HN Debrief

The federal keyword lists that canceled billions in research funding

  • Politics
  • Science
  • Public Policy
  • Education
  • Regulation

The article reports that multiple federal agencies used lists of disfavored words and phrases to review grants, then canceled awards tied to topics or viewpoints the administration wanted gone. The lists were not narrow. They reportedly included terms tied to diversity efforts, climate and clean energy, accessibility, and even whole fields like the humanities. The legal challenge centers on whether grants were revoked for expressing views the administration dislikes, rather than for scientific merit or program fit.

If your work depends on US grants, assume proposal language itself is now a political risk and audit for terms that could trigger blunt screening. More broadly, treat federal science funding as less rules-based and more discretionary than many institutions assumed, which changes planning, hiring, and where long-term research can safely live.

Discussion mood

Overwhelmingly angry and alarmed. The dominant view was that the keyword lists are anti-science, ideologically motivated, and intentionally crude, with many commenters seeing the incompetence as either a feature for chilling speech or cover for a broader project of politicizing institutions.

Key insights

  1. 01

    False positives reached basic technical language

    What makes the policy genuinely destructive is that it does not just hit overtly political proposals. It catches ordinary domain language. Math faculty reportedly stripped out references to inequalities, neuroscience proposals had trouble with cellular diversity, geology grants hit on mineral inclusion, and biologists started dodging the word female by writing around sexed animal models. That turns a grant filter into a contamination source for the research record itself, because people start corrupting titles, abstracts, and methods just to survive screening.

    Do not only review proposals for obvious political flashpoints. Check whether normal technical vocabulary could be misread out of context and decide where precision is worth the risk versus where wording can be safely normalized.

      Attribution:
    • whatever1 #1
    • superhuzza #1
    • beej71 #1
    • floren #1
  2. 02

    Accessibility work got swept in too

    One concrete casualty was accessibility language for screen readers. That is a useful warning because it shows the lists are not just targeting culture-war slogans. They can block routine inclusion work that most software teams would consider baseline usability. The policy boundary is far wider than many people first assume.

    If grants support product, data, or platform work, separate critical accessibility deliverables from politically exposed wording where possible. Do not assume obvious user accommodation work will be treated as neutral.

      Attribution:
    • Forgeties79 #1
  3. 03

    Machine filtering looks like accountability laundering

    Comments tied the weird phrasing on the lists, including terms like “social engineering,” to the possibility that parts of this process were generated or amplified by LLMs or similarly blunt automation. The sharper point was not whether an LLM was literally used. It was that automated scoring gives officials a way to hide ideological choices behind a machine output and blur responsibility for absurd calls. That makes appeals harder because applicants are no longer arguing with a reviewer’s stated reasoning. They are arguing with an opaque pipeline.

    Ask funders and internal review teams for explicit human-review criteria, not just outcomes. If decisions come from automated triage, push early for audit trails and examples because you will need them if a proposal is rejected on nonsense grounds.

      Attribution:
    • tancop #1
    • kombookcha #1 #2
    • secretsatan #1
  4. 04

    The chaos is serving a clear objective

    Several comments rejected the idea that the lists are merely bumbling. The more useful reading is that trauma, uncertainty, and overbreadth are part of the operating model. Russell Vought’s quoted goal of making bureaucrats and agencies feel shut down lines up with a process that does not need to be accurate. It only needs to make institutions hesitate. From that angle, the bad filter is working exactly as intended if researchers preemptively censor themselves and staff stop defending borderline work.

    Plan for chilling effects, not just direct losses. Teams that rely on grants should expect reviewers, administrators, and partner institutions to become more risk-averse even before formal policy reaches them.

      Attribution:
    • spit2wind #1
    • ZeroGravitas #1
    • beej71 #1
  5. 05

    This looks like patronage-state behavior

    A more structural read compared the situation to weaker institutional states where formal rules exist on paper but real outcomes depend on favor, negotiation, and reading elite signals. The mention of the dual state model sharpened that point. The problem is not simply that one administration dislikes certain topics. It is that predictable process is being replaced by discretionary power, which is exactly what makes research and investment environments decay over time.

    Treat this as a governance risk, not just a culture-war episode. If you run a lab, company, or university program, diversify funding sources and jurisdictions because the issue is institutional reliability, not one rejected proposal.

      Attribution:
    • hibikir #1
    • walrus01 #1
    • lokar #1
  6. 06

    People reached for Lysenkoism fast

    The Lysenkoism comparison landed because it captures the core fear. Political doctrine is starting to decide which lines of inquiry are acceptable, and researchers are being pushed to rewrite reality in administratively approved language. That does not mean the United States is replaying Soviet biology in full. It means commenters see a recognizable pattern where ideology begins by distorting funding and ends by distorting what can be truthfully said in a field.

    Watch for second-order damage in literature quality and shared terminology, not just budget cuts. Once researchers start writing around forbidden words, retrieval, replication, and synthesis all get worse.

      Attribution:
    • rsynnott #1
    • bewareofscams #1

Against the grain

  1. 01

    Some banned terms were specific to DoD

    One pushback was that at least part of the list applied to the Department of Defense, which weakens the broadest reading that every listed topic was banned everywhere. That does not make the policy harmless. The stronger reply was that terms like solar energy, hydrogen energy, and electrification obviously have military relevance, especially around contested infrastructure and logistics. The narrower scope changes the map of the damage, not the underlying absurdity of using keywords this way.

    Before reacting to any single term list, check which agency and program it governed. The practical exposure may be uneven, but even agency-specific lists can still hit strategically important work.

      Attribution:
    • meragrin_ #1
    • jaredklewis #1
  2. 02

    Past grant fashion is not the same thing

    Some comments tried to analogize this to earlier periods when grant applications were padded with climate or equity language to match funder priorities. That comparison mostly failed. Steering applicants to claim broader social impact is annoying grantmanship. Canceling existing work through hostile keyword screens is a different category. The distinction that stuck was compulsion to flatter a funder versus punishment for mentioning a forbidden concept at all.

    Do not collapse normal grant politics into what is happening here. If you advise researchers or founders, distinguish between signaling program alignment and navigating outright ideological blacklist behavior.

      Attribution:
    • transcriptase #1
    • Yokohiii #1
    • cwnyth #1
    • ithkuil #1
  3. 03

    Grantmaking was never apolitical

    A credible minority view said the novelty is being overstated because research funding has always been political and grants have long rewarded buzzwords, networks, and proposal craft as much as raw merit. That does not excuse the current lists. It does puncture any fantasy that federal grant systems were previously neutral. The useful lesson is that the old system was already vulnerable to this kind of capture because language games and discretionary judgments were built in from the start.

    If your organization relies on grants, stop treating them as a clean proxy for scientific quality. Build strategy around the fact that funding pipelines mix merit, politics, and grantsmanship even in healthier periods.

      Attribution:
    • AndyMcConachie #1
    • dnautics #1 #2
  4. 04

    Elected governments still face constitutional limits

    One commenter argued that a democratically elected administration should be free to stop funding research it disfavors, and that no one is entitled to public money. The best rebuttal was not that every grant must continue forever. It was that viewpoint-based revocation using a crude keyword dragnet is exactly where constitutional and administrative limits start to matter. Elections authorize governance, not arbitrary punishment for disfavored speech dressed up as routine grant management.

    Expect the important line in these cases to be process and viewpoint discrimination, not a generalized right to funding. If you are affected, document evidence that rejection or cancellation tracked language and ideology rather than program criteria.

      Attribution:
    • adamiscool8 #1
    • wtfwhateven #1

In plain english

dual state ↩
A political model where formal legal institutions still exist, but important decisions are increasingly made through arbitrary power outside normal rules.
LLM ↩
Large Language Model, a machine learning model trained to generate and analyze human-like text.
Lysenkoism ↩
A Soviet-era example of political ideology overriding science, named after Trofim Lysenko, whose rejected biological theories were enforced by the state.

Reference links

Background on political control of science

  • Lysenkoism
    Used as the main historical analogy for ideology distorting science funding and accepted research topics.
  • Dual state model
    Shared to frame the broader shift from rules-based institutions to discretionary power.

Fiscal and political context

Corruption and self-dealing references