The future of NEPA
Why NEPA was not part of the plot in Mad Max: Fury Road
The National Environmental Policy Act, better known as NEPA, requires federal agencies to prepare a “detailed statement” in connection with all proposals for “major Federal actions significantly affecting the quality of the human environment.” This “detailed statement” must address the “reasonably foreseeable environmental effects of the proposed agency action” and “a reasonable range of alternatives to the proposed agency action.”
When NEPA was first enacted in 1970, no one thought it would be a big deal. It has turned out to be a big deal. NEPA-mandated environmental impact statements (referred to as EISs) often run hundreds of pages and take years to compile, leading to pipelines, solar energy facilities, and other projects being delayed for years. After a project is approved, opponents can file lawsuits alleging that the EIS isn’t detailed enough, yielding further delays. NEPA’s detractors characterize the law as a millstone around the neck of economic and technological progress.
NEPA law continues to develop. On December 10, 2024, the Supreme Court will hear oral argument in Seven County Infrastructure Coalition v. Eagle County, a case addressing the scope of an agency’s NEPA obligations. Also, on November 12, 2024, the D.C. Circuit decided Marin Audubon Society v. FAA, an important but confusing decision establishing that certain NEPA regulations aren’t binding.
But the truly significant NEPA-related development is the advent of AI. I’m sorry if my AI posts are getting repetitive, but I don’t see how one can have an intelligent discussion about NEPA without noticing that a technology has fallen into our laps that can reduce the time to create written reports by 99.99%.
In this post I will:
Offer an executive summary of why NEPA is controversial.
Address what’s at stake in the Eagle County and Marin Audubon Society cases.
Explain why the “AI Changes Everything” take, while clichéd and annoying with respect to many fields, is correct with respect to NEPA.
Trains vs. sage-grouse
The Seven County Infrastructure Coalition wants to build an 85-mile railway line in an area of northeastern Utah known as the Uinta Basin. The primary freight would be crude oil.
Building this line requires a federal license. The Coalition applied for such a license from the federal Surface Transportation Board. Under NEPA, the Board was required to prepare an EIS.
The Board did that, and it approved the project. But opponents of the railway line—environmental groups and a local government in Colorado—sued the Board, alleging the EIS was flawed. This dispute has now reached the Supreme Court.
Eyeballing the EIS at issue should give you a sense of why NEPA is so controversial. It’s a spectacular document, spanning 3,600 pages. Here’s a passage from the document, pulled at random:
Any of the Action Alternatives would cross greater sage-grouse habitat, including breeding, nesting, brood-rearing, and wintering habitat, and would result in the permanent removal of and temporary disturbance to that habitat (Table 3.4-12 and Table 3.4-13). Disturbed areas in the temporary footprint would be reclaimed and revegetated following construction; however, affected sagebrush habitat in the temporary footprint would take many years to be restored to pre-construction conditions due to the difficulty in reestablishing this type of habitat (Meyer 1992). Greater sage-grouse could also be killed or injured by collisions with construction equipment, workers’ vehicles, and project-related infrastructure (fences and communications towers). Noise from construction equipment and the presence of people in construction areas could displace greater sage-grouse and cause them to disperse into habitat areas further away from the rail line (Appendix J, Bureau of Land Management Greater Sage-Grouse Resource Management Plan Compliance). There are also several greater sage-grouse leks in the vicinity of all three Action Alternatives within the Carbon SGMA (Figure 3.4-1). The habitat removal and noise associated with construction of the proposed rail line could cause greater sage-grouse to avoid or abandon those leks, especially if construction were to take place during the breeding season.
In case you were wondering, Claude advises that “a sage-grouse lek is a traditional courtship display ground where male sage-grouse gather each spring to perform elaborate mating rituals to attract females.”
Let’s start with a thought experiment. Suppose this EIS could be prepared instantly and at no cost. Would we want it?
Sure. Why not? It has two primary benefits:
It educates the public. If you have a deep interest in the environmental impact of railways in northeastern Utah, then boy, seeing this document must feel like Christmas morning. I’m not trying to make fun of anyone! People with obsessive, esoteric interests are my people.
It forces the agency to consider the environmental impacts of its actions. Personally, I find it hard to get worked up about the sage-grouse leks. Even if their lek happens to fall on the railway line, can’t they find another lek? On the other hand, if there’s some way to build the railway line in a way that helps out the sage-grouse, we should do it. In a world with no trade-offs, it’s better to at least think about sage-grouse mating rituals than to build the railway line, accidentally destroy the leks, and then say, oops.
But trade-offs exist. It took more than two years to get from the Notice of Intent to prepare the EIS to the final EIS, and that is faster than usual. According to a 2018 federal report, the average time from Notice of Intent to final EIS was four and a half years, and a fourth of EISs took six years or more. Six years! Congress recently enacted legislation that tries to improve this situation by imposing deadlines and page limits, but I’ll believe it when I see it.
This might not be problematic if the project could proceed while the EIS is being prepared, but it doesn’t work that way: because the agency is supposed to use the EIS to decide whether to approve the project, the project can’t proceed until the EIS is done.
As a result, NEPA has several drawbacks.
4.5 years is way too long to be pondering whether a new project is worth the environmental harm. Life is short.
NEPA deters worthwhile projects from being built. It’s hard to raise money for a project when construction won’t begin for half a decade or longer.
Even if the project is ultimately built, delay is still bad. If a project is worthwhile, we’re better with it now than five years from now.
More fundamentally, NEPA reflects a policy of status quo bias. You never have to prepare a 3,600-page document assessing the benefits and risks of doing nothing.
Sometimes federal projects, or projects that require federal licenses, make the environment better. Think of solar or wind projects, or Forest Service initiatives that reduce the risk of forest fires. Deterring or delaying these projects in the name of environmental protection might make the environment worse.
NEPA disproportionately affects clean energy projects. The oil and gas industry persuaded Congress to enact various exemptions from NEPA for oil and gas projects. No such exemptions exist for solar and wind projects.
EISs cost millions of dollars.
There’s also the opportunity cost to consider. The Eagle County EIS is amazingly thorough. It is obvious from reading it that the people who created it are competent, conscientious public servants. Can’t they spend their time on other things?
NEPA has drawbacks, but what about its benefits? Well, one important thing to remember is that NEPA—or at least the portion of NEPA that is judicially enforceable—is a purely procedural law. NEPA is not like the Clean Water Act or the Endangered Species Act, which enact substantive environmental protections. It merely requires the agency to write things down. If an agency wants to turn Yellowstone National Park into a nuclear waste dump, and it explains this decision thoroughly, NEPA will not stand in the way.
So NEPA doesn’t, or at least doesn’t directly, protect the environment. Instead, NEPA’s ostensible goals are to educate the public and promote reasoned decision-making. But it’s not clear that NEPA is effective at accomplishing those goals.
Again: look at the Eagle County EIS. Although some people might be interested in particular aspects of this document, approximately no one will read it from cover to cover. A document no one will read shouldn’t be written.
There’s no way the agency decision-makers will read this entire document. They’ll read a five-page executive summary. The full document isn’t helping them make decisions.
I understand the theory that NEPA serves a kind of project management function—the best way to ensure the agency makes good decisions is to force the agency to write everything down. Realistically, however, few of the considerations addressed in the EIS have any prospect of affecting the agency’s final decision. If the goal is to promote good decision-making, the agency would be better off with a shorter, more focused document.
Also, no one voluntarily adopts the project management technique of writing 3,600-page reports exhaustively documenting every consideration. Private companies, who have every incentive to make good decisions, certainly don’t. This suggests that there are better ways to promote good decision-making.
It’s not clear these reports accurately predict the future. According to a 1986 study of 29 EISs, “only 30% of the impacts were unqualifiedly close to their forecasts, with almost as many rated accurate principally by virtue of the vagueness of the forecasts.” Yikes! Why spend half a decade creating a report containing predictions that won’t even materialize?
Tell me what you want, what you really really want
After an agency approves a project, a project opponent can sue the agency for violating NEPA. Such lawsuits can take several forms:
If an agency concludes that an action falls within a “categorical exclusion” from NEPA, a plaintiff can sue and say, no, it doesn’t.
If an agency conducts a preliminary “environmental assessment” and concludes that no full-scale EIS is needed, a plaintiff can sue and say, a full-scale EIS was needed.
If, as in the Eagle County case, the agency does prepare the full-scale EIS, a plaintiff can sue and say, the EIS isn’t good enough.
Therein lies NEPA’s deepest flaw.
Opponents of federally-approved projects, like everyone else, have every right to challenge agency action in court. I certainly don’t mean to accuse anyone of litigating in bad faith. The root problem with judicial review under NEPA is the misalignment between what NEPA plaintiffs actually want and the relief they seek in court.
NEPA plaintiffs allege that EISs are inadequate and seek court orders directing the agency to improve the EISs. In reality, however, they do not care about the adequacy of EISs. EISs serve the purpose of educating the public, but NEPA plaintiffs do not want to be educated by the agency. To the contrary, they think the agency should be educated by them—typically, they allege they brought relevant information to the agency’s attention that the agency ignored.
EISs also purportedly serve the purpose of facilitating reasoned decision-making, but NEPA plaintiffs do not care about reasoned decision-making in the abstract. Instead, the goal of NEPA plaintiffs is to stop the project. This isn’t intended as a criticism. That must, by law, be their goal. NEPA plaintiffs don’t have standing unless they’re harmed by the project and a court ruling in their favor would redress that harm by stopping the project. This means the courthouse door is closed unless the NEPA plaintiff is trying to stop the project (or at least modify it).
So by design, our legal system authorizes plaintiffs pursuing one goal—stopping the project—to file lawsuits seeking a different goal—an improved EIS. This has several drawbacks.
Overlitigation. The vast majority of NEPA lawsuits lose. The court typically holds that the agency’s environmental analysis was good enough. Why do plaintiffs keep bringing these suits? Because of the misalignment between what the plaintiffs want and what they seek in court. The lawsuit might not succeed in yielding an improved EIS, but it will seek in delaying, and potentially stopping, the project.
Unnecessarily long EISs. Agencies know they will be sued over the adequacy of EISs and therefore prepare unnecessarily long EISs in order to fend off lawsuits. This extraneous information in EISs is useless along all dimensions. It’s useless to the agency, which won’t use it in decision-making and merely writes it down in anticipation of litigation. It’s also useless to NEPA plaintiffs, because, again, those plaintiffs do not care about the EIS’s adequacy.
Depressing litigation. NEPA lawsuits often involve disputes over the adequacy of an agency’s explanation on some particular issue. The plaintiff argues, “you should have given a more thorough explanation of consequence X,” and the agency argues, “our explanation is thorough enough.” It’s bickering over the adequacy of paperwork. In a normal world, the parties could just work it out—the agency could add a few sentences of explanation and everyone could go away happy. But because the lawsuit is a proxy fight over whether the project will proceed at all, the parties engage in bare-knuckles litigation over whether a document should have been longer.
The upward spiral. NEPA litigation always goes one way: The plaintiff alleges the document wasn’t detailed enough. No one would have standing to allege the document was too detailed. So all NEPA decisions either (1) leave the status quo intact or (2) hold that more detail is needed. These decisions create a new baseline that agencies must exceed. After 30 years of this, it’s hardly a surprise that we would see a 3,600 page EIS.
I could go on about this, but there’s an abundant literature on NEPA elsewhere. For example, here’s a much more detailed overview of NEPA from the incredible Construction Physics Substack. Instead I’ll turn my attention next to the two big legal NEPA developments—the Supreme Court’s Eagle County case and the D.C. Circuit’s Marin Audubon Society case.
Hypothetical power plants vs. sage-grouse
The Eagle County case will address whether the Surface Transportation Board’s 3,600-page EIS was good enough for government work.
The D.C. Circuit held that the Eagle County EIS was inadequate for two reasons. First, according to the D.C. Circuit, the Board didn’t adequately consider the legal doctrine known as If You Build It, They Will Come. The purpose of the new railway line is to transport crude oil that’s produced in northeastern Utah. If the railway line is built, it will induce developers to build new oil wells, which will affect the environment. Also, the crude oil will be refined, which will also affect the environment.
The Board concluded that it was speculative to try to assess the effects of these hypothetical projects. It also observed that it had “no authority or jurisdiction over development of oil and gas … nor any authority to control or mitigate the impacts of any such development.” The D.C. Circuit held that the Board was required to analyze these upstream and downstream consequences in more detail.
Second, the D.C. Circuit held that the EIS didn’t adequately analyze risks that would arise when trains crossed from the new Utah line into Colorado. Coloradans were worried about train accidents causing oil spills and sparks causing wildfires. The Board did address these issues a little bit, but the D.C. Circuit found that the Board’s analysis was insufficiently thorough.
In the Supreme Court, the railway developers’ argument can be summarized as “Come on!” Is 3,600 pages of environmental analysis really not enough? We’re talking about some new train tracks here. Does the Board seriously have to consider hypothetical oil and gas projects and hypothetical train derailments?
The project opponents respond with their own version of “Come on!” The purpose of this railway line is to stimulate new crude oil production. How can you conduct an intelligent analysis of the environmental consequences of a railway line while ignoring the very projects the line was intended to spur? Also, railway lines are built for trains. Trains sometimes derail. You can’t ignore this reality because it’s “hypothetical.”
Both sides have a point here. The developers’ position makes sense as a policy matter:
Other agencies regulate railway safety and oil and gas development. Requiring the Board to analyze these indirect consequences is double counting.
There’s nothing the Board can do to prevent future derailments or train-related fires in Colorado other than deny the permit. It’s not a situation where, e.g., the Board can insist that the developers use environmentally friendly construction techniques. And I don’t think the Board should deny the permit because of hypothetical future derailments and fires in Colorado. Occasional derailments and fires are inevitable consequences of trains, and I find it weird for the Board to say, “we, the national railway regulator, won’t let you build a new railway line because trains might have accidents on other railway lines that already exist.” And if these considerations shouldn’t affect the Board’s decision, the Board shouldn’t have to address them in the EIS.
Likewise, there’s nothing the Board can do to prevent pollution from future oil and gas production projects other than deny the permit. And again, that’s a bad reason to deny the permit. A future regulator is better positioned to evaluate a specific oil and gas development project. The Board shouldn’t say, “we’ll prevent a developer from building a railway line because we don’t trust future regulators to manage the environmental effects of hypothetical future projects.”
As a practical matter, it seems unlikely that the Board’s decision will change if it adds more paper on these topics when it’s already said that they won’t make a difference. So why are we doing this?
In terms of the public notice function of an EIS, is it really useful for the agency to say, “we’re not an oil and gas regulator and we have no idea what projects will be pursued or approved, but here is an arbitrary estimate with massive error bars”?
3,600 pages is long enough!
On the other side of the ledger, none of this is grounded in law. NEPA doesn’t have a “the report is too long already and this seems unnecessary” exception. Also, in recent years, the Supreme Court has repeatedly endorsed muscular judicial review of agency decision-making, including in several cases involving industry groups’ challenges to environmental regulations. Now that the plaintiff is an environmental group, it seems like a bait and switch for the Court to suddenly hear the gospel of deference to administrative agencies.
On balance, I think the developers’ position will carry the day because it makes more sense. But that holding is a legal kludge that won’t solve the problem of 3,600-page EISs. If you want to change NEPA, you have to think bigger.
It’s optional to do what’s optional.
The second recent NEPA development is Marin Audubon Society v. FAA, which is sort of the ultimate Washingtonian decision.
The Marin Audubon Society case doesn’t answer the question, “how much should we protect the environment?” Nor does it answer the question, “how much paperwork should we create in deciding how much to protect the environment?” Instead, it answers the following question: “which federal bureaucrat should decide how much paperwork we should create in deciding how much to protect the environment?” Depressingly, this is a high-stakes question.
Tourists enjoy flying over national parks. In 2000, Congress enacted a law requiring vendors who want to conduct commercial air tours over national parks to get a permit from the Federal Aviation Administration (FAA). The law provided that before issuing such a permit, the FAA, in cooperation with the National Park Service (NPS), must come up with an air tour management plan—which requires NEPA compliance. To speed things up, Congress mandated that the FAA “make every effort” to act on permit applications within two years.
This didn’t work out. The permit application process deteriorated into bureaucratic chaos that was remarkable even by the low standards of the federal government. Among other things, the FAA and NPS bickered over who would conduct particular NEPA determinations. Twelve years later, there were still no management plans. In 2012, to speed things up, Congress enacted a new law authorizing, in some cases, “voluntary agreements” in lieu of management plans. The agencies were able to achieve some voluntary agreements, but this process also ended up being hideously slow and dysfunctional. Nineteen years after Congress enacted the original law with the two-year deadline, essentially no progress had been made. So the D.C. Circuit issued a writ of mandamus—basically, a court order telling the agencies to do their job.
The FAA and NPS ultimately prepared an air tour management plan for purposes of flights over areas near San Francisco such as Point Reyes National Seashore. But they decided they didn’t need to prepare an environmental analysis under NEPA because the environmental impact would be minimal.
How did they make this decision? Well, there’s a federal agency known as the Council of Environmental Quality, or CEQ, that issues government-wide rules on when environmental assessments do and don’t have to be prepared. The FAA and NPS concluded that, under the CEQ’s rules, an environmental assessment was unnecessary.
As it turns out, for reasons I won’t get into, the FAA and NPS’s decision that an environmental assessment was unnecessary was itself nonsensical. The D.C. Circuit had no choice but to reverse that decision and send it back to the agency. To sum things up so far: after nearly 20 years of irrational dysfunction and paralysis, including disputes over who would handle NEPA analysis, the agencies finally issued a plan that didn’t include an environmental assessment for reasons that were also irrational.
But, in the course of reaching that conclusion, a two-judge majority of the D.C. Circuit panel also held that the CEQ regulations weren’t even binding, because the CEQ didn’t have statutory authority to issue them. For decades, everyone assumed the CEQ did have this authority, but the D.C. Circuit took a fresh look and decided that it didn’t.
This doesn’t mean that NEPA goes away. The statute itself, in its pristine, Doric glory, is still on the books. And this doesn’t mean that NEPA regulations go away. Federal agencies with rulemaking authority, like the Department of Interior, are still allowed to write their own NEPA regulations. The D.C. Circuit’s holding was that the CEQ—the centralized agency that has been writing government-wide regulations on this issue—doesn’t have this power.
As Chief Judge Srinivasan’s dissent persuasively explains, the panel shouldn’t have decided that issue:
None of the parties asked the court to decide this issue. Courts are supposed to resolve issues presented by the parties.
The court vacated the plan anyway because of the agency’s nonsensical reasoning, so the additional holding was unnecessary to decide the case.
The CEQ regulation at issue created an optional exemption from NEPA. So the court held, “a CEQ regulation that says that an environmental assessment is optional is itself optional.” This was an unnecessary holding.
This decision was so weird that I predict it will be vacated by the en banc court. Still, the cat is out of the bag. The decision will show up on Google even after it’s been vacated, and eventually a court will decide the issue in a case where it’s properly raised.
I don’t have a strong view on whether the CEQ currently has the authority to issue NEPA rules that bind the executive branch. The CEQ should have the authority, though. Someone has to decide what NEPA means. I see three options:
No NEPA regulations exist, so agencies fumble around in the dark and judges tell them what to do. The result will be, as Judge Friendly poetically put it, a “judicial oak” that grows from a “legislative acorn.”
Each federal agency creates its own NEPA requirements, yielding numerous inconsistent requirements and World War III when two agencies need to collaborate.
A single, centralized federal agency decides what NEPA requires.
Personally, I am staunchly anti-oak. I also think it’s better to have one set of NEPA regulations than ten. So by process of elimination, the CEQ should write the rules. I understand some people think that the CEQ has imposed too many procedural requirements and wiping out those rules will get rid of red tape. Maybe that’s true, although some CEQ rules—such as the one at issue in the Marin Audubon Society case —are designed to get rid of red tape. But if wiping out CEQ rules will prompt each agency to write their own rules—or, worse, let judges decide in each case when an EIS is required without any regulatory guidance—then boy, be careful what you wish for.
Claude, write me a 3,600 page EIS
Eagle County and Marin Audubon Society are important, but not transformational, cases. By contrast, AI’s impact on NEPA is transformational. Quite simply, AI breaks NEPA.
NEPA requires agencies to produce detailed reports. It doesn’t matter what conclusions the reports reach, as long as they’re sufficiently thorough.
In the past, it was impossible to prepare a thorough EIS without extreme human effort, which at least was circumstantial evidence that the agency took a hard look at environmental issues in making its licensing decision. That is no longer true.
It is an understatement to say that AI helps write EISs. To illustrate the point, I asked Claude this query:
Suppose I am writing an environmental impact statement about a new railway line in the Uinta Valley. I need you to write me the table of contents of that environmental impact statement. It should be highly detailed and granular, with at least 50 different topics. Make sure all topics that reasonably should be included in an environmental impact statement are covered.
Here’s the answer, which I’ll put in a document to avoid email length limits.
You can then ask Claude to fill in each category at any level of granularity. For example, here’s what Claude says about the sage-grouse:
This analysis would withstand judicial review. Judges are ex-lawyers, not environmental experts, and they have no ability to evaluate the correctness of statements in an EIS. Instead, judges merely check whether the reports seem rational and address every point raised by commenters. I promise you that a judge who read those paragraphs would say that the agency adequately considered the sage-grouse.
Claude cannot generate an entire EIS this way. The agency puts a tremendous amount of work into the EIS, including doing its own fieldwork. We have a long way to go before AI-powered robots handle NEPA fieldwork. Still, the agency prepares a lot of the EIS based on its review of documents collected from the applicant and other agencies. These documents can be uploaded into Claude, and Claude can instantly evaluate them and prepare an EIS.
The upshot is that AI can be used to prepare a plausible-seeming EIS, of arbitrary length, detailed enough to withstand judicial review, in a small amount of time—without the agency having actually taken a hard look at environmental impacts.
NEPA detractors may think this is great, but I don’t. I would still like to live in a world where agencies try to mitigate environmental harms as best they can, albeit without spending years preparing reports that span thousands of pages. When I read the Eagle County EIS, I am stunned that the document exists, but also strangely inspired by the competence, knowledge, hard work, and passion about the environment that it memorializes. I don’t want those capabilities going away.
Given that EISs no longer serve their intended purpose in the age of AI, I think a grand bargain is needed, in which current procedural requirements are relaxed in exchange for alternative AI-driven techniques for ensuring that the agency takes a hard look at mechanisms for mitigating environmental harm. Some ideas for such a grand bargain would be as follows.
First, make EISs dynamic. There is no longer any need for a PDF containing a long EIS. The justification for the document—that its very existence shows that the agency took a hard look at environmental issues—no longer holds. And it’s not a useful document for the public. To lay citizens, it’s too long. But it covers so much ground that it can’t spend more than a couple of pages on any particular issue, so for experts interested in particular issues, it’s too short. Rather than waste time and effort writing this document that no one will read, the agency can take all of its raw materials and put it into a sufficiently large context window. People interested in the project can ask whatever questions they want. If they want a general overview, they can ask for that. If they want a detailed analysis of surface water data or whatever, they can ask for that too.
Is this permissible under current law? I think so: I’m comfortable saying that a dynamic EIS satisfies NEPA’s requirement of a “detailed statement.” If not, have AI prepare an executive summary of whatever length makes the courts happy.
Second, severely curtail judicial review. Judicial review in NEPA cases frequently (not always) consists of judges making sure that agencies addressed all comments and gave rational explanations. It’s not really law, it’s more like “did the agency did a minimally competent job at explaining its reasoning?”
AI has its plusses and minuses, but one thing it can definitely do is offer minimally competent explanations. Grinding through thousands of comments and writing a few paragraphs of prose in response to each one is right in the sweet spot of AI’s capabilities. Judicial review of the AI’s work isn’t much of a value-add. Remember, from the court’s perspective, it doesn’t matter if the prose is right or wrong, it only matters that it’s there. And if suggestion #1 is adopted—a switch from static to dynamic EISs—the concept of “check to see whether the agency responded rationally to every comment” becomes irrelevant.
Given the problems with judicial review under NEPA, why not wave a magic wand and make it go away? As long as the agency certifies that it followed an appropriate AI-driven project management program, of which more below, the actual substance of the EIS could be unreviewable. I think this, too, would be permissible under existing law, especially if the agency produces a dynamic EIS that would be capable of offering far more information than EISs in the status quo.
What about the argument that “it’s dystopian to replace judges with AI?” I think that’s always a bad argument, but it’s an especially bad argument here. I recently re-watched Mad Max: Fury Road, a dystopian classic, and the absence of judicial review under NEPA was not a plot point. I can also confirm that no-judicial-review-under-NEPA did not drive the plot in Children of Men.
It is simply not dystopian to replace judges with AI in NEPA cases. I get the intuition that it’s dystopian to have an AI judge preside over a proceeding where due process interests are at stake. But there’s no due process right for a project opponent to have a judge evaluate an environmental impact statement. Instead, judicial review under NEPA serves the practical purpose of ensuring that the agency follows the law. If we can replace judicial review with a more efficient way of achieving that practical purpose, we should.
Third, use AI as a project manager. One of the goals of NEPA review is to encourage the agency to propose and evaluate measures to mitigate environmental harm. The ability to rapidly propose and evaluate such measures is also a strength of AI. If we trade “approve the project, but endure 3,600-page EISs and years of strategic lawsuits” for “approve the project quickly with as many or more mitigation measures than before,” this strikes me as a win-win.
This sounds hand-wavy. But one of the nice things about AI is that AI is excellent at both (a) designing project management programs and (b) understanding its own capabilities. If you want to figure out how to deploy AI effectively in NEPA review, by far the most effective way is to simply ask AI. I’ll again spare you long block-quotes.
In my view, AI puts a world in which there’s both less red tape and more environmentally-friendly development within our reach.
This is an extremely lucid, useful post.
> “Rather than waste time and effort writing this document that no one will read, the agency can take all of its raw materials and put it into a sufficiently large context window.”
YES! This is the future of writing I think; the PRIMARY SOURCES are still valuable and cannot be generated by AI. In fact, really, we don't need Wikipedia anymore; what we still need are all the primary sources it cited to generate the articles. You can then ask an LLM to dynamically generate an article on any topic in any style (iambic pentameter, Dr. Seuss, encyclopedic, sarcastic, right-biased, left-biased) rooted in the same primary sources.
> “A document no one will read shouldn’t be written.”
What a great quote; and it’s unique to you (I checked with Google).