More than"AI Populism": Investigating Data Centers With K-16 Students

Written By Daniel Krutka

Genuine inquiry into AI's socioenvironmental impacts requires multidisciplinary thinking, critical education, and rigorous, community-led research.

 

Civics of Technology Announcements

Consider Commenting: The United States FCC is considering rules related to the E-Rate program, which provides funding to schools and libraries for internet connectivity. They are accepting public comments until October 13. Consider filing one here (the proceedings number is 26-133).

Book Clubs:

Book Reviews:

Merch: We still have Civics of Tech hats available! Order them here.

By Evan Shieh and Jacob Pleasants

The Data Center Backlash Isn't Just “AI Populism”

Do most people actually care about data centers, or just a passionate few? What’s motivating these revolts - is it about AI, or energy costs, or good old NIMBYism?

These are the central driving questions behind “No Data Centers In My Backyard”, a Substack article that recently went viral thanks in part to media coverage by The Ezra Klein Show and Plain English with Derek Thompson (among others). Authored by independent writer Jasmine Sun (who describes her work as an “anthropology of disruption”), the article focuses on several communities in Wisconsin. After speaking with organizers, politicians, and construction workers, Sun concludes that the rising opposition to US-based hyperscale data centers is explained more by process failures and weak messaging around AI’s benefits than it is by material, socioenvironmental concerns. The article ultimately attributes the root cause of data center backlash to “AI populism”, which Sun defines as “a worldview in which AI is viewed not only as a normal technology but as an elite political project to be resisted”.

Sun’s article motivated us to write today’s blog post. As sociotechnical researchers and educators, we resonate with the article’s goal to contextualize the issue of data centers by platforming diverse stakeholders. The socioenvironmental impacts of AI are not uniformly distributed by zip code, and data center development often exacerbates pre-existing issues ranging from regional droughts to industrial pollution and negative public health risks for residents living near hyperscale data centers. Any discourse that properly weighs the societal tradeoffs of AI development must therefore be grounded in locally lived realities.

Yet despite its stated goals, “No Data Centers In My Backyard” ultimately downplays the real environmental risks associated with data center development and undermines the very same local perspectives it claims to represent by incorrectly dismissing data center opposition as fundamentally non-technical in nature. In doing so, the article and its subsequent media attention provides an illustrative case study for the persistent shortcomings of technocracy and the “Abundance Agenda” to faithfully represent the material, socioenvironmental concerns motivating those resisting data centers. We argue that genuine inquiry into AI’s socioenvironmental impacts requires multidisciplinary thinking, critical education, and rigorous, community-led research.

Who Gets to be “Technical”?

Perhaps one of the most revealing passages in “No Data Centers In My Backyard” is when Sun attempts to characterize the opposition to data center construction as indifferent to the technical “facts of the matter.” She writes:

My conversations make me wonder if the technical content of data center deals is mostly beside the point. Closed-loop cooling systems? ‘I don’t believe them.’ Paying millions in taxes? ‘I don’t believe them.’ Creating a thousand jobs? ‘I don’t believe them.’ I hear a reflexive skepticism of every claim. (Emphasis added)

Indeed, Sun is rather dismissive of environmental concerns. She writes that “new closed-loop cooling systems use about as much water as a golf course”, and goes on to state that, aside from a few exceptions, 

...water consumption and air pollution [due to data centers] is generally minimal… these costs are not so different from other industrial developments like manufacturing or solar farms. Those don’t go up without dissent, but have not commanded the shared hatred that data centers have.

It is noteworthy for multiple reasons that the crux of the article rests on this particular argument. First, it rather ironically mischaracterizes the technical nature of data center water consumption by suggesting that closed-loop cooling systems have largely solved the problem of AI’s water usage. Newer data centers do indeed use a closed-loop system that recirculates the fluid that directly cools the GPUs. However, that fluid still needs to be cooled, and a common way that is done is through an evaporative cooling tower that, as the name implies, rejects heat through evaporation and thus consumes a considerable quantity of water. There are alternative approaches such as air-based cooling, but they require a greater draw of electricity, and producing electricity also consumes water. In fact, it turns out that much more water is required to generate the electricity to operate (and cool) the data center than is evaporated via cooling towers. And we haven’t even gotten to the enabled emissions due to the deployment or application of AI models.

Needless to say, this is a technically complex issue. The image of a “closed-loop” cooling system provides a convenient talking point for AI companies to misleadingly claim that they can operate data centers with as little as five drops of water per query, if not effectively “zero water consumption” (despite having been debunked, both of these statistics still circulate on social media and continue to be repeated by generative AI “search engines”). But this is the image that Sun invokes in the article.

But there is a larger issue at play here that goes beyond the technical weeds. What is notable is how Sun positions community members as technically uninformed, exhibiting a “reflexive skepticism” that is unwarranted and indifferent to “the facts.” It’s a rhetorical move that reveals a technocratic (and technosolutionist) ethos [1]. And it’s a move that is far from unique to “No Data Centers In My Backyard”. Dominant media coverage of the tech industry is infamous for its reliance on demonstrably pseudoscientific research, fancy accounting, and sensationalism that masks a lack of scientific rigor under the hollow aesthetics of mathematical terminology. In short, those who are positioned as technical “experts” can often get away with technically dubious claims, while legitimate concerns of “nonexperts” are brushed aside.

Navigating the Data Center Issue with Students

Our goal is not to simply present a takedown of Sun’s reporting. For us, her article serves more as a provocation for asking deeper questions about what inquiry into this topic should look like. As data center investments skyrocket, how can we as educators prepare our students and community members to conduct genuine inquiry on AI’s socioenvironmental impacts? Is it a hopelessly endless cycle of “your facts versus my facts”, as the framing of “AI populism” might suggest? Is the science so complex that it should only be left to doctoral researchers, if not tech industry insiders?

Both of us have recently wrestled with these questions as we taught about data centers. Below, we share our approaches and experiences with two very different groups of students.

Addressing Data Centers with Engineering Students (Jacob)

This summer, I led an investigation of data centers with a group of incoming first-year engineering students at the University of Oklahoma as part of a summer bridge program. I taught a class called “engineering for humans and the environment” that focused on how to promote public welfare as engineers. We worked through technoskeptical examinations of several technological systems, including data centers. A throughline was looking at these as sociotechnical systems and thus exploring their technical, psychosocial, and political aspects.

It turns out, our data center investigation began with more or less the same question that motivated Jasmine Sun’s piece: Why is there so much public resistance? However, we framed this question quite differently, which led to some very different directions for our inquiry. First, we took public resistance seriously as something worth studying. If you’re concerned with public welfare, public resistance is a sign that things are not going well! I gave my students a long list of news reports so that we could identify some common community concerns. Key issues that emerged included environmental impacts but also concerns about transparency, economic impacts, and governance.

Next, though, I posed another question that also parallels Jasmine Sun’s piece: data centers have existed for some time, and the concerns raised about them could also be leveled at quite a few industrial operations, so why the backlash, and why now? Rather than dismiss this backlash as irrational and misinformed, I instead framed this as a case where data center developers have managed to lose their social license to operate. Yes, there are many extractive industries out there, many of which have far greater environmental harms than data centers. But there are reasons why they have managed to maintain public support (not always the best reasons, but reasons nevertheless). And there are also instructive examples where trust has been lost. So, our inquiry led us to look at why this has occurred for data centers, and what would need to change to earn it back.

Public concerns about data centers are myriad, but we started with a few technical problems to investigate further: electricity use and water consumption. My students were rather astonished by the prodigious energy demands of new, large-scale data centers. For what purpose is a GigaWatt-scale data center needed? AI is the obvious answer, of course, so we looked at the energy (and water) consumption from typical uses of AI chatbots. There is a lot of misinformation out there about this, and the best estimates establish that it’s pretty modest. This presents a puzzle: if those massive data centers exist to provide infrastructure for AI, it’s clearly not merely to power your typical consumer-facing chatbot. What, then, is all that computing being used for? [2]

We then got into the environmental impacts. For data centers, you have thermal pollution, noise pollution, and water issues. All of those are significant, but again I pointed out that these are present for essentially all heavy industries and infrastructure; data centers, even large ones, are not uniquely or even especially harmful in terms of their environmental impacts (unlike many other industries, for instance, they are not discharging chemical pollutants into the water supply). I walked students through historical examples of extractive industries, superfund sites, and the complex relationships that local communities have had with nearby highly-polluting industries. Community support has often been rooted in economic investments (though many relationships have been exploitative) and beliefs that the industry is serving a meaningful purpose (e.g., mining strategic minerals for a war effort [3] or providing essential infrastructure). But that support can be tenuous, and there is an ongoing tension between potential short-term gains and long-term environmental costs that will be borne by residents for generations.

For data centers, a core insight is that resistance is not simply about the environmental harms. It’s that the community generally gets extraordinarily little back from a data center project. Aside from short-term construction jobs, hardly anyone is employed by a data center. And while there is the possibility that property taxes could support local government budgets, that economic relationship is not always so simple. It’s also hard to see a data center as “essential infrastructure” that serves a meaningful purpose. Unlike a power plant or aluminum smelter, the infrastructural work that a data center is doing is not only ephemeral, but perhaps even superfluous. 

Imagining Alternatives

With this understanding in hand, I gave my students a rather challenging task. I gave them the role of “engineering consultants” who have been retained by a data center developer who wants to site a new data center in [pick a city]. Their job was to put together a 1-page set of recommendations that would promote public trust, community support, and public welfare more generally. Those recommendations needed to include technical specifications (where it should be cited, how it should be powered, etc.) as well as community engagement and communication strategies. Students recognized that this was a tall order! But they came up with some interesting ideas:

  • Transparency: Be clear about who/what the data center is for, and include financial disclosures. 

  • Don’t Waste Heat: Use “waste” heat for electricity production or home heating.

  • Don’t Waste Components: Find ways of reusing the electronics that are cycled out of data centers (ideally returning them to the community). Given the “RAMpocalypse” going on, this could be quite valuable!

  • Invest in Community: Allocate some budget for community improvements and infrastructure. Provide computing resources to the community as well.

  • Use Sustainable Power: Solar, wind, geothermal, hydro (depending on the context).

  • Sustainable Water Use: Avoid evaporative cooling altogether if possible. Look for ways to collect evaporated water. Treat discharged water.

  • Communicate: Town halls, informational meetings, online engagement, mailings.

This list, of course, doesn’t address some of the broader economic and political issues raised during our investigation (and that appear in Sun’s reporting as well). But coming from an engineering perspective, it’s a good start, and would definitely be a change from business as it is currently done.

Reflections from K-12 Educators (Evan)

As Jacob’s experience demonstrates, empowering students to study the environmental engineering impacts of AI and data centers requires paying attention to the material interests of various stakeholders within local communities. From my experiences teaching critical AI literacy, I’ve learned that introducing complex sociotechnical issues in K-12 classrooms presents a crucial opportunity to broaden participation among students who have been traditionally minoritized in both STEM classrooms and industries. In support of that goal, below are three lessons I’ve learned from working with YDSL’s K-12 teacher fellows:

Lesson #1: Honor Non-Dominant Ways of Knowing

Our students stand to bear the brunt of climate change’s projected global impacts. Unsurprisingly, youth in our current generation have responded by taking more supportive action towards climate justice compared to their counterparts in any previous generation. It is therefore our imperative as K-12 educators to empower our youngest learners with the tools they’ll need to respond to socioenvironmental issues like data centers and AI.

This may seem daunting to introduce at the grade school level, but Oakland-based educator Lyndsay Schaeffer taught us that upper elementary school students are well-poised to navigate the topic of AI ethics and data centers. The key, according to Lyndsay, is to pay attention to the tangible, everyday realities that students are already observing in order to position them as participatory stakeholders on complex sociotechnical issues.

Doing this effectively requires educators to honor non-dominant ways of knowing, particularly for communities of color, whose informal knowledge is disproportionately dismissed in traditional science paradigms. Meaningful student learning happens outside of the classroom - at home, in community groups, or in the neighborhood. For example, Allentown-based civics teacher Shannon Salter shared with us a story about a student living near a data center, who noticed that the trees on the surrounding block were the first to turn color during the Fall. Remarkably, her observation was made nearly a year before mainstream media and research universities reported on the phenomenon of data center heat islands. Unique insights like these are lost when dismissing communal knowledge as “non-technical” or mundane.

Lesson #2: Elevate the Humanities

Conducting rigorous inquiry into socioenvironmental issues rarely presents clean answers, as students must endeavor to process the uncertainties, values, cultures, and histories of their local communities. Perhaps for this very reason, we found that some of the teachers in our program who were best-equipped to teach these topics came from the humanities. English teacher Karren Boatner sees misinformation about AI and the environment as an organic opportunity to teach her students the importance of critical information literacy. For their final project, Karren’s high school students set out to research the positive and negative impacts of AI and data centers on the environment in their hometown of Chicago. Karren found the activity to be a valuable way to engage students of all backgrounds, including those who may have shied away from STEM previously.

Lesson #3: Practice Systems Thinking and Intersectional Analysis

Data centers exemplify how AI’s impacts are not distributed uniformly across society. The local impacts of data centers cannot be extricated from the applications of AI driving data center demand. Among these, supercharging fossil fuel extraction and AI-enabledwarfare are some of the worst sources of modern socioenvironmental harm. Today’s students are keenly aware of the relationship between their individual well-being and broader sociopolitical issues, and the movement to place a moratorium on AI usage in schools may be growing as a result.

Preparing our students to address systemic issues is not an individual endeavor. Civics of Technology and YDSL partnered together this past year to teach New York City Public Schools’ Exploring Equity in AI (EEAI) program. Tellingly, the program drew interest from over 130 teachers across all disciplines, boroughs, and grade bands who produced a diverse array of critical AI lessons and policy proposals spanning a variety of topics. However, as NYCPS program director Christy Crawford informed us, environmental considerations are still largely ignored by most AI literacy programs that primarily focus on narrower “technical” issue.

As educators, we have a responsibility to prepare our students not just as future workers, but as collectively informed owners of civil society. Meeting this goal requires us to embrace sociotechnical complexity and communal lived realities as we challenge our students to build towards a more just society.

[1] Also tellingly, Sun chooses to use golf courses and manufacturing as examples of entities whose environmental impacts are comparable to data centers. Rhetorically, she does this to provide evidence that the impact of data centers is minimal. But this choice feels rather out of touch with the evergreen challenges of drought-prone regions or the intergenerational, localized impacts of heavy industry. Golf courses are a particularly interesting point of comparison. Their water usage may very well be on par (forgive the pun) with data centers. But that might be more of an indictment of golf courses than an acquittal of data centers!

[2] I was content to simply raise the question with my students, but investigating it further could be very worthwhile. While consumer-facing chatbots are most students’ point of entry into AI, it is clear that the bulk of the AI investment is aimed elsewhere (likely enterprise and government contracts).

[3] Being in Oklahoma, I used the specific example of the Tri-state Mining operation and the resultant superfund site. The mine extracted ores that were used during World War II operations and local communities take pride in that history. But they also must live with the despoliation caused by the mines.

Daniel Krutka
Next

Designing the Social World Around AI: Latino Youth Setting the Terms