout as engineering. For example, there are multiple directors and VPs in the engineering department, whereas, in design, there’s usually only one director. A survey by InVision found that less than 30% of companies polled had a UX designer in a director role, while only 7% held a VP role.
Even so, Kawahigashi said much has changed since she first started her design career in 2014. “I went from being literally the only woman designer on an engineering team of 60-70 people in the beginning of my career to joining teams of all-women designers,” she said.
Female underrepresentation in design leads to a world where most products are configured for the male body. Crash-test dummies used to assess the safety features of vehicles are based on the “average” 50th percentile male body. According to a 2011 study by the University of Virginia, that meant female drivers involved in crashes had a 47% higher chance of serious injury in the event of a collision than their male counterparts.
In the midst of a global pandemic, it’s especially sobering to note that most personal protective equipment (PPE) is based on the size and characteristics of male populations from Europe and the U.S. Only 5% of women employed in emergency services said PPE “never” hampered their work, according to a 2017 report, with body armor, stab vests and hi-vis vests all highlighted as unsuitable.
The ramifications of the gender imbalance in design range from life-threatening—until 2020, women in the Air Force wore combat armor designed for men—to inconvenient. While smartphone manufacturers appear to be locked in an arms race to see who can produce the device with the largest screen, women are finding it harder to operate their phones one-handed. The average smartphone size is now 5.5 inches, nearly as large as the average woman’s hand.
When new grads enter the tech industry for the first time, having a mentor or counterpart with whom to discuss things like impostor syndrome (an occupational hazard for anyone in tech), self doubt or a technical goof-up is instrumental to their growth and confidence—something Kawahigashi says is easier to do around other females.
Having more women in the workplace also makes it less likely that gender discrimination goes unnoticed. When she first started her career, Kawahigashi would occasionally experience gendered microaggressions, such as being subjected to inappropriate remarks or a coworker disregarding her input, only to enthusiastically concur when a male counterpart made the same point.
“Too many times I showed up at a job where I was the only woman in the room with executives and I had to give presentations and I felt like no one listened unless my male manager spoke up,” she recalled.
Several years into her career, Kawahigashi says she changed her focus from “just showing up and standing out” to “being a pioneer for women in tech.” Today, she’s a product designer for LearnIn, an upskilling platform for professionals and a Springboard partner. Through this work, she feels she can make an impact by better preparing women for a career in the tech industry.
“What is inspiring to me is seeing people in leadership positions, even if they’re white men, who are openly talking about this situation and saying ‘I want to find a way to fix it’,” she said. “I think it’s about perception and being open to having uncomfortable conversations.”
Lack of gender inclusivity in data science has resulted in well-documented instances of algorithmic bias, like when Google Translate famously returned “she is a babysitter; he is a doctor” when translating phrases from Turkish, a language with gender-neutral pronouns. Or that time a Goldman Sachs algorithm offered a man a spending limit 20 times higher than his wife, who applied for the same credit card.
Over the last 50 years, breast cancer rates in the industrialized world have risen significantly, but a failure to research female bodies and environments means that the data for exactly what is behind the rise is missing.
The gender imbalance in data science has persisted despite a longstanding talent shortage. The Bureau of Labor Statistics predicts that data science roles will grow 15% from 2019-2029, much faster than the average for all occupations, while the average base salary for a data scientist in the U.S. is $122,582, according to Indeed.
While hackathons and coding competitions have been blamed for creating a false sense of competitiveness in the data science field, a possible dissuading factor for women, the issue is likely more elemental.
A survey from BCG of over 9,000 STEM students found that data science has an “image problem.” In other words, there’s a general lack of knowledge about the profession, which was seen as “abstract” and “lacking purpose” by 49% of students surveyed. The same survey showed that landing a job with tangible impact is substantially more important to women (73%) than it is to men (50%). People who view data science as theoretical or wishy-washy are at risk of dropping out of the talent funnel for data science jobs.
Students are likely picking up on the fact that not all companies are succeeding at creating tangible impact from data science and AI, and the general misconception that data science is just about making predictions. There’s also a lack of information around the day-to-day activities of data scientists and what career prospects you can expect, which can be off-putting to both genders.
“I think that what we need to do in general is teach people about the algorithms in their lives,” said Alison Cossette, a data scientist at NPD Group and a Springboard mentor. “They’re everywhere. We live in a 100% curated world where all of your content on Netflix, YouTube and your news is determined by an algorithm.”
Cosette is passionate about promoting data literacy, which she believes is key to dispelling misconceptions around data science. She notes that certain statistical concepts are already becoming mainstream, with health officials using terms like “flattening the curve” during the COVID-19 pandemic and the Simpson’s Paradox being used to explain the outcome of the 2016 presidential election, a phenomenon that describes trends that appear when data is separated into groups, only to disappear when the data is aggregated.
However, the current gender imbalance in data science leads to a harmful disparity in society: the gender data gap. This phenomenon explains why most of the world’s data is based around the male experience. For example, heart disease survival rates for women are substantially worse for women than they are for men. Historically, heart disease research was primarily conducted on male subjects by male scientists and doctors, so male symptoms are considered typical and female symptoms atypical. As a result, women are misdiagnosed up to 50% more often and are more likely to be dismissed without treatment.
Cosette says it’s vitally important for data scientists to increase the transparency of their algorithms to root out biases. For example, requiring businesses to disclose selection criteria when using machine learning models for classification problems, such as justifying why an applicant was rejected for a line of credit or a job, even if it’s just an automated report.
“Shapley values allow you to look at data observations line by line and see which variables gave rise to a particular row or entry,” said Cosette. “Wouldn’t it be nice if an applicant tracking system could tell you why you were rejected for a job? Then we’d have transparency around the algorithmic impact.”
AI and Machine Learning
Recent figures released by Google indicate that just 32% of its workforce is female. Apple’s global workforce was 33% female in 2018, with just 23% of women in technical roles and the highest share (36%) in retail leadership.
Ever wondered why smart speakers feature female voices by default? “Bot creators are primarily driven by predicted market success, which depends on customer satisfaction—and customers like their digital servants to sound like women,” wrote Leah Fessler, a reporter at Quartz.
This is driven by notions that a female voice betokens warmth, a problem-solving attitude, and subservience. However, owners of smart speakers being rude to their devices has created concern for the underlying motivations that lead people to verbally abuse a machine that does nothing but obey their every command. Research released by Unesco claims that digital assistants such as Apple’s Siri and Amazon’s Alexa entrench gender biases by offering responses that are flirty or subservient to queries that are abusive or sexually explicit. When told “You’re a slut” Siri coquettishly responds “I’d blush if I could” or “Now, now” or some other evasive or flirtatious variation. Amazon’s Alexa is a little more blunt: “Well, thanks for the feedback.”
“If anything, walking into a room full of men makes me a louder advocate for women in tech,” said Archana Vaidheeswaran, an AI engineer at Continental and a leadership fellow at Women Who Code, a nonprofit dedicated to inspiring women to excel in technology careers. “It makes me work to increase representation and is a constant reminder of what I am fighting for.”
Vaidheeswaran, who specializes in edge computing, a type of IT infrastructure with decentralized processing power instead of relying on a data center, says that preexisting biases in AI models are exemplified in edge computing because a machine learning model that traditionally runs on a 64-bit or 32-bit server must be compressed to operate on a smaller computing system.
“A huge model that runs on a server cannot run on a small device because it has less memory,” she explained. “One of the things we introduce when we compress the models is bias, and that’s something which we as researchers and engineers did not think about until recently.”
Professional associations like Women Who Code, a Springboard partner, are key to helping women feel more welcome in the tech industry, Vaidheeswaran added. “We need more women leaders to pave the way for future generations of women in tech,” she said. “Women Who Code helps me foster my leadership skills and has given me a safe space to feel like I belong in tech.”
While the overall female participation rate in tech has climbed slowly over the years, the share of women in computer science has been declining since the 1990s, falling from 36%-25% between 1990 and 2013. Even today, just 6% of user profiles on Github are female. These figures are ironic considering women’s vast contributions in programming—from the first known programmer, Ada Lovelace, to Grace Hopper, who is credited with creating the first compiler, a program that lets users create programming languages that more closely resemble natural language.
One theory that explains the gender imbalance in software engineering is that as demand for computer science degrees surges, universities are becoming more selective in their admissions criteria, with many institutions requiring prior programming experience even at the undergraduate level. Unfortunately, women on average are less likely to have taken a computer science course in high school, so many of them find themselves shut out of the profession.
“Men are more likely to be exposed to coding because they probably had peers steering them in this direction and they had male role models to look up to,” said Sapphire Duffy, a software engineer at Kainos, a software company that develops IT solutions for businesses and a leadership fellow at Women Who Code. “On the other hand, women are influenced into different roles such as nursing or teaching.”
While studying IT at Queen’s University Belfast in Northern Ireland, Duffy recalls being one of very few women in her class and suffering from impostor syndrome. A study by Stanford University found four main reasons why many women show an interest in computer science when they first arrive at university, only to switch majors. Two of these reasons—“isolation” and “lack of role models”—are perpetuated by the preexisting shortage of women in computer science, while “loss of interest” stems from computer science courses that are overly theoretical with limited “visible applications” to real-world problems. Women who were attracted to the problem-solving aspect of computer science are turned off by dry coursework, which does not reflect the creativity involved in real software development.
“I love that tech is a sociable career to be in and it’s not what it looks like—yes, I mean those pictures you see where people sit behind a screen all day coding,” said Duffy. “I work with teams from all over the world and build exciting solutions with them.”
Reshma Saujani, the founder of Girls Who Code, wrote in a blog post that simply exposing girls to coding at a young age isn’t enough. Instead, they need support systems throughout their academic career to help them discover computer science courses in high school, feel comfortable selecting a CS-related university major, and finding a job after graduation.
“Software engineering is challenging and people can sometimes feel overwhelmed,” Duffy concurred. “I believe having that type of support system throughout their careers will help more women stay in tech.”