Adding Equity to Environmental Models, with Amanda Giang

Notable Quotes

• Models that leave out equity may undermine policy goals: “We use models often to inform our sustainability interventions, but when we’re not taking equity into account, we can actually end up working against our sustainability goals.” (5:53)
• Setting goals for a model informs the results: “Defining the purpose, figuring out who’s informing that purpose … is such an important point for procedural and recognitional equity. How you design the whole modeling endeavor has huge implications for the sorts of results you get—even from just starting with the thought of, We are interested in equity as an important outcome.” (9:13)
• Models are just one piece in the policymaking puzzle: “One of the challenges is always trying to [convey the] message that the outputs of these models should not be the decisions in and of themselves. We’re hoping that they are inputs to a broader deliberative process where we can talk about some of these complex trade-offs.” (22:00)

Top of the Stack

• “Equity and Modeling in Sustainability Science: Examples and Opportunities Throughout the Process” by Amanda Giang, Morgan R. Edwards, Sarah M. Fletcher, Rivkah Gardner-Frolick, Rowenna Gryba, Jean-Denis Mathias, Camille Venier-Cambron, John M. Anderies, Emily Berglund, Sanya Carley, Jacob Shimkus Erickson, Emily Grubert, Antonia Hadjimichael, Jason Hill, Erin Mayfield, Destenie Nock, Kimberly Kivvaq Pikok, Rebecca K. Saari, Mateo Samudio Lezcano, Afreen Siddiqi, Jennifer B. Skerker, and Christopher W. Tessum
• Ducks: Two Years in the Oil Sands by Kate Beaton

The Full Transcript

Daniel Raimi: Hello, and welcome to Resources Radio, a weekly podcast from Resources for the Future. I'm your host, Daniel Raimi. Today, we talk with Dr. Amanda Giang, an assistant professor in the Department of Mechanical Engineering at the University of British Columbia.

Amanda is the lead author on a really cool recent paper called “Equity and Modeling in Sustainability Science: Examples and Opportunities Throughout the Process.” I'll ask Amanda to describe whether and to what extent computational models have sought to incorporate equity considerations when analyzing complex systems like energy, water, the economy, and the climate. Then, we'll talk about the ways that researchers can start building equity into their models—starting from the conceptualization stage all the way through presentation of results. It's a wonky topic, but it's super important for all of us in the research and policy community, and Amanda explains these issues extremely clearly. Stay with us.

All right, Amanda Giang from the University of British Columbia, welcome to Resources Radio.

Amanda Giang: Thanks so much for having me, Daniel. I'm a big fan.

Daniel Raimi: Thank you. It's great to have you on the show. As you know, since you're a big fan, we always ask people at the beginning of the show how they got interested in environmental topics—whether it started with early life inspiration or whether you came to it later in your career. Can you tell us about that?

Amanda Giang: Yes. For me, I actually think the entry point was my interest in health. I wanted to be a doctor when I was younger. I think Grey's Anatomy had just come out and deeply influenced me. I think a turning point in how I thought about health was actually when I learned about persistent organic pollutants, or “forever chemicals,” produced by human activity. They move all around the world, they're often stored in our bodies, and they can impact our health through generations, because they can accumulate in our bodies.

I think learning about this really helped me have a shift in how I saw a couple of things. The first was just how the environment is such a critical determinant of human health and well-being, and the second was just that this divide between human and environment is kind of artificial. The environment isn't something out there. We are part of it; we embody it; and we're the air we breathe, the water we drink, and the foods we eat. I think that really motivated me to want to work on environmental issues as a way to work toward improved health and well-being and to address health disparities, in particular.

Daniel Raimi: Yes, that's so cool. And it ties into the topic of our conversation today, which is models, which maybe sounds like it doesn't tie in, but it does, because it's the combination of human and natural systems. Modeling is what we're going to talk about today.

Amanda Giang: Yes, absolutely. I'm really interested in that intersection.

Daniel Raimi: As most of our listeners know, computational models are used in all sorts of ways to inform decisionmaking across lots of systems—like energy systems, climate systems, or water systems—where humans are interacting with this mix of natural and person-made systems. The recent article that you published with lots of great coauthors addresses whether and to what extent those models incorporate issues of equity. We're going to talk about that today. Can you start us off first, though, by just defining that term, “equity,” in this context?

Amanda Giang: Yes. I'll just say that—me and my many coauthors, in our definitions—we're really drawing on the work of a huge community of scholars, activists, and practitioners who work on environmental justice issues, so these certainly aren't just our definitions. In this article, we talk about equity as a really multidimensional concept around fairness. Of course, fairness is hugely context-dependent, and it evolves, because what's fair in one place at one time might not be considered fair in another. There are also a lot of different dimensions of equity.

So, you can think, “Fair in what?” We focus specifically on three dimensions. The first one is—one that I think is really the one that's most commonly talked about—distributional equity. Here, we're thinking about how different benefits and burdens are shared or distributed, and that might be across space, across time, or across different social groups.

The second dimension we focus on is procedural equity, and this is really about who gets to meaningfully participate in decisions that affect them.

And then, the last one that we talk about is recognitional equity. For us, this is really about recognizing and respecting different identities, histories, worldviews, and different ways of knowing.

These three dimensions—they're not completely separate. They can really be mutually reinforcing. For instance, we might address distributional equity issues by enhancing procedural and recognitional equity. It's really kind of like an interdependent mix.

Daniel Raimi: Yes. That makes a lot of sense. Again, as our listeners know, computational models that are out there—one of the things that they have to do is they have to flatten complexity and represent some kind of system in a simplified way. As you and your coauthors note, because of that—or well, you can tell me if you think that it's because of that—but historically, many models have largely ignored these equity considerations. I'm hoping you can just help motivate us a little bit by sharing how that lack of incorporation of equity issues in models affects modeling results and how that in turn affects decisionmakers that are relying on models.

Amanda Giang: Yes, as you mentioned, we use models often to inform our sustainability interventions, but when we're not taking equity into account, we can actually end up working against our sustainability goals. As you said, models, by nature, need to kind of flatten the real world, otherwise it wouldn't be the model; it would just be the world.

So, of course we focus on different things at different times, but I think, historically, we haven't always called out when equity wasn't a primary consideration, and that can lead to unsustainable outcomes. One example is, if we think about some of the first generations of models used to inform climate policy, often we were modeling climate drivers and impacts at this kind of aggregated global level. And we would think about—okay, well, how can we optimize utility in this aggregate world across time as we think about it as a single region?

A lot of this had to be done computationally; at that time, we couldn't really represent different parts of the world as easily. But of course, if you think about that, things can be getting better overall in aggregate but in very unequal ways. Maybe things are getting better in aggregate, mostly because things are getting way better for one group and things are not getting better or maybe getting worse for many others. From a policy perspective, that might lead to conclusions that actually work against you, because we know income inequality, for instance, can also be a driver of unsustainable behavior. So, I think this is just recognizing that while not all models have to focus on this—certainly when we're thinking about how we design technology policy and other interventions—it is really important to at least think about what we're missing when we're not including that.

Daniel Raimi: Yes, that's really well said. I'm imagining that you're thinking of the famous DICE model that Professor William Nordhaus pioneered, the Dynamic Integrated Climate-Economy model, and it was really first of its kind and had really powerful and impactful concepts that it brought to bear. Again, it did things at a very coarse level and has been improved upon dramatically by other modelers over the years. Was I right that you were thinking of the Dynamic Integrated Climate-Economy model?

Amanda Giang: Yes. I think that's such a critical and transformative starting point for that. Over time, the Dynamic Integrated Climate-Economy model has also evolved to capture more of these elements.

Daniel Raimi: For sure. One of the really great things about this article is that it really provides some helpful practical steps for modelers, and a lot of the article is devoted to examples of how modelers can bring equity considerations into their models through four stages of the modeling process that are kind of central to the paper and illustrated really well in Figure 1 of the paper. Of course, we'll have a link to the paper in the show notes so people can go check it out if they'd like as they're listening.

I'd love for us to just talk through those four stages now. Can you get us started by helping us understand how equity can be accounted for when modelers are just getting started and defining the purpose of their model and trying to figure out who participates in defining that purpose?

Amanda Giang: Yeah. I think this upstream stage—this beginning of defining the purpose, figuring out who's informing that purpose—is such an important point for procedural and recognitional equity. How you design the whole modeling endeavor has huge implications for the sorts of results you get—even from just starting with the thought of, We are interested in equity as an important outcome, for instance, that we might want to model. One example I'll give comes around marine conservation and bringing attention to how important it is, at these really early stages, to think about the diversity of groups and different ways of knowing that are informing even the modeling questions and the modeling design.

One of the examples that we talk about is the coproduction of knowledge between Indigenous knowledge-holders and Western scientists in terms of informing bowhead whale–population modeling and wildlife management, because at one time, these models were just informed just by Western science, and they led to some conclusions in terms of how the population stock was changing over time and a ban on all subsistence hunting that really would've jeopardized Alaskan Inuit food security, well-being, and community health. But through advocacy and from Alaskan Inuit groups, there was a move toward, How do we actually bring in these different knowledge systems together for that population-stock modeling and recognize that these different knowledge systems might have access to different understandings about how bowhead whales exist in the environment and what data make sense to inform that modeling. And that actually led to different conclusions that then had really big impacts both for bowhead whales and Indigenous communities.

Daniel Raimi: That is such a great example, and I can't resist sharing this fun fact, which is that I've actually been to what was then called Barrow, what's now called Utqiagvik, Alaska. It's the northernmost city in Alaska, and it's right on the Arctic Ocean. This is one community where bowhead whale hunting for subsistence purposes is really important. I got to talk to some people about that while I was up there. There's actually beautiful—the jawbones of bowhead whales are sort of displayed in some public settings, and I was even able to buy a bowhead whale's tooth, a large piece of baleen that is now framed on my wall. I love that example. It's my favorite party thing to show people.

Okay, back to the substance of our conversation. The second issue that you and your coauthors highlight is how modelers can incorporate equity considerations when they're deciding what to model and the appropriate scale of modeling. Can you talk a little bit about that?

Amanda Giang: Yes, I think questions around scale and resolution are particularly important for capturing these distributional aspects. In the example we gave earlier, I talked about some of that global integrated assessment modeling. I think one of the areas where there have been some really interesting results around this is around air-pollution exposure. Work has shown that, when we use coarser resolutions, we're often actually underestimating exposure disparities. To be able to capture that, we actually need to be able to model at finer spatial and temporal scales. That also kind of creates a technical challenge of, How do we do that in a computationally feasible way to be able to capture these really critical exposure disparities? Resolution, I guess, though it's not just about geographic areas. I think there's been a lot of really interesting work that we talk about in the paper—particularly in energy-systems modeling and water-systems modeling—that's really bringing attention to social scale.

Even if we're looking at a small geographic area that's maybe not really capturing the diversity of households, different kinds of constraints that might exist in terms of different kinds of households, or different kinds of agents in that system … Really being able to capture that is really important in terms of understanding, let's say, for instance, technology adoption, when we're talking about, let's say, low-carbon energy technologies that can potentially have a lot of benefit for households economically. Then, let's also say, from a decarbonization perspective, it is really important to actually capture, even within a small unit of space, how different households might vary in terms of their ability to actually access those benefits. I think this has really moved; we can see this in a movement toward agent-based modeling, where we can actually represent a diversity of agents rather than just one single representative household, for instance. And a lot of this has been enabled by the fact that we have more granular data, now, at smaller social scales.

Daniel Raimi: Yeah, more granular data and more computing power, which I was sort of thinking, as you were answering that question—when we think about addressing equity issues from a modeling perspective, does it always add complexity to the model, or are there examples where it could actually simplify things? I'm just thinking this question is important a lot of times, because as models get more and more complicated, it gets harder and harder to understand what's driving some important results. Is that a consideration that you and your coauthors have thought about?

Amanda Giang: Yeah, I think this is a really important question. It's not always the case that more is always better. Like I said, it wouldn't be a model if we were trying to include everything. I do think this question of trade-offs when we're trying to increase scale is something that, for instance, as I mentioned in the air-pollution space, people have been looking at. What is that sweet spot of where we get improvements as we increase the spatial scale, but the trade-offs that we get, let's say, in terms of the amount of computing time we need or the complexity of the models, then kind of makes that less worth it, because we're capturing … the improvement in our ability to capture the patterns drops off; there's that knee in the curve.

I think the other thing, too, is, for instance, when we're doing something modeling where we have better representation of different agents, then that might mean that we need to trade off in terms of the geographic scope of what we're representing. We can't do everything, so thinking really strategically about what we focus on in any given modeling exercise, and then also how we can bring different models together that get at these different aspects to paint a fuller picture of this really complex system, might be another way to get at the reality of those trade-offs.

Daniel Raimi: Yeah, that's great. The third point that you and your coauthors focus on for modelers is giving a sense of how modelers can incorporate equity considerations when they're thinking about the interactions that are going on within their model. This sort of gets to the point that we were just talking about, but can you say more about that?

Amanda Giang: Yeah, when we're talking about interactions, kind of interactions across space and also across time ... Maybe I'll focus on that across-time piece, because I think that's one of the ones where we sometimes struggle. We are trying to understand the dynamics of systems, and equity is not a static thing. As systems evolve, we might make an intervention in a system to address, let's say, some equity concerns, but there might be different adaptive responses and feedbacks in this complex system that might actually either serve to widen some distributional equities or actually help us reduce them. So, capturing that, we have to think carefully about how we can actually capture some of those complex adaptive dynamics. One example that comes up that we talk about is in water-distribution systems and thinking about some of the interactions that you might see in terms of adaptive responses, let's say, to a poor water-quality event.

Some of the adaptive responses that we see for the water-distribution network actually aren't on the water-distribution network, because people, as an adaptive response, will start purchasing bottled water. Of course, that has economic implications for households, as well, particularly when it is lower-income households that are already impacted by poor water quality, let's say, due to aging infrastructure. I think those are some of the kinds of interactions that modelers are starting to try to be able to capture and represent in these human-nature models.

Daniel Raimi: That makes tons of sense. The fourth point that the paper highlights for modelers is, How can equity considerations inform them, when they're thinking about the policy or other types of interventions that are designed to address the issues that are modeled in the system? Can you talk about that?

Amanda Giang: Yeah. One of the points that I think came up across a lot of the examples that we discuss and look at is the reality that equity is not a singular concept. As we started talking at the beginning, equity is really multidimensional. There are different dimensions of it, and it's also not a binary. It can be a continuum in these different dimensions.

When we're thinking about the sorts of equity-related metrics and indicators that we're getting from models, we need to think a little bit more about how we can use suites of metrics that can speak to synergies and trade-offs and these continuums for these different aspects of equity. It's not only a model-design question, but it is also a question of how we visualize and interpret and share these results. And it's tough. Often, let's say, model-output users would like a single metric, so I think this is a place where we definitely need a lot of innovation and work in terms of, How do we make sense of these complex multidimensional equity indicators?

Daniel Raimi: That last point is such an interesting one, and I was thinking about it as you responded. Think about things like the social cost of carbon. Policymakers often want to know what the social cost of carbon is. There are models that will give you an estimate, but the range of projections is very wide, and there's a lot of uncertainty embedded in the modeling systems. When you interact with policymakers or decisionmakers, how do you communicate these complexities to them and what are the challenges that you found?

Amanda Giang: Yes, I mean, one of them is having to disappoint them and saying, “Well, it's complicated, or we have to look at this range.” And I think we've explored a few different methods of doing this, sometimes through interactive visualization platforms, so that at least instead of giving one number, it's like, let's give you something where you can explore these different layers, dimensions, and interactions. Sometimes, we try to design new composite indicators that might bring together different aspects of these different dimensions, and I think that we see that in a lot of areas. For instance, the US Environmental Protection Agency has this environmental justice screening kind of platform, and they have some of these sorts of composite indicators, and that can be really powerful, but it can also be flattening.

I think some of the dangers in that is that, if we freeze and aren't continuously asking questions and making sure that lots of different groups have the ability to ask questions about, Why did you combine those things in these different ways? Or maybe we want to reweight how these things are combined, or maybe we shouldn't just be adding these things together, but we should be multiplying these things together. I think one of the challenges is always trying to message that the outputs of these models should not be the decisions in and of themselves.

We're hoping that they are inputs to a broader deliberative process where we can talk about some of these complex trade-offs in ways that aren't just like, “This number is larger than that number, so we should go with that number.”

Daniel Raimi: Yeah. It reminds me of a really wonderful sentence in the paper, which is, “Models are tools, not solutions,” which is an ethos that I've heard many times, but I've never seen it expressed so concisely. So, kudos on that phrase.

I was actually thinking another very challenging question is, When do you not use a model, even if you could? Are there times when using models and having numbers actually confuses or stymies the decisionmaking process, when the decisionmaking process maybe shouldn't rely on a number at all? Maybe it should rely on some kind of moral consideration, or some kind of other consideration. Is that something that you and your coauthors have sort of thought and talked about?

Amanda Giang: Yeah. First of all, for that line, I just want to credit Emily Grubert.

Daniel Raimi: Nice.

Amanda Giang: [Emily is] really good at summarizing things succinctly and powerfully. This is something that we talk about—that there might be times where we know in our society there's such an overemphasis on quantitative modeling approaches in general, because again, those numbers are powerful, and there's a lot of cognitive authority from quantitative approaches, in particular. But there are times where maybe in the process of doing that, we're obscuring when we know that we're not capturing really important dynamics in those models, or there are, on that recognitional side, things that are not being recognized in there.

Then, maybe by focusing on that output, we're actually kind of taking space away from other voices or other considerations, when we know that our representations are really limited. I think there are lots of times where conceptual modeling or qualitative analysis might be more appropriate. I think, as quantitative modelers, in particular, this really does require us to also have reflexivity and humility about what the limits of our tools are and try to make more space for those who come from other knowledge traditions or have other types of insights to share.

Daniel Raimi: That's really well said. The last thing on this topic that comes to mind is that there are structural reasons why it's hard for modelers to do that—you need to publish papers, you need to try to get tenure, you need to have results that people care about. Sometimes, the structural incentives push against that humility that might be needed sometimes. The last question that I'd love to ask you is about the future. The paper discusses a variety of opportunities to continue improving and enhancing equity considerations in models going forward. Can you talk about a couple of those?

Amanda Giang: Yeah, there's so many. I'm trying to think about which ones to focus on. One I think that is really important is around data governance, because earlier, I mentioned how a lot of our improved ability to maybe capture some of the distributional equity aspects has come from more granular data. That's really powerful, but we also have to be kind of attentive to some of the tensions that can come out with more granular data in terms of questions around privacy, let's say, where that more granular data might actually lead to adverse impacts. For instance, does data on household-level income and expenditures—could it be used in adverse ways to determine who to give loans to and at what interest rates? That actually would work against our goals in using that data.

So, I think that there are a couple of acronyms that we hear when people talk about data governance, and FAIR is one really important one—findable, accessible, interoperable, reusable data principles—and that can really enable more people to participate in this process. But we need to pair that with considerations of the social context of the data, as well. One example of this comes from the Global Indigenous Data Alliance. They've got this CARE framework. So in addition to FAIR, think about the collective benefit, authority to control, responsibility, and ethics around data. So, data is so powerful, but also there needs to be a lot of critical discussions about appropriate uses of data and who gets to control that data.

Daniel Raimi: Yes, that's a great example. It makes me think of a professor from Arizona State University who I saw talk recently. Her name is Krystal Tsosie. She's a Diné person, studies genetics and bioethics, and has done a lot of work on this data-governance question. Super interesting.

Well, Amanda Giang from University of British Columbia, this has been fascinating. I would love to talk for longer, but we're reaching the end of our allotted time. I'd love to ask you now the same question that we ask all of our guests at the end of the show—to recommend something that you think is great. It can be something to read, watch, or listen to. Any medium is acceptable, and it can be related to the environment, but we're actually not that picky about that. So what's on the top of your literal or your metaphorical reading stack?

Amanda Giang: This is literally on the top of the stack of the books that is holding up my microphone, so it’s top of mind. It’s Ducks. It's a graphic novel memoir by Kate Beaton, who's a Canadian graphic novelist, and it's about her time working in the oil sands as a young woman in the late 2000s, trying to pay off her student loans. I think it really gets at both the drivers and impacts of extractive industries socially and environmentally in a way that's both really grounded in the deeply personal, but also gets at these larger systems and structures. It's just really the combination of image and word I think is really powerful.

Daniel Raimi: Yeah, that looks so cool. I've never heard of this: Ducks: Two Years in the Oil Sands by Kate Beaton. Wow. I've got to check this out. We'll have a link to it in the show notes, so other people can check it out, too. That's a great recommendation.

Amanda Giang, one more time, thank you so much for coming on the show. Congratulations on the paper, and thank you so much for sharing it with our audience. We really appreciate it.

Amanda Giang: Thank you.

Daniel Raimi: You’ve been listening to Resources Radio, a podcast from Resources for the Future (RFF). If you have a minute, we’d really appreciate you leaving us a rating or comment on your podcast platform of choice. Also, feel free to send us your suggestions for future episodes.

This podcast is made possible with the generous financial support of our listeners. You can help us continue producing these kinds of discussions on the topics that you care about by making a donation to Resources for the Future online at rff.org/donate.

RFF is an independent, nonprofit research institution in Washington, DC. Our mission is to improve our environmental, energy, and natural resource decisions through impartial economic research and policy engagement. The views expressed on this podcast are solely those of the podcast guests and may differ from those of RFF experts, its officers, or its directors. RFF does not take positions on specific legislative proposals.

Resources Radio is produced by Elizabeth Wason, with music by me, Daniel Raimi. Join us next week for another episode.