 Hi, it's Jerry Mikulski again, Story Threading Unfinished 21. This session between Chris and Deborah is extremely geeky. Now, I'm a geek, but this is kind of above my pay grade. In fact, about eight minutes into the session, I had to stop there, stop playing and go do a bunch of research and move things around in my brain so that I could, in fact, Story Thread a little bit better. But I'm giving it my best swing here. One thing is, it's certainly not outside my awareness that I am a white guy. And the opening conversation here in the session is about algorithmic bias and what people of color notice from technology that people like me didn't notice they had baked right into the technology. And this is a super interesting topic under the big umbrella of the ethics of big data, the ethics of artificial intelligence, all of those sorts of things. So with that as background, let's kind of dive in. So the session is titled, The House is on Fire. And I think that's meant to echo Greta Thunberg talking about the members of the European Parliament to panic like the house is on fire, something that they're not really doing yet. And I haven't been following COP26 very much, which is happening right this minute, just starting. But I hope that they do because the house is in fact on fire. So that was really interesting. I put this session under ethics of big data and some of the other topics I just mentioned. And that comes down to things like algorithmic bias. And so for instance, here is the Algorithmic Justice League, which Deb Raji is a founder of and part of. And so we'll find our way back here to this session, The House is on Fire. And then up to Chris Wiggins, a data scientist who has written for the New York Times is on the faculty at Columbia and has done a bunch of interesting things as well. During this session, they talked to about a lot of academic papers. And I tried to capture most of those papers. So here's 50 years of data science. And unfortunately, I've got them collected alphabetically, but you'll notice if you watched the session. So this is by David Donahoe. And in the beginning in the introduction of this paper, they mentioned an earlier paper called The Future of Data Analysis by an early thinker named John Tukey, a statistician and mathematician who did really interesting things, including coin the term bits for binary digits. But if we go back to 50 years of data science, that's one of many papers they mentioned. This one, this one mentions the common task framework. It happens late in the session. But here's a greater or lesser statistics, a choice for future research written by J.M. Chambers, also a statistician who was part of the project R. If you've ever done statistics in the modern era, you've probably used R, he invented the programming language S. And R is its successor. He was at Bell Labs, which is basically a birth garden for all sorts of famous and useful technologies. If you're a technologist, anyone who is at the Bell Labs during the golden era is a little bit in the pantheon of gods, I'm afraid. And then they also mentioned an article by Jim Balsely. Data is not the new oil, it's the new plutonium. I think that's the article they were pointing to. They just mentioned that data is like plutonium. And I liked in his article, there was a summary that data is amazingly powerful, dangerous when it spreads, difficult to clean up and with serious consequences when improperly used. This is to countervail the very popular notion that data is the new oil, which is a popular thing these days, meant to imply that all this data science and big data is going to lead to large deposits of highly valuable stuff that companies should hoard and then meter out in some way, that that's their chief source of profits. I have a draft of a piece that I need to finish writing titled Data is the New Soil, talking about soil fertility and how, in fact, if we have sovereign, separated data where we own it and release it and maintain it and make it better, that's actually much better for the commons and for humanity. But that gets us off a little bit off the topic over here. They also mentioned a couple other papers, the critique and contribute paper, a practice based framework for improving critical data studies and data science. I had never heard of critical data studies. So I added this paper, I added that I added the authors of this of this paper, except I already had Gina Neff in my brain as a sociologist, married to Philip Howard, and the publisher of the writer of surviving the new economy and venture labor and self tracking. So all of those basically wove their way into the content here. But I want to go back to the bias question that that comes up like immediately and early and just say, Hey, this is actually a really important thing. If you've done any kind of image compression, so there's a standard test image for image processing called Lena, and it's named after Lena Söderberg. The image looks like this kind of speaks for itself. This was a playboy playmate. So Lena Söderberg was a playboy playmate back in the day. She basically becomes this image that the men inventing better image processing routines pass around kind of as an inside joke as part of trying to make this technology better. If bias and sexism and other things could be baked in more deeply, I guess they could, but this is like a not a happy start. And this is not an example that Deborah Chris mentioned in in their conversation. But it's the first thing that came to my head. I'm like, Oh, yeah, I'm familiar with this kind of stuff. They also talked about image net and Alex net and a bunch of other projects, which I've connected in here, but aren't sort of really big ways to go. But one thought and by the way, there's a little scroll bar here. So I've connected up to more and more things. One thing that I kept coming back to that I then elaborated more in my brain, which was and the critiques here are not just about Facebook, but Facebook is a particularly great example to sit and talk about. So the question is, why is Facebook dangerous? So Facebook is actually the largest country on earth. Facebook has more monthly average users than the populations of China plus India. And that's counting all the old people, all the children, all the people who might not be on social networks. So so somehow Facebook grew that large with no oversight. So the question here emerges. If you think of Facebook as a country, what changes? How do you start thinking differently about it? So first, Facebook is huge. Second, users are busy pouring deeply personal information into Facebook all the time in enormous and unbelievable quantities. It turns out that the ability to see who was at the party, what everybody's lives are like, the ability to easily like, connect, share, upload, record, all those kinds of things is incredibly addictive. And in fact, addiction was baked into the frameworks. But then probably worst of all, Facebook's business model, algorithms and policies entrench echo chambers and fuel the spread of misinformation. That's one piece about why Facebook is dangerous. That's really important today to the topic at hand. And then I found a quote, Facebook is Chernobyl. This is a quote by my friend Rika Sifri. Facebook is Chernobyl, a badly designed nuclear reactor that makes a lot of energy and money for its owner by harnessing the inherent power of connecting atoms and people to each other. But without more external oversight and regulation, we'll keep producing toxic effects. Seems like a nice summary of these various topics that we're talking about. So in the conversation, they went around to lots of different perspectives on how these systems kind of break. So Deb says, if you Google wedding, you get a really very Western representation of weddings. And then if you do it in Hindi and maybe having to do the search from India, you'll probably get a very different look and feel for what wedding means. So a lot of people, a lot of researchers just use Google searches to come up with databases thinking that if it came out of the computer, it's probably neutral or objective. And it turns out that not so objective, that wasn't really happening. There was a lot of geekiness here about how to actually try to create objectivity in data and in the algorithms, which data set you choose matters that often machine learning engineers use the data that is most easily available instead of the right data, that incentives are not aligned around rights, harms and justice. And then some question about how to go about creating remedies like reputational damage is what's happening right now. We're seeing a lot of tech clash, a lot of backlash against the tech giants. But is that enough reputational damage? Their stock prices are still flying. They still own the companies. They're still directing them in different ways. And then there was a early in the in the conversation, there's this interesting topic about third party AI algorithm audits, a topic I did not have in my brain. I had a little bit about auditing. And I did have the topic of how do we keep AI from being a threat? And so I spent a lot of the conversation thinking about this question. How do we keep AI and these massive platforms from being a threat to us? And how do we make them more useful? One of my remedies for Facebook in particular is this question. What if Zuckerberg had designed Facebook for citizens instead of consumers? So Facebook is this giant addiction engine that sells off our data and stalks us and a bunch of things like that. What if instead Facebook had been created for citizens whose job is to think together about the problems of the day, to have autonomy and sovereignty over their own data, and then to participate in the public space, in the public sphere? At this point Zuckerberg has made a big deal. In fact, in fact, in the last couple of days Zuckerberg declares the metaverse as Facebook's next big initiative. They renamed the outside company as meta in order to make that point a little bigger. But their idea of the metaverse is this consumer metaverse that sucks. And if instead the metaverse were more like what I'm showing you here, a network of ideas and concepts and evidence, but not just mine. And here we're stuck because I've been feeding this brain for 24 years but I wish I were doing this collaboratively with other people. I haven't been, but I want to and I've got a passion project right now to try to do that. What if that was the metaverse we were walking into so that as part of our civic activities, we could have conversations as I've been trying to have with you story threading unfinished 21. Now, my story threads are pretty one way. I'm recording these at some point you listen to them. There isn't an easy back channel for you to talk back with me, but there could be. And the medium itself could in fact be the back channel and help us make that back channel. So all of these things kind of swirl together into the big juicy topic that Chris and Deb put in front of us in the houses on fire. And I hope this has given you a couple of new avenues to pursue. And when you see the link to this spot in the brain, you'll be able to come back and look up all of the studies they mentioned and other sorts of things. Some of the studies and posts, I'm not sure I got right. So for example, at one point, Chris says that machine learning has known sin, which is a riff on J. Robert Oppenheimer's quote. Physicists have known sin after the Trinity explosion and the bombing of Hiroshima. But and I found this piece by Ben Dixon, and I'm not sure this is the right Ben that they refer to in the session. But I'll be happy to change this and correct it to the right one. Should that feedback get to them and get back to me? For now, I'm very happy to spend this time with you and thanks for watching.