 Hello, everyone, and welcome to the 2.30 p.m. session in the research and education track. As a reminder to our in-world and web audience, you can view the full conference schedule at conference.opensimulator.org and tweet your questions or comments to atopensim.cc with the hashtag OSCC14. This hour, we're happy to introduce a terrific session called Studying Economics in the Virtual World from the lab to the class. Our speaker today is Kevin McCabe. Kevin uses experiments to study the behavior of economic systems that range from the neural systems of the brain to the economic systems of virtual worlds. He interprets economic systems broadly to include the social, political, and economic interactions that lead to outcomes that affect our world. Since 2006, he has been developing economic experiments and data processing pipelines on virtual world platforms. Welcome, everyone. Let's begin the session. OK, thank you, everyone, for showing up. Today, I want to talk about the work we're doing in virtual worlds to do research in economics and to basically study, basically be able to study. So what we'll do is start with just a little background on me. Basically, I'm an experimental economist who studies economics from the viewpoint of the laboratory. That is, how can we set up and design experiments to test economic theories and to build economic systems? My original research was on economic systems design, which is how do you build auctions, systems to distribute goods and services? How do you build contracting systems, property rights systems, and so on? I did that for about the first 10 years of my life. And then the economic systems design approach views the institution that people engage in as a computational device that takes the messages that people send it and turn it into outcomes that they're interested in. As we moved on from economic systems design, I began to realize that institutions were being governed by people and that it became important to understand the personal exchange that people were engaged in. And so we moved much more into the question of, how do people trust each other? How do they negotiate with each other? How do they decide on rules of contact that jointly affect each other, and so on? That led us to the study of neuroeconomics, which is, how does the brain make these decisions? That is, how do we engage in social interaction? And how do we make economic decisions? As I was doing the neuroeconomics research, it began to dawn on me that basically the brain is designed to interact in a fluid world. And when we do an experiment, we often for control and for signal to noise considerations, we end up creating really simple designs. And of course, if you get a simple enough design, you're going to end up getting a theory that will explain your data with enough work. And the question is, does that theory work in the real world? Well, about 10 years ago, I had the chance to meet up with John Lester, who is giving a talk at Harvard, and we got to talking about virtual worlds. And it became obvious to me that it would be interesting to know if the theories that we were studying and getting evidence for in controlled laboratory experiments would be robust to more general social and economic interactions. And it was at that point that the virtual world became a place to consider as an experimental platform. So today, we've done eight research experiments, and it represents a major portion of our research program. But I think I'm going to talk a little bit about one example research design. But I think what to me is interesting is we've been able to leverage the research so that we can start doing demonstration experiments in OpenSim. And these demonstration experiments allow us to take our research back into the classroom. This is something that has been going on in experimental economics for quite some time now. But I think what's new in this bringing the research into the classroom is that not only can students participate in experiments, but they can also build their own economic systems. And this kind of shared social learning is really been what I think is an amazing part of the OpenSim research and teaching we do. So let me talk a little bit about one experiment that we've already done. This one was actually run in Second Life. Since then, we've moved completely to OpenSimulator. But it was our first experiment, an attempt to study collective dilemma problems in virtual worlds. So in a virtual world or in an economic setting, a typical economic problem is what Eleanor Ostrom calls the commons problem. The commons problem is a resource under common control. And because it's under common control, it can often result in individual free writing behavior. Now what do we mean by free writing? Well, it's basically when individuals don't do their fair share of the work. So Ostrom got the Nobel Prize in economics, actually, for studying these systems in the field and the laboratory. And there's been a lot of research done. And the question she asked is, even though this is a dilemma problem, similar to Kay's work on the prisoner's dilemma game that maybe a lot of people have seen already, this is a problem of self or group governance. How do groups govern each other? How do we maintain self-control and not engage in free writing behaviors so we can cooperate? So we decided to build a region that we call Hurricane Island. Now on Hurricane Island, we have 12 people living in houses around the island. The island is subject to periodic storms. So when a storm comes through, it can damage your house. Now why do you care about that? Well, if you're in your house, you can earn income. And the more damaged your house is, the less income you'll earn. So the amount of earnings you make is proportional to the amount of damage your house has. Now you can defend your house from storms locally at your own house. And you can repair your house after it's been damaged. But those are both costly exercises and take time and money. So alternatively, we set up these, if you look on the right here, these self-defense stations or island defense stations, where basically people could go to the island defense station. And if they were there, they could protect the whole island from the storm. So they were offering this public good as opposed to just trying to defend their own house. Now, if we look at the next slide, what we see is a character of the island with eight houses on it that have been, and you'll see that there's two defense stations here, one at N1 called North One and one at N3. Those defense stations only protect people below the defense station. So if you're defending at N1, you protect everybody. But if you're defending at N3, you are only defending purple, red, blue, and green. So the implication of that is that the people below N3 have a threat point. They could say, look, we're only going to defend at N3. And if we do that, if you're a yellow, brown, orange, or teal, you're going to have to protect us at N1. So if they protect us at N1, though, we don't have to protect at N3. So there's kind of an interesting game theory equilibrium here where basically we would predict that brown, yellow, teal, and orange will have to do the defending while purple, red, blue, and green will not. Now what we find out, which is interesting, is that if we make the environment not very complex, we can basically show that people don't follow the game theoretic prescription. They learn to take turns defending at N1, and everybody takes a turn, and they all cooperate. But if we add complexity to the situation, what we find is that the game theory becomes more and more predictive. So this was the first example we had of how virtual worlds give us an experimental platform where we could basically see that the theory works in more complex environments. And so if we had just always been studying this in a very simple environment, we would have missed this observation. I just thought it would be interesting to show people what a chat snippet looks like for one of the less successful, more complex environments. This is an environment where storms can come from both the north and the east, and they don't know which direction the storm's coming from. And they have to coordinate who's going to defend northern storms and who's going to defend eastern storms, et cetera. What we find is that successful groups in the language are showing very successful coordination strategies. That is, they work out and agree beforehand to who's going to cover when. And then there's usually somebody playing the coordinator rule who coordinates this interaction. But if they fail, what happens is they don't agree to the plan. And this was the case with this group. And the storm came in, and the people that were supposed to defend weren't defending. You'll notice the conversation there by Green Gartner saying, look, maybe we should go to our sub-game perfect solution. Maybe we should basically just defend it, N3 and E3, and not try to cooperate as much. At the same time, they're figuring out who the culprit is. So if you look on this next slide, they've gotten to the second storm, and all of a sudden, they've realized that it's Orange and Teal that were supposed to be a team that we're working on in that defense station. It turned out Teal showed up, but Orange never did. And through the conversation, they're starting to identify who the culprit is. Later on, this will turn into verbal punishment, and then later on into more ostrization of Orange because they failed to do their job. Now, what I think I want to do is turn now to how we've been teaching in Second Life. So this is our IFRE summer workshop program. This is for high school students. So we bring in 16 students, put them into teams of four, and have them work for a full week on building a virtual world experiment. Before they come in for eight weeks prior, we have summer interns who are undergraduates from around the country who come in and spend an eight week summer internship program with us. The interns learn how to do research in virtual worlds, and then later on act as mentors in the summer school program. The workshop is funded by IFRE, which is the International Foundation for Research and Experimental Economics. And the guy who started IFRE was Vernon Smith, who won the Nobel Prize in 2002 for his work in experimental economics. And he's been a big supporter of the open sim education efforts in economics. So far, we've been able to run the workshop for four years. We've had a total of 72 students and 16 teams work on projects. We basically emphasize learning by doing the way it works a typical day. The students start out by being in an experiment that illustrates an economic concept. They then get together with each other and myself and discuss the experiment and the economics behind the experiment. In the afternoon, they go out and design and build an experiment in open sim. And lately, the last year, we've run the open sim in Kitely. So students build an experiment. They actually, on the last day, run the experiment with family and friends. Each day, they work on presentations that they're going to make at the end of the five days. And they write up a paper explaining what their experiments about and what economic concept they're studying. This is the 2014 group. Their local area high schools from the northern Virginia area, we usually get representatives from about 12 high schools. We require that they've already taken AP computer science. And before they're allowed to get into the workshop, they have to complete about an eight-hour self-paced instructional tour through our teaching islands on Kitely where they learn basic programming in open sim practices and building practices. So these are highly qualified students, on average. What I thought I would do is just take you through a few slides showing you their typical day. So in the morning, they're participating in an experiment. So we've got 16 subjects running through an experiment at this stage. It takes them about an hour and a half to read instructions and then do the experiment. When their experiment's over, we're able to stream the data out of Kitely to a server. And then we basically use R to create graphical representations of their data. Once we have their data available, then we can talk to them about what the experiment was about, what they were doing in their experiment, and get them to tell us what their strategy was, what they were learning during the experiment, and to talk to each other about, well, how could you extend this to study other things, and so on? So the discussion lasts about an hour. Now, after lunch, they're going to learn some more advanced programming in LSL. At the front of my room, the room there is my graduate student, Peter Twig, who's been doing this with us all four years. And it's just about ready to go on the market. So I'm going to miss him dearly. But anyway, he's in charge of the instructional component, getting them up to speed. The mentors are in the room with them and are there to help them if they have questions, but not to help them actually complete their project. Now, once they get their project, the students start planning out on paper what they're going to do in the virtual world. And then in the afternoons, they spend all of their time building, basically. They do have a short meeting with me to talk about how things are coming along and stuff like that. But they're mostly in the room, and they're mostly working as teams to build their experiment. And here, if you go into that room, you just see this buzz as they're talking about what they're doing and programming in world and building the objects they need. And to me, it's totally amazing because this is the island we give them to start with. It's literally a blank slate. And what you see is the next few slides, I'm just going to show you one of the teams builds in the last year, last summer. So here's one of the students is building an object that's going to end up being the master object for coordinating the message passing between a number of objects in the experiment. And they decided to make the master object the sun. So they're working on the sun right now, getting code in, and so on. At the same time, another student is building some of the interaction elements for the experiment. So there's going to be objects that people have to touch and interact with in order to make decisions in the experiment. And they're just starting to build the basic objects. We find that on the scripting side, students kind of break down into who's most comfortable with scripting. So usually, there's about two people on each team that are doing most of the scripting, although everybody does a little bit of scripting, but usually, two people do most of the scripting. We usually get one person who is kind of just like texturing the island, likes to see visual thinker and likes to think about what the island is going to look like. And they start building. And they get help from other people as they need it. Now here on this slide, you can see their experiment is starting to take shape. And they're starting to get objects built on it. They've got pieces terraformed. Usually, right around now, the other teams start visiting each other's regions and commenting on what's going on and asking questions. So it's kind of fun to watch that interaction take place, too. Now here's the final version of their experiment. So it turned out to be a public goods game. And they've got private and public resources that people can decide to invest in. And this one completely finished. They were able to run their parents through it and basically show them the public goods problem running in open sim life. And then in front of their parents and friends, they do a final presentation explaining both the technical side of their project and the economics of their project. They haven't collected much data yet, but they usually have a little bit of data to show and so on. So we've got funding for another year, so we'll be doing this project for at least one more year. We've been getting National Science Foundation funding for the research. We continue to go strong there. I think in closing, before I finish, to me, one of the most important things that's been going on here is the research is funding the education. That is the building, the tools we build, the environments we build, and so on are paid for by grants. And that allows us to invest the time and money in it. Then the teaching benefits, because we're able to take the research experiments and then reintroduce them as demonstration experiments. So I'm going to stop there and say thank you and check out local chat to see if you have any questions. Somebody asks, is it true that macroeconomics is dead? Interestingly enough, macroeconomics will never die, but there's an economic version of macro called the Microeconomic Foundations of Macro, which is what most macroeconomists do. In fact, as far as I can tell, the large game designers are starting to realize that they need that kind of macroeconomic thinking in their games. Somebody asked, do you sometimes need outside participants in your studies? So far we haven't. We've run largely in-house. When we do teaching, it's a kind of a blended environment classroom, but we're in the classroom. When we do research, we still do it in our laboratory, largely because I'm worried about the experimenter control issues. But I think after the work we've done that we're starting to get comfortable with the idea that we could, in fact, recruit people from Mechanical Turk or Facebook to be in our Kytley experiments. And I think that would be a really interesting direction to go. Another question I got was, do you get a sense that the students will continue to use virtual environments in their future work, both academic and non-academic? That's a good question. It's a little hard to tell. We do a exposed questionnaire with the high school students, asking them how interesting they found it. They, across the board, give the experience an excellent. And we do get about 25% that say, this is something they would be interested in pursuing, if, in fact, they could find a supportive place and so on that's doing it. So I do think so. We've had three high school students now come back and be summer scholars. So come back for the summer to work with us. And our summer scholars and the high school students both tend to end up at really good undergraduate and graduate schools. And they stay in touch. I don't think they get that much opportunity to do virtual worlds in their current research. It looks like that may be it for questions. Well, thank you, Kevin, for a terrific presentation. This session closes out the research and education track presentations. And we're at the end of the second annual Open Simulator Community Conference 2014. Following the conference program, all staff and volunteers are invited to the stage at the keynote one region for pictures and a celebration. And the audience is welcome to join the festivities on your assigned keynote region. In addition, there are still three social events after the conference to keep the excitement going. At 3.30 PM Pacific, Steven Zootfly will be hosting an Educators Birds of a Feather meeting on the Avocon grid to discuss how to encourage more educators to use Open Simulator. And at 5.00 PM, Lucina Wisdomseeker hosts a continuation of the Quest for the Galaxy Language session on the Second Life grid in the Inspiration Island region. Finally, the last social event of the conference is at 7.00 PM Pacific on the Pirates Atoll grid. Dance on the Sunset Beach at Pirates Atoll to the ethereal sounds of dream pop and indie beat. A great time to unwind and socialize after the conference. A tremendous thank you to all of our speakers, sponsors, crowd funders, volunteers, and the hundreds of attendees who braved a host of technical challenges to make this conference a success. Have a wonderful evening, and we look forward to seeing you next year at the Open Simulator Community Conference 2015.