Welcome to the fifth and final module of ethical issues in data science. This module is going to take a slightly different approach than we've taken so far, and it will focus on a particular and very important application, the uses of data science in health care and the ethical issues that arise from those uses. I'd like to provide a little perspective on this choice through a simple visual aid data. Science often is described as a three legged stool, as you see on this screen, with the legs consisting of statistics, computer science and applications. So far we started all of our discussions from issues with data science methods and algorithms, which come from the two legs of the stool that are labeled statistics and computer science. And then in each case we've illustrated issues with several applications. For instance, in the last two lessons when we were talking about algorithmic bias, those applications came from financial issues and from criminal justice and public safety issues. But another approach is to select an application area in which data science is having a particularly important impact and look at ethical issues that are arising from using data science on that application, and that's what we'll do in this module. If we go back to another visual aid, one of the circle diagrams that we used in the very first lesson of this course, you can make a few observations first. Another way of saying what I just did is that so far we've been working our way from the inside out, from the methods to the applications and in this module, we're going to do the reverse. Second, if you take a look at the topics in that outer ring, any one of them could lead to a rich and set of examples of uses of data science in ethical issues that are associated with them. But if we need to pick just one to me, health care and medical applications seem to be an obvious choice for several reasons. First, as we'll see illustrated in this module, that area makes profound use of data science. Second, this area probably leads to more ethical issues than any other one that we could choose, in part because health care in general is linked to so many important and deep ethical conversations. And finally, of course, it's an area that's of importance to all of us, so this module will be organized as follows. The first lesson will look generally at uses of data science and health care and ethical issues that are associated with them. We'll look both at efforts to accumulate huge databases of health data and the general issues that are associated with using AI and machine learning and health care. The second lesson will look specifically at two emerging areas that both have profound consequences and ethical issues associated with them. Genetic editing and neurological interventions. These both are certain to have major impacts on society in the coming years and decades. The third lesson is a bit of a cheap, and there is only partly tied to healthcare and medicine. In my opinion, any course on ethical issues in data science or ethical issues in computing would be incomplete if it didn't comment on what may be the most overarching ethical issue of all in this context. Which is what all of our advances in data science and technology in general, including algorithms and devices that can do many things better than humans can do them, are doing to the future of work. The future of opportunities for human beings to be employed and with that to our basic fabric as humans. So that will be the topic of the third and final lesson. The work that you'll do in this module, besides the reading will include participating in a couple of group discussions at the ends of the second and third lessons, and also a final case study on an article that you'll select related to one of the topics of this module. Of course, we need to discuss the virtual backgrounds first. The last one, I already said, is in Europe, which was appropriate for that lesson on facial recognition, because Europe is perhaps the most active area of the world and looking at policy and laws related to facial recognition. In fact, in the short period, from when I tape that lesson to this one, there's been more activity in Europe on exactly that topic. So the background turns out to be the Swiss city of Lucerne, which is a beautiful city on a large lake that is ringed by mountains. In English that lake is simply referred to as Lake Lucerne, in German it has the far more descriptive name of Vierwaldstatter See. Which means the lake of the four mountain states or cantons. The tower in the photo is a famous water tower that goes back many centuries and is part of the original fortifications of the city. As usual, I've switched parts of the world for the new virtual background. This one is clearly not Europe. It's actually linked a little bit to this lesson. And then it comes from a nation and a part of the world that is particularly interested both in the uses of data science and health care and more generally, in telemedicine. It's probably not hard to guess the country may be more difficult to guess the city. Let's see if anybody can do that, and I'll talk about that at the beginning of the next lesson. Now I'll turn to the content of this first lesson about the roles of data science and health care in the ethical issues that are associated with those roles. They're going to be three parts to this lesson. The first is a brief overview of spectrum of innovations that are underway and anticipated in health care to give you some perspective on two things. How much these are related to data science and the many ethical considerations that are linked to them. We'll look at a couple of articles that look generally at the uses of AI and particularly machine learning and health care, and discuss many of the ethical issues that are related to these will do one of them early in this lesson and one of them at the end. And in between those two we'll look at two articles that talk about efforts to collect massive databases of health records and some of the ethical issues associated with those two efforts in two different parts of the world. So let's start with the Time Magazine article entitled 12 Innovations That Will Change Healthcare and Medicine in the 20 Twenties. A couple of important high level comments, I should say first, that of course, this is just one view of what these innovations would be, and others might have a different list. But I think this is a pretty good representative list, and there are two things that you should immediately observe from this list of topics, which I'll get to in a moment. First of all, almost all of them are related to computer technology, and in each of those cases, their data science aspects to them. Some of them turn out to be overwhelmingly data science topics. Some of them make heavy use of data science, and some of them the data science is more in the background. And secondly, in virtually all of these, you can easily see ethical issues that are attached to them. So of the 12 topics that are mentioned in that article, I would consider at least nine of them to be heavily about computer technology and will list them on the screen now. So those are drone delivered medical supplies, huge health databases, stem cell cure for diabetes, more diverse global bio bank, wristband that can read your mind, cancer diagnosing AI, AI to read every science paper, 3-D virtual hearts and rehab in virtual reality. That was not a prioritized order by the way. I just was using the order in the article. So what are the connections of those topics that I just listed to data science? Well, just looking at a number of them. First of all, the huge health care database is entirely a data science topic and and one that we'll discuss in this lesson. The AI topics, which in reality AI and machine learning topics are also topics that we're about to discuss that have data science at their core. The wrist band that can read your mind is perhaps the most speculative of all of those topics. But if it were ever to come to pass, it would have to transmit and process huge amounts of data quickly. So it would be a very key data science topic, or data science would play a very key role in that topic. The stem cell topic is closely related to genetic editing, which we'll discuss in the next lesson and which couldn't even exist without really profound data science. And then there's other topics on that list, like drones or virtual reality, which aren't primarily data science topics but where data science also plays very key roles. So, that shows that these topics are all heavily related to computing and data science technology. They're also all topics that have ethical issues that are pretty clearly on the surface in fact. For some of them, it's really at the core of how we consider the use or not use of those topics. So two of them, the huge healthcare databases and cancer diagnosing AI are ones that we're going to be talking about in this lesson. The healthcare databases have not only issues in privacy, but others that we will talk about, the cancer diagnosing AI has major issues tied to accuracy, the trust and the trade off between what the AI does and what the humans do. There are other topics on that list that we won't be discussing but also have clear ethical issues. As an example, I'll just take one. The drone delivered medical supplies is not on the surface in ethics issue, but it is if the wrong person gets your medications. So there's clear security and privacy issues that are part of that topic. So let's get right into the consideration of these issues. We'll start with the article AI and medicine raises legal and ethical concerns as it's a nice general look at some of these issues in a good way to get started. I should note that this article is from a forum where the intent is really to stir up conversation more than it is to be carefully researched journalism. So you can take the statements with a grain of salt, but we're using it to inspire discussion and thinking and it's good for that purpose. A few key observations from this article, it immediately refers to the uses of AI and radiology, which is a topic that we'll see again later in this lesson and the balance between the roles of the AI and the doctor. As an aside, I'm not sure if you are aware of this, but radiology is one of the areas of medicine where AI has had some of the earliest and greatest uses because it's proven to be very effective, and this raises several ethical issues that are referred to in this article. One of them is, does the doctor still need to be in the loop? And if so, how much? The second one is, how much does the doctor need to understand about AI? That's the point that will come back to later as well. And another one is, are we putting radiologists out of work and that's just a teeny teaser or aspect of the future of work topic that would be doing later in this module. In another direction, this article makes very interesting points about the predictions of your physical and mental health that could be made from various data and how those could be used. I should mention that this is not just stuff that comes from genetic data. AI and machine learning is perfectly capable of making predictions just based on patterns that people display. So one of the points that the article makes then is that the people who do this sort of thing are not subject to the same regulations and how they can share their information as healthcare organizations are. This is a point that's going to come up in various flavors throughout this module. That the legal as well as ethical restrictions on how healthcare data can be shared by different organizations, are probably just not up to date with what's going on in the world. And this is an issue that needs to be addressed. Perhaps an even more profound one is who do these predictions get shared with? As an example, if you're applying for a job, should your employer see predictions about health care issues that you might have in a year or five years or 10 years. What if you're applying for a lease from the landlord? Or for that matter, what if you're thinking of getting married? Should your potential spouse have access to those issues? And finally for that matter, should you be able to control whether and how you find out about predictions about your own health? These are all important ethical issues that are just now arising and have to be dealt with. A different issue that this article brings up, which is getting us back more to the legal and ethical issues, is that in the United States, the Americans with Disabilities Act, ADA, doesn't prohibit discrimination based on future medical problems. There is a law that says that you can't discriminate based on genetic data, but nothing about other data. So again, the legal and ethical frameworks simply haven't caught up yet to the times. Another point that this article makes is that AI predictions can have errors. Yeah, they can. Humans can make errors too. We've seen exactly that discussion in perhaps the more mundane aspect of hiring, where we talked about how humans can be biased and how algorithms can be biased. So the issue is not only where there can be errors, but the issue is the balance between what the AI does and what the human does. That's more my comment. That's not a comment that was made in the article. The article also mentions the human sight, that if this is just AI that we're losing, for instance, the empathy that you would hope to have from a medical professional if someone had to relate to you concerning results, I'll point out that even that one has two sides. If you're fortunate to be in a situation where you've got good medical care and good medical providers, then sure you want that human empathy when it's useful. On the other hand, if AI is allowing you to get health care when otherwise you couldn't get it at all, which is true in various parts of the world, then it's a very different situation. So I hope this first article has simply given you a sense of the breath of some of the issues of the use of data science in healthcare and medicine and that there are quite a number of ethical issues associated with those uses. Now you might rightfully react and say yes, but a lot of those ethical issues are beyond our scope as data scientists. In fact, many of them are legal and policy issues as well. So what can we do as data scientists? I'd suggest that there are two high level things. The first is to make sure that as you develop and deploy methods that there is thought and conversation about the ethical issues and that the places where they should be addressed, they are addressed. And the second is a theme that has come up repeatedly in this course, that we all have an obligation to educate the public about the methods that we're creating and their ethical implications.
