In a previous article, we discussed some of the most pressing concerns in the field of AI, such as lack of human accountability for bad or unethical decision-making by automated systems, and poor understanding of sources of bias, as well as how these issues are fueled by the over-reliance on “black box” methodologies and supposedly “objective” data. As consumers become increasingly more savvy about the impact that analytics have on their lives, and the industry turns its eyes towards both federal and self-regulation to provide a safety net against these concerns, it becomes imperative for any analytics enterprise to keep these issues on the forefront of their strategy as a way to prepare for shifts in infrastructure, methodologies and considerations that will have to be taken into account once the implications of these concerns have fully taken hold.
On this occasion I’d like to focus on one of the recommendations listed at the end of the aforementioned article; “Hire diverse teams.” The benefits of hiring diverse teams are already well understood; in Diversity Matters, a report published by McKinsey in 2015, it was found that companies in the top quartile of gender diversity were 15 percent more likely to have financial returns above their national industry median. When considering racial/ethnic diversity, this increased to 35 percent. Meanwhile, companies in the bottom quartile for both gender and ethnicity/race were statistically less likely to achieve above-average financial returns than the average companies in the dataset.
A Scientific American article by Katherine W. Phillips, Director at the Sanford C. Bernstein & Co. Center for Leadership and Ethics, highlights decades-long research demonstrating that diversity fosters innovation, creativity and information-sharing, by eliminating hidden biases about how we interface with people, like expecting people with a similar background to us to come to the same conclusions, or share perspectives about information only we have.
In the realm of analytics then, it is easy to draw a line between concerns cited above and benefits to considering a diverse team makeup when it comes to working with data and designing bias-aware decision-making systems. As a field in which significant intersections already exist in areas of expertise needed to truly understand and exploit its methodologies and techniques – software engineering and architecture, statistics, even sociology – it only follows that considering other, less obvious axes of diversity such as gender, race, ethnicity and sexual orientation will lead to the surge in innovation, information sharing and fact-checking that is much needed in the current state of the industry. Companies that don’t foster and cultivate an environment of diversity are at a heightened risk of falling into bias traps regarding the data they work with, its handling, and how ultimately it feeds into predictive or profiling systems. They are also missing out on significant insights that could lead to new opportunities, especially when this data comes from a heterogeneous and diverse background that teams are not prepared to recognize or consider at even the initial exploratory stages of analysis.
The heightened risk for non-diverse analytics teams isn’t just conjecture. Anima Anandkumar, Director of AI at Nvidia and professor at the California Institute of Technology says the risks AI systems will cause harm to certain groups are higher when research teams are homogenous: “Diverse teams are more likely to flag problems that could have negative social consequences before a product has been launched,” she says. In Collaborating with People Like Me: Ethnic Co-authorship Within the US, authors Richard Freeman and Wei Huang analyzed 1.5 million scientific papers published between 1985 and 2008 and found that those written by ethnically diverse groups of authors received more citations and had higher impact factors than those that weren’t. The evidence is clear: Diversity deters bias and encourages curiosity, skepticism and analytical thinking; attributes any analytics enterprise will highly value.
“Companies that don’t foster and cultivate an environment of diversity are at a heightened risk of falling into bias traps regarding the data they work with, its handling, and how ultimately it feeds into predictive or profiling systems.“
And yet, reports like the 2018 Burtch Works Study on data science job demand and salaries state that only 15 percent of data scientists, 22 percent of early-career data scientists and a measly 10 percent of data science executive leaders are women. The 2019 Kaggle State of Data survey reveals an equally concerning gender gap with 84% of respondents identifying as males, a number that has barely moved from previous survey iterations. Studies on minority groups are harder to come by; a 2017 article by consulting firm Priceonomics using student enrollment data from General Assembly found that Data Science courses had by far the lowest total percentage of Hispanic/Latino and African-American students.
As the evidence mounts for diverse analytics teams as a way to identify and deal with bias, produce robust data pipelines and capitalize on cultural and behavioral insights provided by diverse datasets, it is all the more disappointing that these benefits remain untapped for a large majority of enterprise data efforts.
During my tenure at Growth Acceleration Partners, I’ve found its value of “investing in people” to be more than a bullet point in its branding material. No doubt our nearshore nature has encouraged and enabled us to engage with a superbly diverse workforce that I’ve had the pleasure to interact with, but there’s a difference between a company that stumbles upon diversity and one that truly understands it as not only a cultural asset or a “nice to have”, but as an integral part of its framework for growth and innovation to be cultivated and relied upon. In the Ideas to Invoices podcast, CEO Joyce Durst was quoted saying “The key thing about having women at all roles in the company at all levels, it’s really about diversity of thought. (…) We can have diversity to get the very best ideas, innovations and thoughts out on the table (…) That includes diversity in gender, age and multicultural”. GAP recently received certification as a Women’s Business Enterprise by the Women’s Business Enterprise National Council.
GAP offices in Granadilla, Costa Rica
This understanding of – and reliance on – diversity has made our work of building and improving our analytics efforts and offerings easier for both seasoned leaders and new hires. As a software-focused company, building a robust analytics practice presented the challenge of hiring outside the fields we were used to. For the first time we came into contact with actuarial scientists, economists and professionals in other areas, SIGs that were relevant to our efforts and programs in academic institutions that we had had little contact with in the past. All of this forced us to revise our hiring and on-boarding processes, but approaching it as just another source for diverse thought proved to be a successful strategy, and has led to the creation of a robust talent and opportunity pipeline, elevating our position within the industry with meaningful successes under our belt.
What could have been a concern about how new hires would integrate with the rest of the company turned into an example of how a company set up for welcoming diversity in all its flavors can easily integrate people from unexpected backgrounds and find itself reaping technical and cultural benefits in the short term, not only for the specific teams they were hired for but for the company at large.
The value of diversity is well-researched and understood. In light of said benefits as well as the particular pitfalls regarding bias the analytics industry and its consumers are increasingly becoming aware of, diverse teams have been shown to be an effective deterrent for these biases making it a foregone conclusion that diversity must be one of the areas the industry must strive to be better at. By analyzing hiring practices for biases that could be impacting minority hiring gaps, recruiting from unconventional backgrounds and accepting these practices as integral parts of a company’s growth and innovation strategy, we’ll not only begin the long path towards understanding and alleviating these issues, but also be prepared to navigate and solve the ethics, accountability and privacy challenges that will inevitably have to be faced this decade. In her profile for 2019’s 10 Most Influential Women Leaders, Anima Anandkumar lends credence to this notion: “I am hoping this is the beginning of a revolution in AI that promotes diversity and inclusion. I strongly believe this will be a critical ingredient for us to develop better AI techniques that promote fairness, interpretability, transparency, and robustness.”
About Sergio Morales Esquivel
Sergio Morales Esquivel is the Principal Engineer of Analytics Strategy at Growth Acceleration Partners, and a professor at the analytics post-graduate program at Cenfotec University. Sergio leads the Data Analytics Center of Excellence at GAP, where he directs efforts to design and implement solutions to complex data-related problems. Sergio holds a B.S. in Computer Engineering, and an M.S. in Computer Science from Tecnológico de Costa Rica. Outside of work, he enjoys traveling, making games and spreading the love for open software and hardware. You can connect with Sergio on his website or LinkedIn, or send him an email.
