From the course: DJ Patil: Ask Me Anything

Is there a data science code of ethics?

From the course: DJ Patil: Ask Me Anything

Start my 1-month free trial

Is there a data science code of ethics?

“

(light upbeat music) - [Interviewer] So what are some lessons learned when you're trying to set this global guideline or best practices for data collection and disbursement? - Well the first one is, while doctors take a Hippocratic Oath, the administrators and the business people around the hospital may not take the Hippocratic Oath. So they may be perfectly willing to say, hey look, we could sell your data, but may not always have your best interests at stake. Also when a doctor doesn't adhere to the Hippocratic Oath how fast does action take place? Oftentimes it takes too long before action is taken. So we have improvements even in those systems. Now we shouldn't in data science or in technology say, well, see you guys aren't that great, we should actually say, well let's learn from those systems, so how we can even be more effective for society. How can we do more? And part of that starts with the basis of just having an educational framework that starts with a question of how do we have dialogue on this? And that requires classes in ethics, the ability to have case studies for us to have dialogue. One of the ones that's been an idea that's come out of the data science community is just as we have an interview question about cultural fit why isn't there an ethics question asked about data? So you're the data scientist working on our data. We're not supposed to use race as a variable in this model. But we find a proxy for race. Talk to me about what you do next. Now in some cases proxy may be okay. Other cases absolutely not. But what's more important is not how do you make it a definitive line, but how do you approach the problem and how do you suss out the nuances of how to approach it? But what would it look like if every company asked their data scientists or technologists they're interviewing an ethics question? What would it look like if every person who is a data scientist or technologist, when they're interviewing and they had a question, and they said do you have any questions, say hey, how do you approach ethical issues? And judge the company on their response. Maybe what we need is equivalent of some type of descent channel that if you're in this company and working on something and your manager says just do it, and you're like I have a problem, you can escalate it straight up to the CEO or to the board. What if that doesn't work? Maybe there should be a whistleblower protection that you can have where you go to some third-party agency where it says hey, I have a problem with the way we're approaching this. I want to raise my hand and say I'm not sure and you have protection. The time for those kind of conversations is now. And we have to ask those. Those are all embodied in this broader question around ethics that goes beyond an oath, including a checklist when you build a product. A checklist should include these questions of like, well, is there harm? Have we asked how somebody could abuse it? Is there bias in this? What happens when the model drifts? Who updates it? Who's going to manage this? Who's going to keep touching it and make sure that it's not going to go off the rails? What does it happen if we need to shut it down? How do we make sure? What if we have to take away that data? The impacts that we're seeing from data being misused and abused is partly because we have not had those questions asked. People have not done the hard work to make sure that we have done the diligence to have frameworks of that put in. I used to go at conferences and I'd say we should have a session on ethics and there would be a few people that do it. Now there's conferences dedicated to it. And that is a beauty. So we are getting smarter, we're getting better. We need to put it into practice. We need to have it in practice and so that we are actually practicing these measures and getting better about putting it in. (light upbeat music)

Contents