What If Data Scientists Had Licenses Like Lawyers? | Hacker News Reader

What If Data Scientists Had Licenses Like Lawyers?

bloomberg.com

22 pointsabecode16 comments

What If Data Scientists Had Licenses Like Lawyers? .. then very few "data scientists" positions would remain, most of them would become "data analysts", "software researchers (analysis-focused)" or some other title which does not need licensing.

The author of this article did not seem to think about this too deeply -- the bar exam works because it grants extra privileges to those who passed it; and barber certification works because it is easy to tell barber from non-barber. Neither is the case for data scientists.

IMHO, if one wanted to add regulation in this area, the "working" approach would be to add restrictions first -- like only licensed people can work with personal data.

But then we'd have to worry about cases like "there is a team of 's/w engineers' writing the code and a single licensed person who can only press 'run' button". And fixing that would require even more regulations... which will probably transform the B2C software landscape, and it is not clear if it would be beneficial or not.

I dont know how feasible it would be, but an official lic/cert to allow someone to see personal data (I assume data with name, location and other sensitive data like credit card, social security num, etc) is extremely interesting. Pretty sure someone can make a quick list of hurdles and potential exploits, but still... I think theres a public value in that as the accountability could reflect something like architect and engineering plans for a building. The name and lic are on the plan so you can fry the person who made an egregious error. A similar mechanism for exploiting personal data (within legal parameters) would be nifty. I like it, but since products are not owned but licensed by consumers these days means that's a crack induced pipe dream.

> I dont know how feasible it would be, but an official lic/cert to allow someone to see personal data (I assume data with name, location and other sensitive data like credit card, social security num, etc) is extremely interesting

Actually that's brilliant. Forget about the transition period for a moment, what do we want the industry steady state to look like. And I think it's exactly this. You are licensed and trained to work with personal data. Your license is on the line for misuse. You have all the incentives in place for the right behavior, and companies have the right trade-offs to consider when choosing to use it or not.

You can do tons of software development without needing personal data, and when you do need it you have someone in place who is accountable for it.

Also, just like other specialty lics/certs, I want them to command much higher salaries. Increases the cost of even trying to exploit personal data. The company needs a real damn good reason to do so. Taxes, re-certs and randomized audits would be nice too. I dont believe the punishments should be too extreme. We aren't talking about something as serious as heart surgery, but maybe close to legal advice type of standards. Personal data exploits won't neccesarily be a life or death situation except for the rare and extreme cases, just like legal advice, but it surely can easily fuck up your life, just like bad legal advice.

I know that needs fine tuning and it's out of my wheel house to nail down, but it's something I'd like to see.

Yes! I rarely see others mention this. It's always just "licence to call yourself an X." As though the job title itself is something important.

The other thing to note is that certain jobs can be easily offshored. If you can't get a local (data scientist/software engineer/ journalist) to do what you want, you can just pay a foreigner to do it. The same is not true of doctors and lawyers, due to the extra privileges you point out.

Doesn't GDPR already require that companies restrict and document who accesses PII?

Yes, let’s take a fast moving industry and make them take exams based on what was hot 10 years ago.

I work with actuaries. Basically they are accredited data scientists. Oh scratch that, they would have been data scientists if the CAS exams hadn’t stagnated their knowledge. instead of keeping abreast of the industry, they devote all their time to study for antiquated exams that just serve as barriers to entry, but it does keep their jobs safe.

Also ones of the key arguments the article uses isn’t just applicable to accredited professions, everyone can go to jail or be fined if they lie to a judge. This is the first major reason the authors cites... I propose authors should have to be accredited so they can lose their license when they produce negligently thought out arguments.

> “Its drafters could also borrow ideas from actuary licensing, such as a thorough understanding of linear algebra and statistics”

Dear lord! Noooo. This Person is out of their mind. The cas exams promote rote memorization of simplified formulas and problems rather than actually reasoning statistics.

these days lawyers decide who we can kill overseas and haven't been disbarred to my knowledge: https://rb.gy/ejzin2

As a data scientist, and owner of a data science company, this idea is certainly tempting on the surface but unlikely to be effective or necessary in practice. I feel like in the long run, the market is generally pretty good at weeding out bad apples. There will always be companies who want to hire on the cheap and have to learn the hard way that qualified talent comes at a premium.

> The ethical standards would have to recognize the peculiarities of the job. As opposed to lawyers, who interact directly with judges and clients, data scientists interact primarily with their employers, typically large technology companies. Their loyalties are split between the people who can fire them and the public they might be harming.

The point of the license wouldn't be to protect you from them. It would be to protect us from you.

I don't think it's as easy as the author suggests, but I do think it's an interesting idea.

Part of why everyone has to do esoteric leetcode problems every time they interview with a new company is because there is no uniform hiring standards. Imagine taking the Bar exam each time a lawyer wanted to jump to a new firm.

Licensing could help improve the signal to noise ratio for hiring standards at least.

The concept of science and licensing bodies is incompatible. Peer review is sufficient for scientific articles. The idea of licensing scientists puts us back into the middle ages. The opportunities to politicize and raise barriers of entry to scientific practice have an understandable appeal. Think of unions, hair dressers having to spend $5k on cosmetology degrees, M.Ed.'s. Hopefully this licensed profession would be given a name other than scientist - maybe data engineer, analyst, doctor, or technician to avoid confusion. What's needed is more peer review of data related practices, not credentialing and fake authority.

Wow. Talk about bad ideas! Anyone who thinks the bar exam increases the competency of lawyers doesn't know a lawyer.

I would openly welcome this. As a data Scientist who’s seen some dodgy shit happen within the various companies I’ve consulted with, I think this would go a long way in preventing dangerous models reaching production.

I think a better comparison should be with how the actuarial industry conducts itself. They have been doing “data science” for decades but equally so are bound by professional standards and expectations.

If data scientists had licenses, what do you think the exam for new 2020 grads would look like? Would it include methods introduced in the 2010s? I can't imagine an exam keeping up with the pace of the field. And which field is that?

If there were to be an exam I'd be on board with, it'd be questions about GDPR and CCPA compliance. Even those are moving quickly as courts decide what the laws mean.

LOL, the entire industry would vanish

Reply on news.ycombinator.com