News from Linde Institute Refreshhttps://lindeinstitute.caltech.edu/news2023-10-10T19:50:00+00:00Linde Institute Refresh Stafflindeinstitute@hss.caltech.eduCopyright © 2024 California Institute of TechnologyCaltech's New Center for Science, Society, and Public Policy Hosts Research Conference on Conspiratorial Thinking2023-10-10T19:50:00+00:00Cynthia Ellerceller@caltech.eduhttps://sites.caltech.edu/newspage-index/Caltech_Hosts_Conference_on_Conspiratorial_Thinking<p data-block-key="7idkv">The <a href="https://lindeinstitute.caltech.edu/research/csspp">Center for Science, Society, and Public Policy</a> (CSSPP) at Caltech mounted its first research conference on September 14–15, 2023, to address the phenomenon of conspiratorial thinking from disciplines as diverse as English literature, political science, economics, neuroscience, and psychiatry. Mike Alvarez, co-director of CSSPP and professor of political and computational social science, sees this initiative as characteristic of what CSSPP aims to do. "We keep trying to identify topics that connect with research that Caltech faculty are doing but that also connect to broader policy conversations that are going on outside," he says.</p><p data-block-key="4kc85">Caltech's interest in conspiratorial thinking has its intellectual roots in the COVID-19 pandemic. "The pandemic was a weird time when we were all sitting at home trying to figure out how we could do our work and what we might be able to do to help," Alvarez says. "Ralph Adolphs [PhD '93 and Bren Professor of Psychology, Neuroscience, and Biology] began a study called the <a href="https://coviddynamic.caltech.edu/">COVID-Dynamic</a> Longitudinal Study that put together these really fascinating batteries of questions to ask people about how they were doing emotionally during lockdown. This dovetailed with work that I was doing with two of my graduate students on vaccine acceptance back before covid vaccines were even available."</p><p data-block-key="8d4ef">What began with more narrowly focused questions related to the pandemic became more general concerns for Caltech faculty about misinformation and conspiratorial thinking. "I began a collaboration with Ramit Debnath, a sustainability fellow at the University of Cambridge, looking at misinformation on climate change and what leads to the dissemination of misleading information, especially on social media," Alvarez explains. "Meanwhile, John O'Doherty's lab [O'Doherty is Caltech's Fletcher Jones Professor of Decision Neuroscience] moved into researching the potential psychological determinants of people who might be susceptible to conspiratorial thought—things like aversion to ambiguity or being less willing to seek uncertainty."</p><p data-block-key="708er">"As we started to poke around the literature on conspiratorial thinking, we found a lot of fantastic research on this and decided to bring some of the leading scholars in humanities, social sciences, psychology, neuroscience, and political science to Caltech, under the auspices of CSSPP, so that we could let the Caltech community see the cutting-edge research on this topic," Alvarez says. "When we reached out to people to ask if they'd be interested in participating in a conference, they were all really excited to come to Caltech for this event. So, we assembled an amazing all-star cast."</p><p data-block-key="e5fpm">The conference kicked off on September 14 with a presentation by Elise Wang, assistant professor of English at Cal State Fullerton on medieval European blood libel conspiracies. (Blood libel is a belief that Jews use the blood of Christian children for ritual purposes. It is false, of course, but such charges led to the persecution of Jewish communities in Europe, and the belief is still found in anti-Semitic literature today.) Wang set the initial terms of discussion for the conference by asserting that conspiracy is its own narrative genre with a consistent set of characteristics that have persisted over time and across a very wide variety of conspiratorial theories. Key to this genre is a certain flexibility and resilience that allows believers to fill in details, connect the dots in various ways, and talk their way around even logical contradictions.</p><p data-block-key="49g15">Wang's talk was followed by a philosophical/epistemological take on conspiracy thinking, and then presentations of new research from Alvarez's group by Yimeng Li (PhD '22), now a postdoc at Florida State University, and Debnath.</p><p data-block-key="cp15m">In the afternoon, Adam Berinsky of MIT presented on "The Root of False Beliefs," a topic he has been working on for more than a decade and which is now gathered into his just-published book from Princeton University Press, <i>Political Rumors: Why We Accept Misinformation and How to Fight It.</i> Additional talks were given by political scientists Betsy Sinclair (MS '04, PhD '07) of Washington University in St. Louis and former student of Alvarez's, and Joanne Miller of the University of Delaware, both experts in conspiratorial thinking. Sinclair spoke about her research into identification with partisan political groups and the extent to which individuals are willing to adopt false beliefs if those beliefs preserve their partisan identity. Miller reviewed the secret plots that are so often the essential backdrop of conspiracy theories, calling back to Wang's notion of the narrative structure of conspiracies to better understand when and why "not seeing is believing."</p><p data-block-key="anm9c">The second day of the conference turned toward psychological questions that probe the precise dispositions of those who favor conspiracy theories. Nadia Brashier of UC San Diego began her talk by asking if conspiracy theorists think too much or too little, concluding that the answer is both: some supply incredible detail and craft intricate conspiratorial worldviews; others seem content to take a lot on faith, endorsing a conspiratorial worldview without worrying too much about the fine points. Gordon Pennycook, a professor of psychology at Cornell University, shared his results that show that conspiracy believers are "dispositionally overconfident" and greatly overestimate the number of those who agree with their views; often, he noted, these individuals claim that more than 50 percent of people share conspiratorial beliefs that are actually confined to 5–15 percent of the population. Lisa Kluen, who moved to the Laboratory for Brain and Cognitive Health Technology at McLean Hospital after a postdoctoral fellowship at Caltech in O'Doherty's lab, presented research on the "cognitive attributes" of those who most readily champion conspiracy theories. "What we found," Kluen says, "is that individuals who subscribe to conspiracy beliefs more readily attribute outcomes to the involvement of hidden agents. Also, they seem to seek less information before making decisions, and their decision-making seems less guided by reward."</p><p data-block-key="5rve4">Perspectives from clinical psychology and psychiatry rounded out the afternoon, with speakers offering additional thoughts on the cognitive traits of conspiracy believers and explaining the ways in which conspiracy belief can be clearly distinguished from delusional thinking and from the paranoia experienced by schizophrenics.</p><p data-block-key="ea93g">The conference closed with a presentation by Dutch social psychologist Sander van der Linden of the University of Cambridge, who researches fake news and seeks to "inoculate" people against conspiracy theories through a variety of online and interpersonal games and exercises. How to work against socially destructive conspiracy theories was a through line in the conference. "How do we prevent the spread of conspiracy theories? How can we help prevent them from going down the rabbit hole? And if they have gone down these rabbit holes, are there ways we can persuade them to at least be open to new information?" are all questions Alvarez hoped to raise when designing this conference.</p><p data-block-key="5udl9">"By having an event like this, we want to highlight the kind of research that we do [at Caltech] in the social sciences, ranging from the quantitative and more observational work to the really detailed psychological neuroscience work that John O'Doherty's group does, and frame it in the context of all the other work that's going on in the area of conspiratorial thinking," Alvarez says.</p>Preserving Natural Resources through Policy2023-09-15T06:33:22.483224+00:00Cynthia Ellerceller@caltech.eduhttps://sites.caltech.edu/newspage-index/Hannah_Druckenmiller_Profile<p data-block-key="7idkv"><a href="https://www.hss.caltech.edu/people/hannah-druckenmiller">Hannah Druckenmiller</a> joins the Caltech faculty this year as an assistant professor of economics. An environmental economist, Druckenmiller focuses on governmental policies concerned with preserving natural resources: whether the policies succeed or fail, how their costs and benefits can be better conceptualized for decision-makers, and how the effects of climate change can best be ameliorated through improved regulation.</p><p data-block-key="3dou">Druckenmiller earned her PhD in economics at UC Berkeley and most recently worked for Resources for the Future (RFF), a nonprofit in Washington, D.C., focused on environmental economics and policy research.</p><p data-block-key="etkmp"><b>What environmental policies are you most interested in?</b></p><p data-block-key="4nqf7">Two of the most significant environmental regulations in the United States at the federal level are the Clean Water Act and the Clean Air Act. Although these regulations have been in place for more than 50 years, they're constantly being reassessed and reinterpreted. The scope of the Clean Water Act, for example, is repeatedly debated and redefined by the Supreme Court, presidential administrations, and state litigation. In fact, the Environmental Protection Agency has had different interpretation of which waters are regulated by the Clean Water Act under Presidents Obama, Trump, and Biden. And the change in the scope of environmental regulation is significant—we estimate that 30–40 percent of regulated waters lost federal protection when the Clean Water Act was reinterpreted by the Trump administration.</p><p data-block-key="8lfli"><b>How can the scope of the Clean Water Act change so dramatically?</b></p><p data-block-key="3t3od">The regulation is written in an ambiguous way. It's not clear exactly what is protected. The Clean Water Act protects the "Waters of the United States." This phrase clearly includes navigable waterways like the Mississippi River, and clearly excludes a small puddle in your backyard. Debates center around whether the law protects intermediate cases like isolated wetlands or ephemeral streams that flow a few days per year. In order to decide whether a water resource—like a lake or a river—is regulated, the government sends out an engineer from the Army Corps. These are case-by-case decisions, and we don't have even a ballpark figure for what percent of streams or wetlands in the United States are regulated. One of my projects is trying to use machine learning to map this. Basically, we're creating an algorithm trained on all these case-by-case decisions that have been made historically to decide what the possibility is of any particular resource being regulated. We're trying to understand what was originally regulated, how that scope changes when you get a narrower interpretation of what's protected, and then what the downstream consequences are.</p><p data-block-key="53n6f"><b>What types of downstream consequences have you looked at?</b></p><p data-block-key="6bd4s">I'm interested in how ecosystem services change when you alter environmental protections. Under the Trump administration's interpretation of the "Waters of the United States," a significant share of wetlands was deregulated. I wanted to understand how this would impact the flood protection services that wetlands provide. We found that converting 1 hectare of wetlands (roughly the size of 2.5 football fields) to built-up land increases property damages from flooding by more than $12,000 per year. These costs are rarely borne by the developer who converted the wetlands to another use—they are mostly borne by downstream community members who no longer benefit from the wetlands' ability to trap and slowly release water that would otherwise cause flooding.</p><p data-block-key="9eff5"><b>How do you look at effects of climate change?</b></p><p data-block-key="6sh1l">I recently did a project on how to disincentivize development in areas that are most likely to be affected by climate change. We already have high levels of development in places that are at risk from coastal flooding and sea level rise. One big question is, how can we manage retreat from those places? A second question is how we can stop future development in those risky places.</p><p data-block-key="1ecg5">A longstanding hypothesis in economics is that, intentionally or not, the government subsidizes development in environmentally fragile places by providing risk management tools like insurance, or funding for infrastructure such as roads, water lines, and sewage systems, and by providing disaster assistance when problems arise. When there's a hurricane and the federal government comes in and gives the locality or individuals money to recover, that's indirectly subsidizing them for living there in the first place.</p><p data-block-key="7c1m6">So our question was, if you got rid of government subsidies in these fragile places, could that alone prevent development there? Or are these such desirable places—they are along beautiful coastlines—that governmental subsidies, or their lack, wouldn't make a significant dent in the amount of development you see? We looked at this program from the 1980s called the Coastal Barrier Resources System, which was a policy that removed these types of subsidies for development in designated areas along the Atlantic and Gulf coasts. We wanted to see the long-term effect of the policy. Forty years later, do we see much lower levels of development in these places?</p><p data-block-key="b694p">The challenge to studying the Coastal Barrier Resources System is that designated areas were not randomly assigned along the coast. They were intentionally selected by land-use planners because they were considered risky or because the land was thought to have environmental value. We were able to find controls—similar places not affected by the Coastal Barrier Resources System program—by running a machine-learning procedure intended to mimic the process by which land-use planners designated target areas in the 1980s. This allowed us to find places that were statistically indistinguishable from treated areas in the 1980s, but which did not enter the program. Then we compared outcomes in the treatment and the control areas. What we found is that this policy did a lot to reduce development. On average, the protected areas had 85 percent lower development levels.</p><p data-block-key="6osuf">What I found most interesting is that we also found evidence that the areas without federal incentives for development created spillover benefits to surrounding communities. By conserving natural land, ensuring that it wasn't converted into built-up area, we saw flood-protection benefits in the areas surrounding wetlands. We even saw higher property values in those areas because they're next to a natural amenity: namely, these pristine coastal areas.</p><p data-block-key="2sfu4">That was interesting to me as an economist. It implies that federal programs to reduce disaster exposure don't need to conflict with local interests in maintaining the tax base. A lot of localities don't want these protected areas in their jurisdiction because they think that if they kill development it will lower their property tax revenue. But what we found is that restricting development in these coastal areas has no net effect on property tax revenue across the counties they're located in. There are higher property values in surrounding areas and lower property values in the designated protected area, so most often these two effects cancel each other out.</p><p data-block-key="3veoe"><b>Do you think removing subsidies for development in wildfire prone areas could work the same way?</b></p><p data-block-key="bgu6p">My coauthors at Resources for the Future and I are hoping to work more on this. Some of my collaborators do a lot of work on wildfires, and we're really interested to see if you could institute a similar policy in the wildland-urban interface. There are differences between coastal areas and wildland-urban interfaces, so it's not obvious that the same policy instrument would work, but that's something we're interested in looking at.</p><p data-block-key="373vf"><b>How did you become interested in environmental policy?</b></p><p data-block-key="7g24t">I grew up in New York, and we spent a lot of time by the ocean. I always wanted to do something related to marine biology. When I was in high school, I attended a program in the Bahamas called The Island School. You go there for a semester, and all of your academics are place-based: your science class is marine biology, your math class is focused on celestial navigation, and in the humanities all the literature we read was written by authors from the Caribbean.</p><p data-block-key="8gqi5">The Island School is an amazing place, and it really solidified my desire to study marine biology. So when I went to Stanford, I chose an interdisciplinary major in environmental science with a focus on oceans. The major required several classes in economics since economics is central to understanding environmental policy. I took my first economics class and fell in love with it as a framework for understanding why the environment is regulated, when regulation is successful, and when it's not.</p><p data-block-key="86ai8"><b>What are you looking forward to at Caltech?</b></p><p data-block-key="eaf23">The thing I'm most excited about is that people here really think differently. It seems like you're not constrained to your disciplinary box as much as you might be at other institutions. There's access at Caltech to world-class scientists who are interested in collaborating with social scientists. I hear ideas here that I haven't heard before.</p><p data-block-key="8pjqh"><b>Will you be participating in the Center for Science, Society, and Public Policy (CSSPP)?</b></p><p data-block-key="10u6q">Yes, the <a href="https://lindeinstitute.caltech.edu/research/csspp">CSSPP</a> was a big draw for me at Caltech. The center supports engagement between science and policy, with priority research areas including climate change and sustainability and artificial intelligence (AI). I'm increasingly interested in how we can use AI systems to improve the monitoring and enforcement of environmental regulations.</p><p data-block-key="5l94l">For example, the Clean Air Act is enforced based on a system of 900 ground-based monitors across the United States. That's less than one per county, so the law is not binding in many places that have noncompliant air quality where there simply isn't a monitor. I have a new project that asks whether there's a path to using satellite data directly for enforcement of the Clean Air Act. And if not, is there a path to using satellite data to at least inform the placement of new monitors?</p><p data-block-key="8t9il"><b>So there were no satellite monitors when the Clean Air Act was first written?</b></p><p data-block-key="eegm6">Correct, but it's more complicated than that. Satellites cannot directly measure chemicals or pollutants as you would ordinarily think of them. They measure things that are correlated with pollution, like aerosol optical depth. That's basically just a fancy word for how hazy the atmosphere is. Then you can use a statistical model to convert that into an estimate of a specific pollutant, like particulate matter. It's hard to enforce a regulation based on something this uncertain, especially when the regulation is very costly. I mean, if you're noncompliant with the Clean Air Act, your county has to curb pollution-generating activities like industrial production or traffic. This basically means less economic activity in your region.</p><p data-block-key="c8v15"><b>What are you going to be teaching at Caltech?</b></p><p data-block-key="5l8ds">I'm starting with environmental economics at the undergraduate level this year. I've met some of the PhD students in economics since I arrived. It seems like the students here are really bright and curious. I also hope to work with postdocs coming through CSSPP.</p>Experimental Economics in Theory and Practice2023-07-24T03:06:39.694113+00:00Cynthia Ellerceller@caltech.eduhttps://sites.caltech.edu/newspage-index/Experimental-Economics-Theory-and-Practice<p data-block-key="7idkv">The social sciences have to face a notoriously difficult challenge that begins with their very name. Just what sort of science are these "social" sciences? Can they really help us study and understand human society in the same way the natural sciences promise to improve our understanding of the natural world? Certainly it seems that they should. Why should we not be able to observe, analyze, and even quantify human behavior if we already feel comfortable doing the same for volcanoes and rivers, marigolds and fruit flies?</p><p data-block-key="qodq">In practice, things have often been much dicier. From the Book of Genesis on, we humans have demonstrated a somewhat (entirely?) overblown assessment of our own singularity perched atop the natural world. And yet, the quest to find experimentally verifiable answers about human behavior continues. How could it not? If there is one thing that would be extremely profitable for us to understand, it would be ourselves. It is this quest to construct hypotheses about human behavior, test them, and share the results that motivates Caltech's summer program in theory-driven experimental economics, now in its second year.</p><p data-block-key="2vhsh">"When I was considering moving to Caltech," Professor of Economics Charlie Sprenger, who is also executive officer for the social sciences, explains, "[Professor of Economics] Marina Agranov and I brainstormed some ideas for creating educational opportunities for graduate students who are interested in the intersection between structured theories of choice and experimental tests. It was Marina's idea to develop a visiting student summer program on the topic as a way to help students build their networks, get feedback on their own projects, and inspire each other."</p><p data-block-key="drpgp">The summer program reaches out to graduate students and other interested scholars in experimental economics via professors and researchers in the field. This year's summer program brought students from UC Santa Barbara, University of Michigan, UC San Diego, Ohio State University, Columbia University, Princeton University, and UC Berkeley to the Caltech campus. "This is the first opportunity we've had where we can meet grad students from the same department at different universities," said Jack Adeney, a doctoral student in the social sciences who came to study at Caltech by way of NYU Abu Dhabi and the University of Cambridge, and who attended the program.</p><p data-block-key="efbqm">This year's weeklong intensive ran from June 20–24 in Dabney Lounge. It began with lectures and presentations given by Caltech faculty, including Sprenger, Agranov, Antonio Rangel, Kirby Nielsen, and Thomas Palfrey, and Doug Bernheim of Stanford University. Students learned how to formulate theories about human decision-making and then construct controlled economic environments to test these in a way that reasonably matches real-world situations. They were also exposed to new data-collection tools such as tracking visual fixations, neural activity, and decision times that enhance standard laboratory experiments on subjects' economic choices.</p><p data-block-key="fiplk">The last two days of the summer program gave students the opportunity to present their own research and get feedback from professors and fellow students alike. This, as much as anything else, is the motivation for the program.</p><p data-block-key="3j1bs">"Marina and I are quite blunt on this point," says Sprenger. "We want students to be inspired to conduct theory-driven experimental research, write wonderful job market papers, and receive offers for postdoctoral positions and assistant professorships that will inspire the next generation to follow their path."</p>Algorithms: The Cause of and Solution to Our Modern Problems? A Conversation with Eric Mazumdar2023-05-19T21:06:15.136254+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/algorithms-the-cause-of-and-solution-to-our-modern-problems-a-conversation-with-eric-mazumdar<p data-block-key="nit18">Many academics choose to study one thing and one thing only. Eric Mazumdar is not one of those academics. As an assistant professor of computing and mathematical sciences and economics at Caltech, Mazumdar uses tools and ideas from economics to understand emerging problems in machine learning. Some of those problems have grown immensely more complicated with the development of algorithms that steer our choices about the clothes we wear, the food we eat, the movies we watch, the information we digest online, or even whether to grant someone bail or give them a home loan. But algorithms are a double-edged sword; for all the complexity they add to our lives, their behaviors can be easier to model and understand than human behaviors.</p><p data-block-key="4bg8n">We recently spoke with Mazumdar, who came to Caltech in 2021 after receiving his doctorate in electrical engineering and computer science from UC Berkeley that same year.</p><p data-block-key="85l3b"><b>You study economics. You study computer science. You got an undergraduate degree in electrical engineering. How did you transition from electrical engineering to economics? Is there a through line there?</b></p><p data-block-key="8rrc2">My journey was actually a little bit more complicated. When I started on my bachelor's, I was actually interested in bioengineering for a while, and then I realized more and more that the part I liked most about the research I was doing was the math and the modeling of the dynamics of biological systems. That took me from bioengineering to electrical engineering and computer science.</p><p data-block-key="3fk8d">Then I started my PhD working in a field known as control and <a href="https://en.wikipedia.org/wiki/Dynamical_system">dynamical systems</a>. It's a relatively old field, but, at the time, I was thinking that the field's tools hadn't been applied yet to a really pressing area—this emerging area of machine learning in society, where the dynamics of the algorithms have real impacts on people's lives. The decisions that these machine learning algorithms make can have consequential impacts, but the dynamics are really complex because you have algorithms interacting with people.</p><p data-block-key="b7fr9">To me, the through line is thinking about the underlying dynamics in all of these systems.</p><p data-block-key="2nk7h"><b>Can you define dynamics for me? Is there a field-specific definition of dynamics?</b></p><p data-block-key="771kq">Dynamics are really just any process that evolves over time. A lot of our day-to-day interaction with algorithms happens in dynamic ways and at different scales. For example, on a small scale, you have individuals interacting with Netflix or TikTok, and their preferences change, and content changes over time. But then you can have dynamics on larger scales, like traffic dynamics. Anyone who commutes knows that traffic patterns change and evolve throughout the day. In the morning, traffic tends to move toward urban areas and in the afternoon, traffic moves toward residential areas. That's a time-varying process. Dynamics are just any time-varying process.</p><p data-block-key="8tbmc"><b>I would imagine that these kinds of dynamics can get rather complicated when you talk about something like Netflix, because it's not</b> <b><i>just</i></b><b> people's preferences. The algorithm is also adapting to their preferences, so maybe someone starts out with a preference but then the algorithm learns from them and it feeds back to them, and it might change their preference.</b></p><p data-block-key="215us">You use exactly the right word there. They get complicated because of these feedback loops between algorithms and people. That interaction, as it plays out over time, can get extremely complicated and give rise to behaviors that we wouldn't want to happen or things that are surprising.</p><p data-block-key="6h8fq">Examples of this are pretty common. One of the ones that we were looking at recently was what happens if multiple firms are trying to sell the same product on Amazon, and they're all trying to set their prices. If they use the wrong pricing algorithms, the fact that they're interacting with each other can cause prices to go crazy and spike unnaturally. It shows that the feedback loops in these systems are super important and are not very well understood yet.</p><p data-block-key="508s7"><b>What is it about economics that interests you?</b></p><p data-block-key="6dl2">A large class of algorithms is being deployed into everyday life for decision making that has an impact on people's lives. If these algorithms are making consequential decisions for people, we have to take into account people's objectives and people's preferences and how people reason. Those are ideas and tools that I think economics has historically dealt with. Instead of reinventing the wheel, I think that there's a huge amount of benefit we can get from integrating ideas in economics into algorithm design and machine learning. That's what I work on.</p><p data-block-key="4uh9i"><b>Machine learning and artificial intelligence (AI) have been in the news a lot lately. Is this a challenging field to be in because of how quickly things are changing?</b></p><p data-block-key="4t2g5">For sure. On the one hand, things have changed very fast even since I started grad school, which was 2015; the recent progress in deep learning and large language models like GPT-3 and 4, in particular, has been really impressive. However, on the other hand, I think that things have moved so fast that a lot of important questions have gone unaddressed. I think there's a huge amount of work to be done at the interface of economics and machine learning in terms of understanding the fairness, bias, and manipulability of algorithmic decision making. As an example, in recent work, we trained a simple algorithm on top of a large language model (like GPT) to predict skills and jobs from resumes, and we found that by adding a simple signal in only 0.1 percent of the training data, we could consistently get a resume scored highly for skills that it actually wasn't associated with. The fact that these models, which can be so impressive, are also so easy to manipulate is really telling.</p><p data-block-key="3ghca">An overly simplistic way of thinking about why we've missed out on understanding these problems is that machine learning has progressed extremely rapidly in tasks that are similar to pattern recognition, but in decision making, I think it's moved a little bit slower. To me, that's because decision making, fundamentally, has to deal with uncertainty, planning through dynamics, and the behaviors of other agents. Those are problems that people are still thinking about in economics and that we don't have huge amounts of data for. Adding the complexity of algorithms doesn't really make the problem any easier.</p><p data-block-key="ake3j"><b>Economics is an old field, but machine learning is relatively new. What does it add to the mix when you bring these machine learning tools into economics?</b></p><p data-block-key="dcllb">I think we understand machine learning algorithms a lot better than we do people, because, from a certain standpoint, we can control what algorithms do when we use them in game-theoretic settings. We have a lot more control over the end behaviors that can happen. That tends to actually simplify some of the problems in the end.</p><p data-block-key="5govm"><b>How do these algorithms simplify economics problems for you? Are they able to recognize patterns that are difficult for humans to recognize? Is it that they are able to work through a big pile of data more easily?</b></p><p data-block-key="fuk6g">They make it easier because we have more math to describe them. There are a lot of mathematical models for how people make decisions in economics, <a href="https://en.wikipedia.org/wiki/Behavioral_economics">behavioral economics</a>, and <a href="https://en.wikipedia.org/wiki/Mathematical_psychology">mathematical psychology</a>, but if we're coding an algorithm, we know exactly what's happening. That kind of certainty allows us to simplify and understand the dynamics we study.</p><p data-block-key="2mafn"><b>What brought you to Caltech?</b></p><p data-block-key="48jg8">One of the things about Caltech that's exciting to me is how interdisciplinary it is. I'm jointly appointed in the Division of the Humanities and Social Sciences and CMS [<a href="https://www.cms.caltech.edu/">the Computing + Mathematical Sciences department</a>], but I also have an affiliation with the <a href="https://www.cms.caltech.edu/academics/grad/grad_cds">Control and Dynamical Systems</a> program. The opportunity to work across disciplines and talk to people in different areas is a very rare thing in academia, and I think that's something that Caltech does extremely well because of the small size. I can interact with the economics faculty, with the experimental economists, with the control folks, with the robotics folks, or with people who work on networks.</p><p data-block-key="7eqrr">As an example, we're doing a research project right now with one of my students that is exploring how the use of algorithms in social media impacts polarization in society. We're trying to model the interaction of an algorithm over a large population of people. Modeling large populations of agents is something that's commonly done using partial differential equations (PDEs), and Caltech has a really strong core of theorists in applied math who work on understanding PDEs. Just by talking to them, we have gotten several ideas on how to model these populations of people when they interact with learning algorithms. We used that to come up with a model that tells us when and how polarization can emerge.</p><p data-block-key="brel2"><b>OK, my last question: Where is your favorite place to travel to, and why?</b></p><p data-block-key="d1jr9">My mom is from France and my father is from India, so I've had to go to India and France quite a lot, but the one trip that has stayed with me was a trip to Egypt when I was pretty young. We were able to start from the bottom of the Nile and go all the way to the top, and see all of the sites. It was just incredible to see all of these historical sites that had been there for millennia. That's always been one of my favorite trips. It's such a different culture and history than the ones that I'm used to in France, India, and the U.S.</p>New Caltech Center Sheds Light on the Future of Generative AI, Innovation, and Regulation2023-05-19T20:22:00+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/generative-ai-regulation-kevin-roose<p data-block-key="o08ft">From weather forecasts and disease diagnosis to chatbots and self-driving cars, new applications of <a href="https://scienceexchange.caltech.edu/topics/artificial-intelligence-research/artificial-intelligence-definition?utm_source=caltechnews&utm_medium=web&utm_campaign=cseai">artificial intelligence</a> (AI) continue to multiply. More recently, the widespread availability of tools that can create content—whether code, text, images, audio, or video—such as ChatGPT and DALL-E, has thrust "generative AI" into the spotlight.</p><p data-block-key="bh852">As applications proliferate, so do complex questions about how to ensure responsible use of generative AI. To explore the societal implications of AI technology and how policymakers might approach regulating it, the Caltech <a href="https://lindeinstitute.caltech.edu/research/csspp">Center for Science, Society, and Public Policy</a> (CSSPP) <a href="https://lindeinstitute.caltech.edu/research/csspp/csspp-events/csspp-roundtable-conversation-may-5-2023">hosted</a> a conversation among researchers, industry representatives, and the public on Caltech's campus. The CSSPP was <a href="https://www.caltech.edu/about/news/new-center-aims-to-help-shape-public-science-policy">established</a> in early 2023 to examine the intersection of science and society, provide a forum for the discussion of scientific ethics, and help shape public science policy. The center is affiliated with <a href="https://lindeinstitute.caltech.edu/">The Ronald and Maxine Linde Institute of Economic and Management Sciences</a>.</p><p data-block-key="20ejd">"We believe that scientific knowledge and technological prowess are essential to any meaningful evaluation of the impacts of AI on society," said Caltech president Thomas F. Rosenbaum, the Sonja and William Davidow Presidential Chair and professor of physics. "This is true for the positives and for the negatives: whether it be lifesaving improvements to health screening, powerful tools for artistic creation, and new ways of approaching science or potential upheavals in the job market, propagation of false information, and new weapons of war. Only through this type of informed evaluation can we amplify the salutary aspects of technological development and counter its dehumanizing capacity."</p><p data-block-key="9hm5g">The event featured an introduction on the state of generative AI from <i>New York Times</i> technology columnist <a href="https://www.nytimes.com/by/kevin-roose">Kevin Roose</a> (who famously had an <a href="https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html">unnerving conversation</a> with Microsoft's Bing chatbot).</p><p data-block-key="5b9rc">In his keynote, Roose reminded the audience of the power of shared responsibility and knowledge. "One of the advantages that AIs have over humans is that they have networked intelligence: When one node in a neural network learns something or makes a connection, it propagates it through to all the other nodes in the neural network. When one self-driving car in a fleet learns about a new kind of obstacle, it feeds that information back into the system," he said. "Humans don't do that, by and large. We silo information, we hoard it, we keep it to ourselves. And I think that if we want a realistic shot at competing and thriving and succeeding, and maintaining our agency and our relevance in this new era of generative AI, we really need to do it together." While on campus, Roose also participated in a Q&A session with nearly 50 Caltech students.</p><p data-block-key="74blb">In a subsequent panel discussion moderated by R. Michael Alvarez, professor of political and computational social science and co-director of the CSSPP, experts in law, gaming and technology, and academic research shared thoughts on the positive and negative potential of generative AI.</p><p data-block-key="ct8ij">The optimistic outlook centers on AI's power to advance science and engineering, for example, by making it possible to predict genome sequences of new COVID-19 variants before they appear in nature, design better medical equipment, and mitigate climate change.</p><p data-block-key="59gn0">"How do we capture CO2 and store it underground? How do we plan for the right reservoir and the right amount of CO2 to store? These are the kinds of complex processes that our human minds can't even grapple with," said Anima Anandkumar, Bren Professor of Computing and Mathematical Sciences and co-leader of the <a href="https://www.ai4science.caltech.edu/index.html">AI4Science</a> initiative at Caltech. "We are using [generative] AI, and we are doing it much faster. Along with it comes the benefit of being able to come up with new discoveries, new inventions."</p><p data-block-key="6ndj8">Speaking from a more skeptical perspective, panelists raised concerns over intellectual property and copyright, bias, and large-scale misinformation.</p><p data-block-key="c7se0">Additionally, when generative AI technologies are coupled with the massive amount of personal data consumers share with social media algorithms, our own biases can become vulnerable to manipulation, Carly Taylor, a data scientist and security strategist at Activision Publishing pointed out. "All of us are capable of being bamboozled," Taylor said. "Everyone has confirmation biases, and in many cases across social media, we have spent every single day for years telling Facebook, Instagram, and LinkedIn exactly what we are biased toward by what we search, what content we consume, or with whom we engage … As a risk, that can become completely exploitable."</p><p data-block-key="774c0">Justin Levitt, Gerald T. McLaughlin Fellow and professor of law at Loyola Law School, shared his pessimism about AI's impact on democracy in the United States, including the ability to rapidly spread election misinformation. "Democracy depends on a set of different opinions and a set of common facts, and generative AI is going to be great for giving us an infinite array of disparate facts," he said. "That's a disaster for democracy."</p><p data-block-key="3hl60">Sean Comer, an applied researcher at Activision and its Infinity Ward development studio, saw a silver lining in recent anxiety over generative AI. "Maybe it gives us the elephant in the room to address the attention economy, which a lot of misinformation tends to stem from," he said. "Maybe it's a necessary evil that can force us to deal with these things."</p><p data-block-key="9sqfg">Importantly, speakers addressed the role of research institutions like Caltech. For example, avoiding biases in new AI models will take the kind of critical thinking and rigorous testing for which academia is known. The CSSPP will continue to foster these types of conversations by bringing policymakers to campus for lectures, colloquia, discussion panels, and workshops in addition to developing undergraduate and graduate courses that cover issues in scientific ethics and policy, and consider how policy may be augmented by scientific ethics and expertise.</p><h5 data-block-key="o08ft">Learn more about the <a href="https://lindeinstitute.caltech.edu/research/csspp">Caltech Center for Science, Society, and Public Policy (CSSPP).</a></h5>No Magic Number for Time It Takes to Form Habits2023-04-17T19:00:00+00:00Whitney Clavinwclavin@caltech.eduhttps://sites.caltech.edu/newspage-index/no-magic-number-for-time-it-takes-to-form-habits<p data-block-key="ktcf2">Putting on your workout clothes and getting to the gym can feel like a slog at first. Eventually, you might get in the habit of going to the gym and readily pop over to your Zumba class or for a run on the treadmill. A new study from social scientists at Caltech now shows how long it takes to form the gym habit: an average of about six months.</p><p data-block-key="3tdqc">The same study also looked at how long it takes health care workers to get in the habit of washing their hands: an average of a few weeks.</p><p data-block-key="4ecqm">"There is no magic number for habit formation," says Anastasia Buyalskaya (PhD '21), now an assistant professor of marketing at <a href="https://www.hec.edu/en/faculty-research/faculty-directory/faculty-member/buyalskaya-anastasia">HEC Paris</a>. Other authors of the study, which appears in the journal <i>Proceedings of the National Academy of Sciences,</i> include Caltech's <a href="https://www.hss.caltech.edu/people/colin-f-camerer">Colin Camerer</a>, Robert Kirby Professor of Behavioral Economics and director and leadership chair of the T&C Chen Center for Social and Decision Neuroscience, and researchers from the University of Chicago and the University of Pennsylvania. Xiaomin Li (MS '17, PhD '21), formerly a graduate student and postdoctoral scholar at Caltech, is also an author.</p><p data-block-key="d0jpl">"You may have heard that it takes about 21 days to form a habit, but that estimate was not based on any science," Camerer says. "Our works supports the idea that the speed of habit formation differs according to the behavior in question and a variety of other factors."</p><p data-block-key="9htrk">The study is the first to use machine learning tools to study habit formation. The researchers employed machine learning to analyze large data sets of tens of thousands of people who were either swiping their badges to enter their gym or washing their hands during hospital shifts. For the gym research, the researchers partnered with 24 Hour Fitness, and for the hand-washing research, they partnered with a company that used radio frequency identification (RFID) technology to monitor hand-washing in hospitals. The data sets tracked more than 30,000 gymgoers over four years and more than 3,000 hospital workers over nearly 100 shifts.</p><p data-block-key="4asci">"With machine learning, we can observe hundreds of context variables that may be predictive of behavioral execution," explains Buyalskaya. "You don't necessarily have to start with a hypothesis about a specific variable, as the machine learning does the work for us to find the relevant ones."</p><p data-block-key="99roa">Machine learning also let the researchers study people over time in their natural environments; most previous studies were limited to participants filling out surveys.</p><p data-block-key="dsfor">The study found that certain variables had no effect on gym habit formation, such as time of day. Other factors, such as one's past behavior, did come into play. For instance, for 76 percent of gymgoers, the amount of time that had passed since a previous gym visit was an important predicator of whether the person would go again. In other words, the longer it had been since a gymgoer last went to the gym, the less likely they were to make a habit of it. Sixty-nine percent of the gymgoers were more likely to go to the gym on the same days of the week, with Monday and Tuesday being the most well attended.</p><p data-block-key="6hjke">For the hand-washing part of the study, the researchers looked at data from health care workers who were given new requirements to wear RFID badges that recorded their hand-washing activity. "It is possible that some health workers already had the habit prior to us observing them, however we treat the introduction of the RFID technology as a ‘shock' and assume that they may need to rebuild their habit from the moment they use the technology," Buyalskaya says.</p><p data-block-key="3vf53">"Overall, we are seeing that machine learning is a powerful tool to study human habits outside the lab," Buyalskaya says.</p><p data-block-key="44jl0">The study titled "<a href="https://www.pnas.org/doi/10.1073/pnas.2216115120">What can machine learning teach us about habit formation? Evidence from exercise and hygiene</a>" was funded by the Behavior Change for Good Initiative, the <a href="https://lindeinstitute.caltech.edu/">Ronald and Maxine Linde Institute of Economics and Management Sciences</a> at Caltech, and the <a href="https://neuroscience.caltech.edu/">Tianqiao and Chrissy Chen Institute for Neuroscience</a> at Caltech.</p>New Center Aims to Help Shape Public Science Policy2023-03-30T16:38:00+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/new-center-aims-to-help-shape-public-science-policy<p data-block-key="ix9rj">Sometimes it seems as though scientific advancement occurs at such a rapid pace that its effects on society are barely considered until they have already happened. A new center established at Caltech seeks to examine this intersection of science and society, provide a forum for the discussion of scientific ethics, and help shape public science policy.</p><p data-block-key="68f5l">The <a href="https://lindeinstitute.caltech.edu/research/csspp">Center for Science, Society, and Public Policy</a> (CSSPP) will be affiliated with <a href="https://lindeinstitute.caltech.edu/">The Ronald and Maxine Linde Institute of Economic and Management Sciences</a> in the Division of the Humanities and Social Sciences (HSS), and it counts two of the division's faculty members as its co-directors. They are <a href="https://www.hss.caltech.edu/people/r-michael-alvarez">R. Michael Alvarez</a>, professor of political and computational social science; and <a href="https://www.hss.caltech.edu/people/frederick-eberhardt">Frederick Eberhardt</a>, professor of philosophy. The center has also hired two postdoctoral scholars.</p><p data-block-key="595eh">"We represent different aspects of HSS in an initiative like this," Alvarez says. "I come from the social sciences, and in the social sciences, we've had a long history of faculty who have been directly involved in various types of public policy over the years. That includes me, in particular when it comes to elections and election administration. I've got some experience and a little bit of academic understanding about how politics and the public policy process works."</p><p data-block-key="fj7h7"></p><embed alt="A portrait of R. Michael Alvarez. He stands in front of a book case and smiles." embedtype="image" format="RightAlignSmall" id="47310"/><p data-block-key="e45ab"></p><p data-block-key="1caq1">Eberhardt, whose own work has examined the philosophy of science, says he began to consider science's responsibility to the public several years ago when he realized that although Caltech had courses on artificial intelligence (AI), it had none on the ethics of AI.</p><p data-block-key="dh6hl">"I was very concerned that this is something we needed to teach our students," Eberhardt says. "This is something they need to know and think about, especially nowadays. So, I put that course together. It has been extremely well received over the years.</p><p data-block-key="66a8k"></p><embed alt="A portrait of Frederick Eberhardt. He smiles and is outdoors." embedtype="image" format="LeftAlignSmall" id="47311"/><p data-block-key="2dra"></p><p data-block-key="8ejpd">"I think there's a feeling of responsibility among the philosophers at Caltech that we need to do our part in helping people consider these issues," Eberhardt adds. "What brings me to the table is this obligation that something needs to be done, quite apart from the fact that there is a wealth of tricky and interesting topics to work on in this space."</p><p data-block-key="66bbi">Eberhardt's experience with the ethics of AI may be helpful as the center launches. <a href="https://scienceexchange.caltech.edu/topics/artificial-intelligence-research?utm_medium=web&utm_campaign=cseai&utm_source=caltechnews&utm_content=&utm_term=">Artificial intelligence</a> has been a topic of interest lately because of the release of ChatGPT, a chatbot with advanced linguistic abilities that often produces factually incorrect material. Besides its lack of veracity, ChatGPT's ability to "write" lengthy papers and essays from a prompt has caused concern among educators who say it may enable students to produce assignments without doing the work themselves.</p><p data-block-key="85v3s">"The whole debate around ChatGPT and the large language models has people in academia asking what do we do with this? What <i>can</i> we do with this? And how will it change the way students can and should write papers? It's obviously a difficult discussion to have, and it's not obvious what the solutions are."</p><p data-block-key="1mdgd">Eberhardt adds that concerns about <a href="https://scienceexchange.caltech.edu/topics/artificial-intelligence-research/trustworthy-ai?utm_medium=web&utm_campaign=cseai&utm_source=caltechnews&utm_content=&utm_term=">how AI is used</a> go beyond its potential for abetting plagiarism. These AI systems have also found a home in the legal system, where they have been criticized for automatically providing harsher sentences for Black people convicted of crimes than white people convicted of similar crimes.</p><p data-block-key="1cuc2">"These automated decision processes are being used in incarceration and for granting parole," he says. "We didn't have a discussion about them in time as a society, and now we're scrambling to find out what sort of recourse we have for decisions made by these automated procedures."</p><p data-block-key="94bg1">Similar discussions should be held, Eberhardt and Alvarez say, about when and how biological tools like gene editing are used, and about other ethical issues regarding biomedical research.</p><p data-block-key="c75jm">For those kinds of conversations to happen before a new technology becomes entrenched, perhaps in an irresponsible way, Eberhardt and Alvarez say scientists and researchers need to understand the process of how scientific results are translated into policy. Having such conversations with policy makers would also help the public understand why research is important, and in turn, it would help the policy makers decide what kind of rules or priorities to lay out.</p><p data-block-key="8hpdv">"There's this huge gap between doing the scientific research, considering the impact of that sort of scientific research, communicating that scientific research out to the public, understanding what the regulations are, and how to shape those sorts of regulations," Eberhardt says. "The underlying goal of this center is to make students, postdocs, and faculty literate in this disconnect and make them able to bridge it. I think the hope is that we might actually have some influence on policymaking."</p><p data-block-key="a1lnu">Alvarez says students will be a vital part of that effort.</p><p data-block-key="c4q80">"It's very clear that there are a lot of Caltech students who want to make a difference in the world. They want their science to matter," he says. "What we hope is that, through public events, connecting students with policy makers, and helping them understand how to communicate their research in policy-relevant and public-relevant ways, more students will get involved directly in policy making and be better equipped to be consumers of policy information. Whatever career path they pursue, we hope they will be better able to make their science relevant to the world around them."</p><p data-block-key="6t9sg">Eventually, the center will offer courses for students, a discussion forum for Summer Undergraduate Research Fellowship (SURF) students on science policy careers, as well as research projects for students and faculty members. The center's public programming efforts will begin May 5 with <a href="https://www.hss.caltech.edu/news-and-events/calendar/csspp-workshop-panel">a panel discussion on generative AI</a>.</p>Reducing Procrastination with Tailored Incentives2023-02-07T18:25:00+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/reducing-procrastination-with-tailored-incentives<p data-block-key="mpdad">If you have an assignment due in two weeks, do you work on it right then and there, or do you wait a week-and-a-half and rush to finish it the night before it's due?</p><p data-block-key="7j82u">If you're like many people (and the writer of this piece), you're going to wait until last-minute panic sets in before you buckle down and get your work done.</p><p data-block-key="c8knd">Procrastination is common, and often it is even expected, but that does not mean it is completely harmless. In some situations, it can have a significant effect on productivity. For Caltech's <a href="https://www.hss.caltech.edu/people/charles-sprenger">Charlie Sprenger</a>, professor of economics, procrastination is a problem that can be solved with the tools of his trade, and he has started in a place where its effects can be acute: health care.</p><p data-block-key="lsdu2"></p><embed alt="A portrait of Charlie Sprenger. He is outdoors and wears a button-down shirt." embedtype="image" format="RightAlignSmall" id="46109"/><p data-block-key="2rnd3"></p><p data-block-key="ep0cv"></p><p data-block-key="88thb">In a paper published in the <i>Journal of the European Economic Association</i>, Sprenger and colleagues from UC San Diego, the London School of Economics, Germany's FAU Erlangen-Nuremberg, and the University of Pittsburgh show how they used economics principles to tackle the problem of procrastination during polio vaccination drives in Pakistan.</p><p data-block-key="1vik4">During those drives, health care workers are tasked with going out into neighborhoods and administering vaccinations to a set number of children over a two-day period. The system used to track those vaccinations was not exactly robust: The workers carried with them a piece of chalk and tallied the number of doses they administered on the outside walls of homes.</p><p data-block-key="a96cm">Eventually, Pakistani health authorities decided it could do better with a mobile phone app, which is where Sprenger and his colleagues came in. The app provided the authorities with better data—data that showed the workers were procrastinating. If a worker was expected to administer 200 doses in two days, they might do 50 the first day, and then rush to finish the remaining 150 on the second day, for example.</p><p data-block-key="8joei">"That's the standard force of procrastination that most of us are familiar with when we're under deadline," Sprenger says. "We postpone the unpleasant task of doing the work until close to the deadline and then have to rush and maybe don't provide the best work that we can, given our time constraints."</p><p data-block-key="cvvbt">Because of the logistics of distributing vaccines and their need to be kept refrigerated, health authorities want them administered in a smooth (predictable and consistent) way. A last-minute rush is not ideal.</p><p data-block-key="f24r8">The researchers conducted interviews with the workers to get a better idea of how they wanted to perform their work—essentially how much work did they want to do on day one and on day two. Using those preferences, Sprenger and colleagues developed tailored incentives for some of the workers.</p><p data-block-key="br2ps">"You could think about various ways to implement this," Sprenger says. "You could say, ‘Look, I'll pay you 10 cents per vaccine on day one and 5 cents per vaccine on day two. Doesn't that make you want to do your work earlier?' Or you could think about it differently, saying vaccines that you do on day one count for two times as much as vaccines that you do on day two."</p><p data-block-key="4d67s">By tailoring these kinds of incentives to individual workers based on their stated preferences, the research team showed that it was possible to reduce procrastination substantially. (They generate "behavior around 30 percent closer to the policy target of equal allocation," to be specific.)</p><p data-block-key="72cnf">That is a lesson that Sprenger says could be applied in many policy areas.</p><p data-block-key="eldd3">"It didn't need to be Pakistan, it didn't need to be government health workers. It could have been anything," he says. "This fits into a bigger agenda of using the tools of experimental economics to measure preferences and change incentive schemes based on those measurements in an individualized way."</p><p data-block-key="1r7o2">As far as the app goes, Sprenger says the Pakistani government turned off the preference/incentives feature after the team completed its study, but the rest of the app is still in use, allowing Pakistan to monitor its vaccination efforts.</p><p data-block-key="41rpn">The paper describing the research is titled "<a href="https://resolver.caltech.edu/CaltechAUTHORS:20230113-708423000.11">Using Preference Estimates to Customize Incentives: An Application to Polio Vaccination Drives in Pakistan</a>." Sprenger's co-authors are James Andreoni of UC San Diego, Michael Callen of the London School of Economics, Karrar Hussain of FAU Erlangen-Nuremberg, and Muhammad Yasir Khan of the University of Pittsburgh.</p><p data-block-key="f681i">Funding for the research was provided by the International Growth Centre the Pershing Square Fund for Research on the Foundations of Human Behavior, and the Department for International Development Building Capacity for Research Evidence pilot program initiative.</p>Caltech and Activision Publishing Team Up to Combat Bad Behavior Online2022-12-15T19:51:00+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/caltech-and-activision-publishing-team-up-to-combat-bad-behavior-online<p data-block-key="n6trf">Whether it is trolling, racism, sexism, doxing, or just general harassment, the internet has a bad behavior problem. Researchers from Caltech and <a href="https://www.activision.com/company/aboutus">Activision Publishing</a>, a video game publisher, are working together to bring their combined expertise to address this behavior in video games.</p><p data-block-key="96b9u">Because this kind of toxic behavior makes the internet an unpleasant place to be, there have been many attempts over the years to make sure people behave themselves online. In the earlier days of the internet, websites often relied on moderators—volunteers or staff—who were trained to keep discussions and content civil and appropriate. But as the internet continued to grow and harmful behaviors became more extreme, it became apparent that moderators need better tools at their disposal.</p><p data-block-key="mhvk">Increasingly, the online world is moving toward automated moderation tools that can identify abusive words and behavior without the need for human intervention. Now, two researchers from Caltech, one an expert in artificial intelligence (AI) and the other a political scientist, are teaming up with Activision on a two-year research project that aims to create an AI that can detect abusive online behavior and help the company's support and moderation teams to combat it.</p><p data-block-key="pq54">The sponsored research agreement involves <a href="https://www.eas.caltech.edu/people/anima">Anima Anandkumar</a>, the Bren Professor of Computing and Mathematical Sciences, who has trained AI to fly drones and study the coronavirus; <a href="https://www.hss.caltech.edu/people/r-michael-alvarez">Michael Alvarez</a>, professor of political and computational social science, who has used machine learning tools to study political trends in social media; and Activision's data engineers, who will provide insight into player engagement and game-driven data.</p><p data-block-key="d1okq">Alvarez and Anandkumar have already worked together on <a href="https://www.caltech.edu/about/news/ai-metoo-training-algorithms-spot-online-trolls">training AI to detect trolling</a> in social media. Their project with the team that works on the <i>Call of Duty</i> video games will allow them to develop similar technology for potential use in gaming.</p><p data-block-key="8dudn">"Over the past few years, our collaboration with Anima Anandkumar's group has been very productive," Alvarez says. "We have learned a great deal about how to use large data and deep learning to identify toxic conversation and behavior. This new direction, with our colleagues at Activision, gives us an opportunity to apply what we have learned to study toxic behavior in a new and important area—gaming."</p><p data-block-key="de33q">For Anandkumar, the important questions this research will answer are: "How do we enable AI that is transparent, beneficial to society, and free of biases?" and "How do we ensure a safe gaming environment for everyone?"</p><p data-block-key="6982u">She adds that working with Activision gives the researchers not only access to data about how people interact in online games, but also to their specialized knowledge.</p><p data-block-key="f9l16">"We want to know how players interact. What kind of language do they use? What kinds of biases do they have? What should we be looking for? That requires domain expertise," she says.</p><p data-block-key="5vqjr">Michael Vance, Activision 's chief technology officer, says the firm is excited to work with Caltech.</p><p data-block-key="9f2m2">"Our teams continue to make great progress in combating disruptive behavior, and we also want to look much further down the road," Vance says. "This collaboration will allow us to build upon our existing work and explore the frontier of research in this area."</p>Individual Decisions and Complex Economies2022-12-02T19:34:09.249581+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/individual-decisions-and-complex-economies<p data-block-key="wrxzb">Every day, all across the world, people make billions of choices, such as how much they are willing to pay for vegetables, how much time they should spend fixing their car, or whether they will keep their Netflix subscription or switch to another streaming service. At the individual level, you might call these personal preferences, but when you add them all up, you have an economy.</p><p data-block-key="e618q">The field of behavioral economics attempts to predict big economic trends by first trying to understand the individual preferences that serve as the building blocks of group behavioral trends. Peter Caradonna, a new assistant professor of economics in the Division of the Humanities and Social Sciences, is working to build better tools for studying these individual preferences. We recently talked to him about his research and what excites him most in his field.</p><h3 data-block-key="5mcvi">What is the relationship between what you study and what the average person thinks of as economics?</h3><p data-block-key="7bmua">Things like the state of the nation's economy, or the stock market, or phenomena like business cycles are really complicated and have lots of moving parts. One of the ways that we try to better understand these tremendously complex systems is through the use of simplified models. But we want these models to be "micro-founded." What this means is that underneath all the other bells and whistles, the models should be built on a foundation of individual, rational actors trying to do whatever they think is in their self-interest.</p><p data-block-key="aftjm">Now to specify things like what we mean by self-interest, we need to be able to describe, in a stylized way, the ways people act, make decisions, and evaluate different kinds of trade-offs. And this is where what I study comes in. Ideally, we'd like to know how good our assumptions about individual behavior are, really to be able to compare competing sets of assumptions, and see how well they perform in practice. Because if these basic assumptions are off base, it is going to bleed into every other type of analysis we could hope to conduct. But this turns out to be a very difficult exercise. So I'm very interested in developing new, and better-performing tools to try and really test and select amongst these models of how people behave.</p><h3 data-block-key="8doul">You used the phrase, "rational actor." I think a lot of folks would say, "People are not always rational." How much of a challenge for your field is it that people can be irrational?</h3><p data-block-key="a4a1j">When I said "rational actor," I was accidentally slipping into some technical jargon. When economists use the phrase "rational," we mean it in a slightly different way, I think, than a layperson might interpret it.</p><p data-block-key="baqbd">For us, all rationality means is that a person has an internally consistent ranking in their head of the relative desirability of different outcomes, or consequences of their actions. Rationality doesn't necessarily mean that I'm all cool, calm, and collected. Likewise, irrationality doesn't mean that I'm acting a little crazy. It also doesn't necessarily mean that I'm a sociopath who only cares for myself at the expense of others.</p><p data-block-key="80l9n">To economists, all rationality means is that an individual has this basic, consistent ordering called a preference that says, "I like this more than that." Anyone who satisfies that basic property, we call a rational actor.</p><h3 data-block-key="r8vu">What does an experiment in economics look like? Can you paint a picture for me?</h3><p data-block-key="f0mrt">One of the hardest parts of studying these models of individual decision making is that preference orderings—these things which we view as driving all individual behavior—are unobservable. So if I have a theory that says "people prefer to hold assets with a certain kind of risk pattern over another," it's hard to test because I can't just see these preferences directly. Instead, I have to indirectly infer them by observing people's behavior. This makes having access to a lab environment really valuable. Oftentimes, experiments involve presenting subjects, in a carefully incentivized way, with specific sets of things to choose between. If I just wanted to test something very simple, like whether or not a subject was rational, that is, has some consistent preference, I could provide them with a collection of different prizes—an apple, a banana, and a cantaloupe, for example.</p><p data-block-key="360b5">I could first say, "Here's an apple, and here's a banana. Tell me which one you like more." Then I could do the same thing with the banana and a cantaloupe. So maybe they choose an apple instead of a banana and chose a banana instead of a cantaloupe. From that, I'd infer that these choices reflect their hidden preference.</p><p data-block-key="80kr7">Now, if I wanted to test if there is any kind of basic ranking that could describe the agent's preferences more generally, I could then say, "Here's an apple, and here's a cantaloupe." Now, if you already told me you like an apple more than a banana, and you like a banana more than a cantaloupe, you ought to like an apple more than a cantaloupe, at least if you're rational. If their choices reflect that, great! But if not, then you get into very interesting waters trying to quantify the <i>magnitude</i> of their irrationality.</p><p data-block-key="b74qj"></p><embed alt="An apple and a banana with a greater-than sign open toward the apple." embedtype="image" format="MiddleAlignLarge" id="45189"/><p data-block-key="17e6v"></p><embed alt="A banana and a cantaloupe with a greater-than sign open toward the banana." embedtype="image" format="MiddleAlignLarge" id="45190"/><p data-block-key="fr6c2"></p><embed alt="An apple and a cantaloupe with a question mark between them." embedtype="image" format="MiddleAlignLarge" id="45191"/><p data-block-key="qb7c"></p><p data-block-key="bk3k6">More generally, we like to try and build up these little artificial encounters that are able to reveal things about how the agent is thinking and what tradeoffs they consider when making decisions. Maybe the agent is given some budget of money that they're allowed to spend in the lab on bundles of different amounts of food and drink. Maybe they're choosing different portfolios of financial assets with uncertain returns, or rewards being delivered at different points in time. Maybe they have to choose how they allocate time between more or less computationally complex tasks.</p><p data-block-key="7k9h4">These choice-type experiments are often very revealing about the drivers of the decision-making process. One thing my research touches on is studying what types of questions we can ask subjects to obtain the most revealing data, and developing new statistical tools to better analyze it.</p><h3 data-block-key="739vg">Are there any trends in your field right now that are especially exciting to you?</h3><p data-block-key="a0t45">There's a whole host of very interesting new work being done on these kind of problems, both from theoretical and empirical perspectives. If I had to pick one thing, lately there's been some very interesting work using modern machine-learning techniques to see how good a model is at predicting behavior that I'm interested in learning more about.</p><p data-block-key="3ur1n">More generally, one of the things that I find particularly appealing is that there really isn't any kind of broad consensus on the "best" way to study these models. At a technical level, there's lots of very interesting innovation going on using a whole range of mathematical and statistical tools, some of which are only just starting to be put to use in economics more broadly. So it's a very exciting time.</p><h3 data-block-key="5c8ih">When did you decide that economics was the field for you, and what was it about economics that was interesting?</h3><p data-block-key="eq0md">When I was first entering college, I really wanted to be a diplomat, which in hindsight would've been a terrible fit for me. I got interested in economics originally because it seemed at the time to be getting at the cogs and wheels behind lots of international affairs questions I found interesting. I thought that a lot of times, the various incentives faced by actors on the global stage boiled down into economic considerations.</p><p data-block-key="6aph7">When I was an undergraduate, I took a math class over the summer as a requirement for my major, and I had this instructor who had just finished his PhD and was going off to a postdoc position in the fall. He was young, and he was exciting, and he was cool, which I had never seen in anyone who did math before.</p><p data-block-key="empds">I just remember sitting in his class—it was multivariable calculus—with eyes like saucers for the entire eight weeks. I was so smitten, I declared as a math/econ major immediately that fall. And as time went on, I got more interested in the technical, theory side of economics, originally because of the mathematical tools involved.</p><p data-block-key="1tpbc">These days, one of the most appealing parts of economics to me is that, in its modern form, it's so young compared to, say, the natural sciences. There are so many basic questions that we as a field just aren't that close to satisfactorily answering yet, and that makes it this kind of wonderful intellectual melting pot, where all kinds of tools and ideas from a whole host of different fields are being combined in new and exciting ways.</p><h3 data-block-key="434ks">How do you like to spend your free time when you are not working?</h3><p data-block-key="631ln">One of the things I really enjoyed doing when I was younger was hiking. I did a whole lot of hiking in the Adirondack Mountains—I was born on the East Coast—growing up. But the kind of mountains we have back East are nothing like what there is out here. I'm really looking forward to exploring some of the wilderness out around Pasadena and the Los Angeles area more generally. This past weekend, I hiked Mt. Baldy with one of my colleagues, which is the highest mountain in L.A. County, and while it was some steep hiking, it was just gorgeous. I'm really excited to get out more around here, and hopefully this spring to do some hiking out in the national parks!</p>A Conversation with Pawel Janas2022-10-21T18:39:00+00:00Emily Velascoevelasco@caltech.eduhttps://sites.caltech.edu/newspage-index/a-conversation-with-pawel-janas<p data-block-key="igmvg">When a crisis occurs, often the only thing a person can control is how they respond. This is true not just for individuals, but for families, companies, and governments.</p><p data-block-key="90ikj">How governments respond to crises, specifically financial crises, can have profound effects on their economies and the livelihoods of their constituents. For Pawel Janas, a new assistant professor of economics at Caltech, understanding how governments respond to crises is also an academic pursuit. Janas is an economic historian with an interest in how local governments react when they face financial hardship.</p><p data-block-key="7grhg">We recently spoke with him about his research, what brought him to Caltech, and the sport he plays semiprofessionally.</p><h3 data-block-key="u06j">Can you tell us about your research?</h3><p data-block-key="20m65">I consider myself an economic historian who also knows finance. To me, economics is telling stories with data and some fancy math, but I think it's very dangerous for people to have the same story in their minds without ever challenging that story. A large part of what economic historians do is revisit these economic stories that have been told through generations and see what we can learn from those episodes.</p><h3 data-block-key="e2nc9">What are some specific examples of things that you study?</h3><p data-block-key="8g6b3">I research financial crises—what happens when there's a huge downturn in the economy. My specific research interest has so far been the Great Depression and what happened when cities all of a sudden couldn't pay their debts and became financially constrained. I want to see what cities and their administrations did to cope with a financial crisis.</p><h3 data-block-key="3fc6g">Why is this kind of research important? What can you learn that is applicable today by looking at things that happened during the Great Depression?</h3><p data-block-key="2p6b0">That's a great question, but I think I want to push back on that. I don't think all the research that we do at Caltech should be about policy or about being relevant today, because we have no idea what the future is going to bring.</p><p data-block-key="8dm10">But cities and counties <i>are</i> borrowing a lot of money today, and we want to know what happens when financial markets freeze and they can't borrow as much as they want to. That's going to happen at some point, and it turns out that historical episodes give us the right environment and the right laboratory to look at these core economic questions without having a very direct 2022 policy implication.</p><h3 data-block-key="mdo6">When you look at how the Great Depression affected local municipalities, what do you see?</h3><p data-block-key="4l9u4">You see trade-offs. Cities have a variety of things that they do, right? They provide education; they provide police; they provide fire protection; they paid welfare back in the day, and they also build a lot of infrastructure. So, cities do a lot of things, and they of course borrow their money and they pay off those debts.</p><p data-block-key="12pca">My research shows that the cities that were very highly indebted were also the same cities that were cutting back on the public services and public goods that we would expect them to continue funding.</p><h3 data-block-key="3ml9j">What are some of the projects that you are working on right now?</h3><p data-block-key="heik">I have three projects that deal with the Great Depression. The first project is what I just described with the cities when they go bankrupt.</p><p data-block-key="ffp63">My second project is about schools, again during the Great Depression. Schools back in the day were funded mostly through local property taxes, and the Great Depression decimated the local property tax base, so schools were being severely underfunded. This project looks at which types of students kept going to school during the Great Depression and which ones dropped out. I'm trying to understand the implications for those people down the road, whether they returned to education and what their labor outcomes were.</p><p data-block-key="dgpo0">The third project looks at the question of what happens when banks close. During the Great Depression, banks in this one region closed at a higher rate than similar looking banks in other regions. This actually offers us a really nice experiment for studying the effect of bank closures on economic activity.</p><h3 data-block-key="a210t">How do you like to spend your free time?</h3><p data-block-key="ab98c">My biggest and only hobby is playing Ultimate Frisbee, and I've been doing it for 15 years now. I played on my high school team in Colorado. I also played during college in Colorado, and when I moved to Chicago for my PhD, I played in all the club teams in Chicago and the semi-pro teams in Chicago. Right now, I'm still traveling to play on the club team in Chicago.</p><h3 data-block-key="7eo1c">What is it about ultimate frisbee that appeals to you?</h3><p data-block-key="amp0s">The fact that these grown adults spend a lot of money and time playing this game means that the community is very strong. I don't know in what other area of life you'll find a bunch of men and women in their late 20s and early 30s spending a lot of money and time doing a sport that's not professional, and that they don't get paid for. It's a really special group of people.</p>Caltech Charts the Course to a Green Electrical Grid2022-08-24T20:30:00+00:00Caltech Media Relationsmr@caltech.eduhttps://sites.caltech.edu/newspage-index/smart-grid-solutions-california-renewables<p data-block-key="1oekw">California's energy grid, an engineering marvel 150 years ago, is due for a makeover. Now, thanks to an interdisciplinary group of researchers, Caltech is working to transform energy systems by developing a "smart grid": a flexible, responsive, efficient, system that incorporates renewable energy sources while meeting growing power demands.</p><p data-block-key="5ei8p">Californians increasingly are experiencing the <a href="https://scienceexchange.caltech.edu/topics/sustainability/effects-climate-change?utm_source=caltechnews&utm_medium=web&utm_campaign=csesustainability">effects of climate change</a>, from drought and water shortages to extreme heat. These challenges highlight the urgent need to transition away from greenhouse gas-emitting fossil fuels and toward renewable energy sources such as wind and solar. In 2018, California <a href="https://www.energy.ca.gov/sb100">committed</a> to supplying electric customers with 100 percent renewable and zero-carbon energy sources by 2045.</p><p data-block-key="f34g8">A crucial step in that transition is restructuring the power grid, the complicated network of hardware and software that brings electricity to our doorsteps. With support from public and private partners, Caltech engineers, economists, mathematicians, and computer scientists are devising and testing the underpinnings of tomorrow's grid. Their advancements include everything from the creation of intuitive algorithms and hardware that ensure that electrical vehicle (EV) charging stations draw power when it is cheap and abundant to the establishment of new economic structures that prevent market players from manipulating energy prices.<br/></p><blockquote data-block-key="3lpl1">"A lot of the infrastructure has already reached a place where it can be integrated and implemented," says Adam Wierman, professor of computing and mathematical sciences. "There are smart devices for sensing and communication, and solar and wind are cost competitive. That means that there will be progress in the short term. The question is when we will hit this wall where the current architecture stops being enough—where power outages start to become more likely, when costs start to go up. Our research is targeted on getting through that wall, so that when you get there, you can still progress."</blockquote><h3 data-block-key="4chid"><b><br/>Yesterday's Electrical Grid</b></h3><p data-block-key="dg28v">Conventional grids have relied on coal-fired or nuclear power stations located near areas of high electricity demand. Renewable energy generation systems, like solar panels and wind turbines, are often located in open spaces far away from urban areas, thus creating the need for more storage (e.g., pumped hydroelectric energy storage or batteries) and greater transmission capability (e.g., power lines).</p><p data-block-key="1eu34">The grid also was designed to distribute a steady and predictable flow of electricity created by burning fossil fuels on demand. Renewable power from wind and solar fluctuate depending on weather and are difficult to predict. Without <a href="https://scienceexchange.caltech.edu/topics/sustainability/battery-technology-power-grid?utm_source=caltechnews&utm_medium=web&utm_campaign=csesustainability">long-term, large energy storage</a>, this can result in an imbalance between the amount of electricity produced and the amount needed at any given time.</p><p data-block-key="3m00a">Compounding these challenges is the expectation that electricity demand will soar as more people drive electric vehicles and install electric appliances in an effort to decarbonize. The grid of the future must be able to grow to accommodate this increased demand.</p><h3 data-block-key="4m93o"><b>Caltech Smart Grid Solutions</b></h3><p data-block-key="98e7d">When energy supply is less predictable, energy demand needs to be more flexible to avoid overtaxing the grid. Smart scheduling—pulling power from the grid when renewable energy is available—could make the grid more compatible with the variability of wind and solar power.</p><embed alt="man with electric vehicle and charging station" embedtype="image" format="MiddleAlignLarge" id="43034"/><p data-block-key="8r1o8"></p><p data-block-key="9udgi">Caltech <a href="http://smart.caltech.edu/">researchers</a> have developed mathematical tools that determine when to use and when to conserve power based on the energy available on the grid. These tools solve the problem of how to stabilize voltages on the grid even when energy from renewable sources fluctuates. This new approach can be applied to the distribution networks that take power from larger substations and deliver it to houses, buildings, streetlights, and other energy consumers in a region.</p><p data-block-key="b9jeo">With a smart scheduling-enabled grid, you could signal to your dishwasher whether you need dishes right away, or if the job could wait a few hours until the grid is less stressed. The enormous data centers run by Google, Amazon, and other companies could initiate energy-intensive activities, such as the archiving of data, only when solar and wind energy are available.</p><p data-block-key="91llo">Caltech already has launched a solution aimed at a significant consumer of energy: EVs. Steven Low, Frank J. Gilloon Professor of Computing and Mathematical Sciences and Electrical Engineering, former Caltech graduate student George Lee, and former <a href="https://resnick.caltech.edu/">Resnick Sustainability Institute</a> Graduate Research Fellow Zachary Lee invented the <a href="https://ev.caltech.edu/">Adaptive Charging Network</a>, a smartphone-enabled platform that parking garages can use to maximize the efficiency of EV charging stations. For example, one person may leave their EV parked for a full workday but need relatively little energy to top off the battery. That person could wait until late in the day, when energy is more available and cheaper, to charge. Another individual, who needs to charge before driving to a lunchtime meeting, could begin to receive electricity immediately. This technology, licensed through a company called PowerFlex, is now operating in Caltech's parking structures and is being deployed nationwide.</p><p data-block-key="9p6ca">But smart scheduling alone is not enough. A grid that runs on renewable energy also requires a major overhaul in how the grid is managed and regulated.</p><p data-block-key="a7865">Today's electrical grid relies on a centralized approach, meaning all data are collected and control decisions are made at a central management center to make sure enough electricity is directed to where it is needed. The smart grid, on the other hand, uses a decentralized approach, in which everything from dishwashers to data centers can interact with and glean information from the grid to optimize energy use without human intervention. This requires new algorithms designed with cybersecurity threats in mind from the outset.</p><p data-block-key="fq3su">Today's system also uses markets that predict customer electricity demand a day ahead of time, which allows power plants to generate and sell enough power to meet that demand. This market structure doesn't work when fluctuating renewables come into play. Caltech researchers, like Wierman, are doing theoretical work looking at the smart grid and the network of markets it will produce. One potential problem is that renewable-based markets open the door for companies to manipulate prices by turning off generators. Whereas the operational status of a normal generator can be monitored, solar and wind power make it nearly impossible to verify how much electricity should have been produced because it is difficult to know whether it was windy or sunny at a certain time. Wierman is creating market designs that would reduce that risk.</p><p data-block-key="8pv2t"></p><embed alt="caltech professor working with chemistry lab equipment" embedtype="image" format="MiddleAlignLarge" id="43036"/><p data-block-key="4rv9d"></p><p data-block-key="70lgg">Caltech's hometown of Pasadena is a convenient laboratory for the Institute's smart-grid researchers: The city has promoted the use of solar panels and EVs, providing a real-world, real-time example of the challenges that come with incorporating renewables into a conventional grid. Caltech is working with Pasadena Water and Power on a project to strategically install and program batteries that could store solar and wind energy for use when energy supply is low.</p><blockquote data-block-key="d0tgq">"Engineering grand challenges are inspiring much of the research at Caltech, especially in the broad area of sustainability," says Harry Atwater, Otis Booth Leadership Chair of the Division of Engineering and Applied Science, Howard Hughes Professor of Applied Physics and Materials Science, and director of the Liquid Sunlight Alliance. "Inventing and developing technologies to accelerate the transition to a sustainable energy infrastructure forms a large part of our research portfolio—our faculty, students, and alumni are at the forefront of solving large-scale problems that will have long-term impact."</blockquote><p data-block-key="r2rfo"></p><hr/><h3 data-block-key="5jeeg"><br/><br/>Interested in learning more? <a href="https://caltech.us5.list-manage.com/subscribe?u=eaa490ec46ef24d22d04308c6&id=36a33f2ada">Subscribe to Caltech Weekly</a></h3>