Demand data centers keep Americans safe. 

Detected Manipulations & Suggested Changes

Appeal to emotion

Using emotionally charged stories or language to persuade rather than focusing proportionally on evidence and uncertainty.

“In Morrow County, Oregon, families are living through a crisis. According to a Rolling Stone investigation, mothers have suffered miscarriages. Neighbors are battling rare cancers.” “Big Tech will complain. Some companies may threaten to locate elsewhere. But no community should accept development that requires sacrificing clean water or public health.” “The only remaining question is whether our leaders will act in time — or wait until the damage is done.”

3 suggestions

Clarify the evidentiary status of the Oregon health issues, e.g.: “In Morrow County, Oregon, families are facing serious health concerns, including reported miscarriages and rare cancers. A Rolling Stone investigation has raised questions about possible environmental contributors, though no definitive causal link to the nearby data center has been established.”

Reduce absolutist, fear‑based framing, e.g. change “no community should accept development that requires sacrificing clean water or public health” to “communities should carefully weigh potential economic benefits against any credible risks to water quality and public health, and require strong safeguards where risks are identified.”

Replace the dramatic closing (“act in time — or wait until the damage is done”) with a more neutral call for due diligence, e.g.: “Policymakers now face a choice about how proactively to address these risks through regulation and technology standards.”

Conflating correlation with causation / Post hoc fallacy (implied)

Presenting or strongly implying a causal link based mainly on temporal or spatial correlation, without sufficient evidence.

“Local officials have raised concerns about dangerously elevated nitrate levels in the community’s drinking water following the siting of a nearby Amazon data center. The investigation reports that the facility’s massive water consumption — up to five million gallons per day — may have accelerated nitrogen migration into the aquifer faster than natural filtration can occur. Amazon strongly denies any connection between its operations and these health problems.” The structure juxtaposes miscarriages and rare cancers with the data center’s water use and nitrate levels, which can lead readers to infer a causal link, even though the article acknowledges Amazon’s denial and uses “may have.”

3 suggestions

Explicitly separate correlation from causation, e.g.: “While elevated nitrate levels have been detected and some experts hypothesize that high water withdrawals could influence nitrogen migration, current evidence does not establish that the data center caused the reported health problems.”

Add context about other potential nitrate sources (e.g., agriculture, septic systems) and the state of scientific investigation, to avoid over‑attributing risk to data centers alone.

Include any available regulatory or independent assessments (if they exist) that either support or question the hypothesized link, and clearly label hypotheses as such.

Oversimplification

Reducing a complex technical, economic, or regulatory issue to a simple, one‑sided narrative or solution.

“We can build data centers today that eliminate operational water consumption entirely — and with the right policy reforms, they can be built at lower total cost than conventional designs.” “The solution is zero-water cooling through immersion systems with dry heat rejection.” “In other words, we can build cheaper data centers without draining aquifers or compromising drinking water.” “The technology exists. The economics can work. What remains is political will.”

3 suggestions

Qualify claims about feasibility and cost, e.g.: “In many cases, it is technically feasible to build data centers that eliminate operational water consumption, and under certain policy and market conditions, these designs may be cost‑competitive with conventional systems.”

Acknowledge trade‑offs and uncertainties, such as energy efficiency impacts, site‑specific constraints, and the maturity of immersion cooling at very large scales.

Replace “The solution is zero-water cooling…” with “One promising approach is zero‑water cooling…” and briefly mention alternative mitigation strategies (e.g., reclaimed water use, hybrid systems, improved efficiency) to avoid presenting a single silver‑bullet solution.

Unbalanced reporting / Selective use of sources

Presenting mainly one side’s evidence and arguments while giving minimal or cursory space to opposing views or uncertainties.

The article cites: - A Rolling Stone investigation and local officials’ concerns. - Modeling by researchers affiliated with Caltech about premature deaths. - Multiple industry analyses and academic studies supporting zero‑water cooling. By contrast, the industry’s position is summarized in one sentence: “Industry representatives argue that zero-water cooling is too expensive. That claim does not withstand scrutiny.” No specific industry data, detailed counterarguments, or independent critiques of zero‑water systems are presented.

3 suggestions

Include more detail on the industry’s cost and feasibility concerns, with specific examples or data, and then respond to them with evidence rather than dismissing them in a single sentence.

Present any known limitations, risks, or open questions about zero‑water cooling (e.g., reliability, maintenance, lifecycle environmental impacts of dielectric fluids) and how proponents propose to address them.

Clarify the scope and limitations of the cited Caltech‑affiliated modeling (assumptions, uncertainty ranges, peer‑review status) and, if available, mention any differing estimates from other research groups or agencies.

Biased language / Framing effect

Using loaded wording and framing that nudges readers toward a particular judgment rather than neutrally presenting facts.

Examples include: - “unchecked data center growth” - “regulators are signaling that some degradation of water quality may be acceptable to support data center growth.” - “Big Tech will complain. Some companies may threaten to locate elsewhere.” - “no community should accept development that requires sacrificing clean water or public health.” These phrases frame data centers and regulators in a negative light and portray industry responses as self‑interested and dismissive, without evidence for those characterizations.

3 suggestions

Replace “unchecked data center growth” with a more neutral phrase such as “rapid data center expansion” or “significant growth in data center development.”

Rephrase “regulators are signaling that some degradation of water quality may be acceptable…” to a more descriptive formulation, e.g.: “Under current antidegradation standards, regulators may allow some reduction in water quality when they determine that the social or economic benefits of a project outweigh the environmental costs.”

Change “Big Tech will complain. Some companies may threaten to locate elsewhere.” to a neutral description of likely policy debates, e.g.: “Some companies may argue that stricter standards could affect their siting decisions or project costs.”

Hasty generalization / Extrapolation from limited cases

Drawing broad conclusions or policy prescriptions from a small number of cases or early evidence.

“Whatever ultimately caused the health crisis in Oregon, Ohio communities are already feeling the strain of unchecked data center growth. We know how to prevent similar harm. We know it can save money.” The article moves from one highlighted case (Morrow County) and general modeling to broad claims that “we know” how to prevent harm and save money, without fully addressing variability across regions, technologies, and regulatory contexts.

3 suggestions

Qualify the scope of the conclusions, e.g.: “Emerging evidence from places like Morrow County and national modeling suggests that, in some contexts, data center growth can strain water and air resources. There are promising strategies that may reduce these risks and, in some cases, lower long‑term costs.”

Avoid categorical statements like “We know how to prevent similar harm” and instead use “There are evidence‑based approaches that can substantially reduce the risk of similar harms.”

Note that further empirical research and pilot projects are needed to validate cost and risk‑reduction claims across different regions and scales.

Unsubstantiated / Overstated claims

Making strong assertions that go beyond the evidence presented or are not backed by specific citations in the text.

“Industry representatives argue that zero-water cooling is too expensive. That claim does not withstand scrutiny.” “Taken together, these measures could reduce total project costs by more than 10% while eliminating operational water consumption altogether.” “We can build cheaper data centers without draining aquifers or compromising drinking water.” The article references “multiple industry analyses, pilot deployments and academic studies” but does not name them, summarize their methods, or indicate uncertainty ranges. The >10% cost reduction and categorical dismissal of industry concerns are presented as settled facts.

3 suggestions

Cite specific studies or reports (with authors, institutions, and dates) and briefly summarize their key findings and assumptions when making quantitative claims like “more than 10%.”

Qualify the strength of the conclusion, e.g.: “Some analyses suggest that, under certain conditions, these measures could reduce total project costs by around 10% or more while eliminating operational water consumption.”

Replace “That claim does not withstand scrutiny” with a more evidence‑based comparison, e.g.: “While industry representatives cite higher upfront costs, several studies find that, over the facility’s lifetime, zero‑water systems can be cost‑competitive or even less expensive when policy incentives and avoided water costs are considered.”

False dilemma (implied)

Presenting a situation as having only two opposing options when more nuanced or intermediate options exist.

“But no community should accept development that requires sacrificing clean water or public health.” “The only remaining question is whether our leaders will act in time — or wait until the damage is done.” These lines frame the choice as either adopting the advocated policies/technologies now or accepting serious harm later, without acknowledging intermediate regulatory approaches, partial mitigations, or the possibility that some data centers may pose low risk under existing rules.

3 suggestions

Acknowledge a spectrum of policy options, e.g.: “Policymakers can choose among several approaches, from incremental improvements in permitting and monitoring to more ambitious requirements for low‑water or zero‑water designs.”

Rephrase the closing to avoid a binary framing, e.g.: “The key question is how quickly and to what extent our leaders will strengthen safeguards and encourage lower‑impact technologies.”

Clarify that not all development necessarily “requires sacrificing” clean water or health, but that safeguards are needed to ensure those outcomes are protected.