Data Scientists Fear Extinction

SOA has expressed the fear of actuaries being replaced by data scientists, At least this data scientist claims data scientists will be replaced by those with extensive domain knowledge.

1 Like

Interesting article, thanks. It may have some parallel with the experience I had years ago.

I learned how to program at college and many of us loved to program and hoped to get a job programming. When we got out, there weren’t many programming jobs to be had, as most of the essential programming had been done and what was left (that which didn’t require domain knowledge) was outsourced to countries where labor was much cheaper. This meant a pivot towards roles that were a hybrid of domain and technical knowledge - the roles that were harder to outsource.

Perhaps there will be a similar outcome for data scientists. Not so much from outsourcing (although some of the tasks will outsourced to automation) but when enough of them get domain knowledge, it may become a necessity to have domain knowledge.

Listen, actuaries have nothing to fear from data scientists “coming for their jobs”. There’s no data scientist in the world who wants to spend hours a day poring over the minutiae of insurance regulations and auditing spreadsheets to ensure compliance with all of them. It’s a red herring to say that actuaries will be outsourced by “data scientists”. Actuaries use some generic data science techniques, just like they use marketing techniques and IT techniques and policy administration techniques in the service of their actual responsibility, ensuring that the insurer has enough money to pay future claims. DS ain’t gonna get that bogged down in pedantic details.


I don’t think the specialized data scientist will go extinct.

But a couple of advances in parallel computing and deep neural net expanded modeling to new domains in the 2010s especially. To do this required knowing programming, math, and research skills. Mostly it was PhDs who had these skills, and there has been a great demand for them.

As these new models become more routine, these PhDs who specialize in learning how to do new things will be replaced with people more specialized in those applications, and who probably will have less formal education. The PhDs will move onto something else. It seems like it was finance in the 2000s, data science in the 2010s and 2020s, and we will see what in the 2030s.

Actuaries do need to be ready to pick up these new kinds of models. We see this already with some of the actuarial modeling software, which allows domain (insurance) knowledge to more easily substitute for programming and modeling knowledge.

1 Like

Domain knowledge and an understanding of the limitations of the models/tools/methods is usually the best combination. People who use sophisticated fancy tools without bothering to understand the quirks of the data generating process will get into trouble. My personal opinion is that a lot of the black box techniques coming from computer science are not suitable for most applications in economics, business, finance, etc. Look at what happened at Zillow, for instance.

What you’re saying is that an actuary should be a generalist. But the trend has been to get rid of generalists and instead having specialists interacting together with a tiny group of generalists acting like “glue”. The premise here is that the demand for actuaries will go down as data scientists take over some of the core actuarial work. Do you really want to hire someone who is average at everything instead of having a bunch of pros working together?

They won’t go away. Automating the dumb stuff will hopefully get fewer of them working on how to make us buy more stuff and more of them working on more substantial things.

Ya’ll funny.

1 Like

Welcome, Riley. I remember you on the AO.

Good to hear you still like actuaries enough to find our new forum.