Interview with Alice Wong, Founder and Principal data scientist, Hyperplane Consulting

While times are changing for the better, the harsh fact remains that men have long dominated the tech space. The dominance of men in leadership roles in business and technology and ways to include more and more women – often become topics of discussion in the media.

To get more and more women to pursue STEM-based careers, reading stories about women who have broken the glass ceiling and taken up challenging roles in the tech space can act as a great source of inspiration. Today, we look at the inspiring journey of Alice Wong. She has extensive experience in the field of data science and has worked at Novartis as Principal Biostatistician and Manager, data science at PwC, among others. She founded Hyperplane Consulting in 2017. She is also a popular creator on LinkedIn and extensively talks about various facets of data science and AI. Wong holds a BA in Economics from the University of Pennsylvania and an MS in Biostatistics, Harvard University. 

Econometrics got me hooked

Wong had already decided on economics, and then, later on, biostatistics, before the data science hype. It was more of a case of aptitude combined with an interest in problem-solving using empirical data or evidence that helped her make this choice.

She says, “I was an economics major, and econometrics got me hooked on Statistics. When I talked to my professor teaching probability about applying to graduate programs in statistics, he suggested Biostatistics as it was newer (spawned a few decades ago) and was more fast-paced in innovation. Not to say, healthcare was a hot topic back in the late 2000s. As you may or may not know, Biostatistics is just like Statistics and is methodology- rather than Biology-oriented in its coursework, so much so that Biostatistics and Statistics programs are ranked together in US News and Rankings.”

Statistics has always been a core field of AI

Wong added that statistics has always been a core field of AI, analytics or data science, whatever you might call it these days. Hence, it was a natural option for her to work in an analytics (as it was called back then before the data science title became prevalent) job right after graduation. 

Download our Mobile App

Getting buy-in from stakeholders

Throughout her career, Wong has undergone challenges with which she has dealt effectively and come out with solutions. Some challenges she encountered as a data scientist include:

• Coming from a very mathematical and coding-oriented field, Wong’s first instinct was always to explain the models in math or ‘easy code’ like SQL. It was only after much experience that she realised most people preferred high-level explanations in lay terms with details relegated to an appendix. Overcoming this challenge just came with more experience, and the recent proliferation of free materials to explain and distil Data Science to a wider, less technical audience has also helped her grasp some patterns in how to explain things in a lay way.

• Getting buy-in from stakeholders: One of the things that would help is keenly observing who gets buy-in and comparing data science with something like its rival (and often still more popular) like Business Intelligence. Then one should think about the distinguishing points of data science like it produces statistical significance, which tables and graphs cannot produce, as the latter does not put together effect sizes and sample sizes in a formula to tell about the level of generalisability to the population of interest.  

Staying updated depends on the role

As technology keeps undergoing changes rapidly, it becomes extremely important for a data scientist to stay updated with the changes to make a long-term career in the field. 

Wong adds, “For an NLP role, the common expectation would be to keep up with the very latest trends from last month or even week. For statistics and machine learning, the repertoire remains fairly standard over time.”

As for tools, Wong believes it depends on one’s approach and personal philosophy. One can learn the tech stack required by a company only after they have requested it or learn a large smattering of tools on the off chance that some company will need one or two of those tools. 

Desire to learn and humility

A good data scientist has some traits that set them apart from others. As per Wong, some of these traits are:

• One who cannot be replaced by AutoML. What this means is that even if they use AutoML, they should be able to piece together different solutions to create a more predictive and/or robust solution.  
• Good statistical inference skills would also help guard against replaceability by AutoML as AutoML in its current state overlooks a few statistical deficiencies such as having single-value mean imputation resulting in decreased variance and the resulting inaccurate p-values.  
• Desire to learn and humility.  

For a college student interested in pursuing a career in data science and AI, Wong has some advice. She adds, “Know the market. In some markets, deep learning jobs are very common for juniors and less so in others. However, my own observation is that, even for deep learning jobs, hiring managers want to see an ability to use simpler models correctly as a first resort, so learning the statistics and ML staples is important even for such roles. There’s also quite obviously a generational gap in the data world that’s rarely openly spoken of, so if you want to impress hiring managers, you’ll really need to try to get into their heads.”

Data Science and AI are glamorous terms

As we know, data science and AI remain heavily male-dominated fields. Wong says, “Data Science and AI are glamorous terms, and whoever’s the most aggressive about grabbing these terms will grab them by asserting the rules of what constitutes a data scientist. Since a data scientist is a social construct, not a natural creature like a bird in the bush for you to detect, whoever is most aggressive about defining the criteria to be a data scientist will have their wishes come true,” states Wong.

Use cases for data science are male-dominated too

The topics that are widely promoted as use cases for data science, like crypto trading, are also very male-oriented. It’s a positive feedback cycle where, once certain topics are set in motion, they attract even more attention to the males promoting them, and then it increasingly seems like data science and AI are male-dominated spaces.

Casual sexism or unconscious bias

There is also the issue of casual sexism or unconscious bias that’s hard to erase because it cannot easily be ‘proven’. Other common microaggressions are automatically assuming the male peer or junior teammate is the female’s manager, paying attention to someone’s words only when a male says them, even after a female has already said the exact same thing. 

Equal voice in determining how a data scientist is defined

To undergo a change, Wong feels that we need to examine the systems in place that lead to these issues in the first place.

A question that needs to be answered is if relevant players from different disciplines have an equal voice in determining how a data scientist is defined. We should also examine the forces at play in causing the ‘pipeline’ problem.

Peggy Orenstein’s ‘Flux’ points to research showing that females do a bit less well at math than males in high school in the US, but by the time they get to college, performance levels out, and a little more than 40 per cent of math majors in the US are female.  

Wong adds, “Are females discouraged from becoming math majors in college just because of their high school performance? If that is the case, what can be done to retain their interest in college? Given that the percentage of females in math and statistics are still pretty high, even if not equal to males, the more pernicious problem I foresee for gender parity in data Science is, as I’ve mentioned, the increasing cornering of statistics and Math majors into jobs with the ‘Research’ title instead of according them more lucrative jobs like data scientist ones.”

Greater focus on AI ethics

The UN has very recently set goals to standardise AI ethics across all countries. There will probably be a greater emphasis on AI ethics, but probably also more complications as you try to impose standards that may not be compatible with the cultural mores of all countries.

Wong concludes, “I have also done my own analysis on data-related trends in earnings transcripts, etc., and speculate that all trends are cyclical. This emphasis on MLOps today will probably give way to something else in one or two years.”