Whoever said that arts degrees were useless in today’s data-driven economy weren’t seeing the big picture. Or perhaps the problem is that they were never introduced to John Hearty, data science manager at NuData Security and former philosophy student.
As the world’s economy becomes more technology-driven, increasing numbers of people are choosing STEM degrees in the hopes of future-proofing their career prospects. Earlier this month, a conservative UK think tank even went so far as to suggest that arts degrees were not good value for money and should therefore be scrapped.
That is, frankly, a completely myopic approach to both education and the skills economy. To say that an arts degree such as philosophy is totally useless to someone completely overlooks the transferable skills that can be earned from those types of degrees. John Hearty, the data science manager at Mastercard company NuData Security, is a prime example of this.
“I was one of those kids that had too many questions, so at my high school it was generally agreed that philosophy was a good fit for me,” explains Hearty, reflecting on his initial decision to pursue a philosophy degree at Nottingham University. “Employability was not a factor in my decision-making then.”
After his undergraduate degree, Hearty signed up for a postgraduate degree, then travelled and meandered through “various odd jobs”. At age 25, he decided he wanted to begin pursuing a career. He found a university that would let him pick up an MSc in computer science despite having, as he puts it, “no computing background or demonstrable [computer] skills”.
Even before the course began, Hearty steeled himself for hardship. He spent the summer before he started his course devouring programming books and networking textbooks. “I wrote lousy code constantly. I thought about writing code. I buried myself in the stuff.”
Luckily, his fears were unfounded. He was thrown into the deep end, but landed on his feet. Preparedness served him well. “Suddenly, I was all machine learning, all the time. I was learning R, then Python, reviving my knowledge of linear algebra, calculus, discovering PCA and SVMs and boosting ensembles. I was learning techniques, theory, code practice and visualisation approaches all at the same time. I was solving problems. I was hooked.”
After his studies, Hearty became one of the first members of the data science team at Xbox, where he worked in gaming for a while before transitioning into financial security. Now, he oversees the creation of security, identification and fraud prevention solutions by NuData Security’s R&D team.
Is a formal data science education no longer necessary?
Hearty maintains that gaining formal qualifications is no longer as necessary as it once was. “What really matters is that you have the technical knowledge and practical experience needed to deliver effective and safe solutions.”
Even though philosophy may seem to many an unusual discipline from which to move into data science, Hearty argues that philosophy skills are useful to data science. “Data science is a field with a lot of conditional decision-making. … Creating something useful in these complex and conditional circumstances is usually about working out what matters, and which tools and data are going to provide the greatest upside during the time available.
“Philosophy tends to equip you well for reasoning through ambiguous and interdependent problem spaces, by equipping you to objectively identify what really matters, separate concerns and produce a practical, success-maximising plan. … Philosophy also arms one well for the hypothesis-driven, logical practice of data science.
“I think I developed decent listening skills and a habit of analysing and assessing inputs [in my philosophy degree]. It is not necessary to study philosophy formally to train in these areas, and I would suggest dipping your toes into philosophical writing if you’re interested in trying.”
The ethical crossover
The ethical burden upon data scientists has become something of a hot-button issue in the field today. Though many university courses in data science do try and equip future professionals with a moral sensibility, Hearty argues that this isn’t quite enough.
“I believe that making ethical data science practice the norm takes a combination of personal responsibility, beneficial corporate engagement, and successful public discourse and policy.
“When I say ‘personal responsibility’, I mean the responsibility of individual practitioners to learn the ethical principles related to data science, reflect on those principles’ relevance to their personal data science practice and context, and then find ethical means to achieve their goals.”
Next, he advises that data scientists learn ethical principles and then put them into practice by asking themselves the ethical implications of what they are doing. Hearty picks out sampling bias, a common problem that plagues machine learning specialists trying to create accurate facial recognition software or criminal justice case-processing automation.
“Once you’ve reflected on how your practice and principles relate, you can start to make decisions about sampling, processing and performance functions based on your principles. At this point, you’re functioning ethically.”
Machine learning in particular is an area that Hearty worries about from an ethical perspective. Or, more specifically, he worries about the “regulatory weakness” therein. “Andrew Ng has said that ‘AI is the new electricity’. I tend to think of machine learning as something more akin to ‘the new gunpowder’, something that is tremendously empowering or hazardous, depending on the ends to which you apply it.”
