If the number of women pursuing STEM careers is increasing, and employers are stating that gender balance is a top priority, what’s going wrong? Sara Sanford and the AdaMarie Inclusion Lab set out to understand what has worked, what makes it effective, and how organizations can gain an edge by becoming career destinations for the next generation of women in STEM.
So far, approaches to retaining early-career women in STEM have fallen short. Despite the best of intentions, most DEI efforts have failed to achieve organization-wide change. To identify meaningful solutions, AdaMarie sought to understand what’s not working, so we can find solutions that will.
These are the blockers keeping employers and employees from reaching their full potential:
Silos Instead of Systems
In our research, we found that the majority of inclusion efforts in tech companies focused on recruiting underrepresented employees, conducting diversity trainings, and hosting empowerment-focused events such as all-day conferences and diversity celebrations. Many efforts put the onus back on underrepresented employees to motivate each other to overcome persistent obstacles. While communities such as affinity groups may provide a source of solidarity, no amount of mutual support can overcome inherently biased systems.
The roots of gender imbalance in the workplace aren’t addressed by these initiatives. The causes are systemic, under the surface. They’re part of the day-to-day functions that become the muscle memory of organizations. Here are a few examples:
- Allocation of projects, clients, and accounts: In firms that evaluate employee performance based on client tenure, satisfaction, and spending, women often appear to be under-performing, when compared to their male counterparts. These evaluations, however, don’t tell the full story. Accounts and clients are often allocated inequitably. Because managers assume that men are better at working with clients, they will assign or transfer them accounts that already have successful track records to start with, avoiding “risking” these accounts on women. Women are assigned inferior accounts, or given new accounts instead of transfers. The superior client relationships that men have access to can lead to gendered differences in performance-based evaluations and pay. Women who feel they’ve been assigned clients with relatively successful track records, however, are evaluated as performing similarly to their male peers.
- Names being left off of reports or patent proposals: Female survey respondents across disciplines in STEM reported having their names left off of papers to which they contributed, being left out of invitations to present on projects they supported, and not being named as inventors on patents. Our findings echoed a 2022 study published in Nature , which found that “Women are less likely to be named as authors on articles or as inventors on patents than are their male teammates, despite doing the same amount of work.” A data set covering 10,000 research teams controlled for the size of the researchers’ roles in their projects, and found that men who played a similar-size role as women were twice as likely to be named on scientific documents.
- Unstandardized performance evaluations: Our survey results and interviews found that women were more likely to state that they had been assessed inaccurately in organizations that did not have quantifiable criteria in their performance evaluations. Making matters worse, employees in tech companies were the most likely to state their employers lacked standardized or well-defined evaluation criteria. In addition, women that were rated as underperforming in technical areas were more likely to experience penalties (being removed from important projects, for example) than men who were rated as underperforming.
Big Data = Big Bias
As STEM employers become increasingly dependent on AI, a pernicious assumption is spreading that software will allow them to escape the influence of bias. The mathematical models that underpin AI hiring and performance evaluation models, however, hold immense potential to scale bias, rather than limit it.
Algorithms don’t just appear. Human beings, with bias, create them. For processes such as hiring, this has negative consequences for qualified, underrepresented applicants, and the employers who want to hire them; machine-learning-based recruiting systems have been shown to downgrade resumes that include the word “women” or otherwise indicate that the applicant was a woman. Multiples studies have found that resumes submitted by graduates of women’s colleges, for example, or that include the term “women’s chess club” were rated less favorably, compared to resumes of similar caliber submitted by men.
This biased sorting is a result of the ways that machines “learn” in the algorithm development process. To learn how to hire, machines are fed years of hiring history, which informs algorithms that determine whom to accept or reject. Hundreds of thousands of resumes marked “accepted” or “rejected” are scanned, interpreted, and sorted into a framework for what “hirable” looks like. Because tech companies have historically hired a disproportionate percentage of men, especially for technical roles, men are coded as hirable.
Despite the risks posed by AI-driven hiring tools, over half of managers in STEM companies report using them.
Information Overwhelm
Expectations on managers and executives to champion DEI continue to rise, but leaders who want to do the right thing often become overwhelmed. Many managers and executives start to explore DEI strategies and become paralyzed by an ever-changing DEI vocabulary, contradictory messages, and seemingly endless nuance. If they do implement new inclusive policies, they find a new segment of their employee base is angry at them.
Employees echo similar sentiments. Those from traditionally well-represented backgrounds want to support their peers as allies, but they don’t know where to start or are worried they’ll say the wrong thing. Women and employees of color are encouraged to stand up for themselves, only to be told they’re too aggressive or arrogant. Other employees see these consequences and then avoid speaking up.
Whatever the particular mix of overwhelm looked like in the organizations we researched, the outcomes were repeated: underestimated employees eventually realize their only choices are to assimilate or leave, and the businesses and employees both miss out.