We have all read stories of facial recognition software that fails to recognize dark-skinned faces, or robo-loan officers that deny mortgages to certain groups. As a growing body of research has made clear, algorithms created by non-representative groups have resulted in AI that perpetuates the inequities already prevalent in our society. As more companies rely more heavily on data and AI, these problems of algorithmic discrimination may only become worse.
Most companies know this by now. What they’re trying to figure out is: how can they avoid becoming yet another bad example?
The short answer is, thinking critically about the data you’re collecting and how you’re using it needs to be everyone’s job. Expanding the circle of who is in the room helping to question, build, and monitor algorithms is the only way that we will develop responsible AI. Doing that work requires data literacy — the ability to parse and organize complex data, interpret and summarize information, develop predictions, or appreciate the ethical implications of algorithms. Like math, it can be learned in beginner and advanced modes, spans multiple disciplines, and is often more practical than academic.
Building up data literacy in an organization can also help diversify the data teams who are at the forefront of making critical decisions about how data will be collected, processed, and deployed. The importance of diverse data teams is something I learned firsthand over more than a decade as a quant fund manager. It’s a commonly held belief that more diverse portfolios outperform because they reduce risk. But it is analogously true that diverse teams outperform because they reduce the risk of groupthink. By investing in data literacy across the enterprise, businesses can bring more divergent and creative perspectives to bear on both mitigating the risk of algorithmic bias — and identifying other efficiencies and opportunities that data can often reveal.
But a look at the data tells us that most companies are still struggling to build data literacy. Ninety percent of business leaders cite data literacy as key to company success, but only 25% of workers feel confident in their data skills. Not only that, but some estimates suggest that nearly nine in 10 data science professionals are white, and just 18% are women. Research from General Assembly indicates that when it comes to diversity, data science lags behind even other tech-oriented disciplines, like digital marketing and user experience design.
Why, despite the obvious need and increasing urgency, are we not teaching data literacy systematically and at scale?
Here are some of the strategies to better understand data literacy.
Data literacy is not a technical skill. It is a professional skill. Encourage all of your employees — marketers, sales professionals, operations personnel, product managers, etc. — to develop their data literacy through quarterly engagement sessions that you host, where you cover topics like data-driven decision making, the art of the possible in AI, how data connects to your business, ethics & AI, or how to communicate using data. This kind of organization-wide emphasis is the basis for a transformation to a data-first culture.
The world of data is big, filled with buzzwords and misunderstanding. Develop a view as an organization which components of data literacy matter most to your organization — if you are a financial services firm, it may be probability and risk measurement; if you are a technology firm, it may be experimentation and visualization. In your L&D sessions, develop learning content that uses this language and demonstrates how it connects to your business in multiple departments, so employees can connect all the dots between data literacy and their workflows.
One thing we recommend to all of Correlation One’s clients is to empower employees to generate new business ideas that apply their data literacy. For example, suppose your company is in the music industry. As part of your L&D program, have employees develop project proposals that leverage their newfound understanding of data literacy — combining it with the knowledge they have of the industry, they will generate surprising new ideas for cost savings or revenue generation. Just as importantly, you will be empowering them to drive a new data-first culture from the bottom up.
Take your current process for approving ideas or setting budgets. Then add mechanisms that reward data-driven thinking. For example, require managers to include clean visualizations in their proposals, or to build dashboards that track their KPIs quantitatively and in real time. If you can shift your managers’ decision-making from intuition to data by granting faster project approvals or larger budgets for proposals made using data-driven thinking, you will quickly get the behavior you seek from your managers through incentive alignment.
Subscriptions to education and training platforms like Coursera often fall flat in organizations that are looking for lasting transformation. That is because learning is much more effective when it is social (done with others), personalized (done with expert feedback), and contextual (connected directly to the business problems you are solving). Developing these personalized, social, contextual learning programs is more resource intensive, but the benefits in terms of employee engagement with the material, employee retention of the material, and empowering your employees is worth it.
Perhaps most importantly, my experience both before and during Correlation One has helped me understand that data is not a vertical — it is not just one job family, like a data scientist or data engineer. Instead, data is a horizontal — it is a skillset that cuts across a growing number of jobs in every field. A marketer is a better marketer with data skills. A product manager is a better product manager with data skills. And so on for operations, engineering, sales, and even HR. Not everyone needs to know how to code. But soon everyone will need data literacy.
Ultimately, data literacy is about much more than machine learning and data science. And it’s about more than AI. Data literacy is simply about humans coping better in a data-infused world — which is why we need it now more than ever.