Bennett
LightNews LightNews
banner
bennettmcintosh.com
Bennett
@bennettmcintosh.com
150 followers 340 following 87 posts
Science & Technology Studies | History of Sci & Tech || How do data & scientific communities make each other? Current projects on: social genomics; environmental regulation

No offense but if you follow me with no bio I will probably block
Posts Replies Media Videos
bennettmcintosh.com
Add Harvard to the list of elite universities whose alumni databases have recently been targeted by social engineering attacks in the last few weeks.

www.insidehighered.com/news/tech-in...

[I didn't even go there!]
November 22, 2025 at 5:03 PM
bennettmcintosh.com
your post put this image in my head so now I'm putting it online
a doctored photo of a WeWork building, in which the letters "w-e" on the building's sign have been replaced by the letters "j-e-e" in a shoddy photoshop job. The building has a glass facade framed by tan concrete on the top floors and dark gray stone on the bottom floors. Palm trees slightly soften the edges but do not mitigate the imposing looming of the building
November 14, 2025 at 4:24 AM
bennettmcintosh.com
Also, for a great read on why replacing the "race" terminology with the "species" terminology doesn't solve the problems here, I highly recommend Claire Jean Kim's DANGEROUS CROSSINGS

www.cambridge.org/core/books/d...
> "The easy way to make sure that there is that moral struggle between good and evil is just to say that individuals of a particular group are inherently evil." > One of the most enduringly legible symbols that a character is different and therefore more disposable is race. Of course, the fact that this is also true in the real world is the reason it became such an effective shorthand.
November 13, 2025 at 12:54 AM
bennettmcintosh.com
this throw-away line stood out: Musk, Meloni &c see cosmopolitan exchange as erasing old "specificities" but queers & creoles & immigrants are behind countless *specific* cultural innovations. Labeling them undifferentiated Others or genetically-doomed subhumans is a worldview designed to hide this
paragraph highlighting the ways that reactionaries love Lord of the Rings: Thiel and his minions like Joe Lonsdale & JD Vance name their companies Palantir, Andruil, Erebor & Narya; Dubya's surrogates used Jackson's films as a metaphor for the war in Iraq. Highlighted is a line about how Italian PM Georgia Meloni is a "defender of 'Italianity' against what she sees as the dilution of immigration [and] a Tolkein obsessive who sees in hobbits, dwarves, and elves the 'value of specificity'"
November 13, 2025 at 12:54 AM
bennettmcintosh.com
It is ever cheaper & easier to produce new science undermining key regulations. Technical/"objective" signals of merit don't prevent this, and relying on social signals like funding source breeds resentment (and, as reanalysis becomes cheaper/easier, they may grow less reliable too)
"However, if we distinguish between replication (performing a study de novo) and reanalysis (re-examining an existing study’s data), several things become clear. First, recent clashes are part of a long conflict over data access, stretching back at least to the 1990s: while some actors may be interested in preventing the consideration of certain data, others are interested in reanalyzing data to challenge the scientific consensus, and in fact have done so. Second, using procedure or methodological quality to adjudicate reanalyses often fails, leading to a recourse to reputational markers that amplifies conflict. Scientific transparency is a"
October 29, 2025 at 10:30 PM
bennettmcintosh.com
In v2 of our preprint on "Adversarial Reanalysis," Kelsey Ichikawa, @nicolecnelson.bsky.social, & added our take on Gold Standard Science as an example of how #OpenData challenges science's existing tools for self-governance

osf.io/preprints/me... #sts #metascience
MetaArXiv title on osf.io: "Adversarial reanalysis and the challenge of open data in regulatory science"
October 29, 2025 at 10:30 PM
bennettmcintosh.com
the dreck calling itself "bold" & "truth-telling" is fascinating

were you not meeting artificial quotas or were you:
- calling your employees orangutans
- threatening them into silence about rape
- hiring men who can't drive forklifts over women who can
- destroying evidence about all of the above?
“Maybe you should suck it up and actually compete on the ground that we have in 21st-century America?” That is what feminists think happened, but they are wrong. Feminization is not an organic result of women outcompeting men. It is an artificial result of social engineering, and if we take our thumb off the scale it will collapse within a generation. The most obvious thumb on the scale is anti-discrimination law. It is illegal to employ too few women at your company. If women are underrepresented, especially in your higher management, that is a lawsuit waiting to happen. As a result, employers give women jobs and promotions they would not otherwise have gotten simply in order to keep their numbers up. It is rational for them to do this, because the consequences for failing to do so can be dire. Texaco, Goldman Sachs, Novartis, and Coca-Cola are among the companies that have paid nine-figure settlements in response to lawsuits alleging bias against women in hiring and promotions. No manager wants to be the person who cost his company $200 million in a gender discrimination lawsuit. As the first accuser in the racial discrimination case against Texaco, Bari-Ellen Roberts was anything but surprised when she heard the epithet-laced recordings that have disgraced the oil giant. "People would tell me, 'You can't believe the language that goes on in the meetings,'" said the 44-year-old senior financial analyst about secretly made tapes of high-level meetings that ignited public outrage when they surfaced last week. "I had heard from managers that I was referred to as that little colored girl, that I was called uppity," she said. "We were told that other African-Americans in other Texaco locations were referred to as orangutans and porch monkeys.” She learned the contents of the two microcassettes just before a transcript of the meetings was filed last week in White Plains Federal Court. EEOC's suit alleged that Coca-Cola Bottling Company of Mobile, owned and operated by Coca-Cola Bottling Company Consolidated, and CC Beverage Packing, Inc. refused to hire Martina Owes, an applicant for two vacant warehouse positions, because she is female. While Owes had the required warehouse and forklift experience, the company chose to hire less qualified men for the available positions. EEOC also charged that by not preserving all application materials related to those positions, the company violated federal record-keeping laws. (WMT) will groan in dismay at the $250 million punitive damages award won by female employees of Novartis (NVS) in a New York federal court today. Walmart is next on the docket of major corporate gender discrimination class action cases. A jury found Novartis unfairly treated its female employees by not promoting them as much, discriminating against workers who got pregnant, and most infamously threatening a female drug sales rep after she reported to HR that was raped by one of her customers.
October 17, 2025 at 12:09 AM
bennettmcintosh.com
they're not employing stupid methods towards laudable goals; rule-by-fiat is their goal and *pretending* they have laudable goals is the method by which they intend to achieve it!
These are laudable goals. This is what is good in the compact. The bad in the compact is often in the methods proposed for achieving those goals: The idea that viewpoint diversity is best achieved through administrative bean-counting and, implicitly, viewpoint-based quotas. A range of threats to the First Amendment, including the erasure of a line between criticism of conservative views and incitement to violence. The unwillingness to rely on the illegality of incitement as sufficient protection against it. The micromanaging of admissions and tuition-setting that reflects an imperfect understanding of those processes, of the economic realities of colleges and universities as businesses, and of First Amendment standards protecting freedom of association. Work that must be done intellectually - for instance, achieving a culture of debate on campus — is here treated as work that can be accomplished by fiat.
October 7, 2025 at 11:13 PM
bennettmcintosh.com
It's awkward phrasing, but I don't read this as barring anything that *isn't* education/advocacy/support, so much as saying that those explicitly *will* be allowed, even in forms that might otherwise violate other rules. But possible that I (& bsky!) have not thought through how this could go wrong!
E. Mental Health & Wellbeing: We support a healthier community by prohibiting content that could harm people's mental or physical health. Do not share content, methods, instructions, or promotion of self-harm, suicide, eating disorders, and extreme dieting practices. Do not share methods, instructions, depictions, or promotion of dangerous stunts or abuse of dangerous or controlled substances. 5. Protected Expression These guidelines protect legitimate expression that serves the public interest and supports community wellbeing. The following content is welcome on Bluesky when appropriately labeled and contextually framed: Journalism, Analysis, Education, and Advocacy: Factual reporting, academic research, or educational content about matters of public interest, safety, or legal significance. Anti-violence advocacy, safety campaigns, and prevention resources that raise awareness without providing harmful methods or glorifying harm. Content warnings about dangerous viral challenges or emerging safety threats. Support, Recovery, and Mental Health: Personal recovery experiences or survivor stories shared to promote healing and community support rather than encouraging harmful behavior. Sharing verified mental health resources, crisis prevention information, and support group discussions. Transparency and Public Information: Publicly available information from official sources such as court records, business filings, property records, or related to government transparency and public officials.
September 21, 2025 at 11:53 PM
bennettmcintosh.com
Doomers, who treat super-intelligent AI as likely and threatening, treat intelligent as almost a magic power, but it's much more heterogeneous, spiky (and, I would add relational) than that; But the AI-as-Normal-Tech argument seems naive about the incentives to use risky AI in, e.g., Mil Tech
Second, what exactly does this argument take “intelligence” to mean? It seems to be treating it as a unitary property (Yudkowsky told me that there’s “a compact, regular story” underlying all intelligence). ==But intelligence is not one thing, and it’s not measurable on a single continuum. It’s almost certainly more like a variety of heterogenous things== — attention, imagination, curiosity, common sense — and it may well be intertwined with our social cooperativeness, our sensations, and our emotions. Will AI have all of these? Some of these? We aren’t sure of the kind of intelligence AI will attain. Besides, just because an intelligent being has a lot of capability, that doesn’t mean it has a lot of power — the ability to modify the environment — and power is what’s really at stake here. One of the most glaring flaws of the normalist view is that it doesn’t even try to talk about the military. Yet military applications — from autonomous weapons to lightning-fast decision-making about whom to target — are among the most critical for advanced AI. ==They’re the use cases most likely to make governments feel that all countries absolutely are in an AI arms race, so they must plow ahead, risks be damned.== That weakens the normalist camp’s view that we won’t necessarily deploy AI at scale if it seems risky.
September 17, 2025 at 3:31 PM
bennettmcintosh.com
Who called it a topologically improbable political compass and not an Ezra Klein Bottle
this is like a really crowded meme, but the left side is a political compass, twisted around so that the lib-right and auth-left segments connect (as do the lib-left and auth-right) and the right side is text about how your political compass results are "Abundance Tankie" who is somehow a fan of both Kleinism-Thompsonism and Marxism-Leninism who will turn the DSA into a corporatist engine of state capacity or something
September 6, 2025 at 2:56 AM
bennettmcintosh.com
I still think about this (from 2018!) regularly. So many people reaching seeking solace from the "privilege of an apocalypse"

proteanmag.com/2018/12/11/n...
Two paragraphs from the linked essay. The passage begins by pointing out that climate change brings the timescales of geology and human life into "nightmarish combination" and then continues: "The paralyzing effect of discussing this leads some to conceive of climate change as Armageddon, the end of the world as we know it. But this too is inaccurate. In the near future, we will be denied the privilege of an apocalypse. There will be no point at which external forces will summarily and magnificently end human enterprise, because there will be no recognizable end. Like an engine starting to wheeze, sputter, and break down, the expanding growth of our economy will begin to be undercut by the intractable difficulties of adapting to these new conditions."
August 20, 2025 at 5:51 AM
bennettmcintosh.com
Still better than dairy milk on basically every metric, and the good news is that there are even better options -- soy milk is high-protein, low impact, and delicious; imo The King of Milks ourworldindata.org/environmenta...
charts from Our World in Data showing the environmental impact (in land use, greenhouse gas, freshwater use and eutrophication) of dairy milk and various alternatives. Dairy milk is by far the most impactful; only in water use do any of the plant alternatives even reach half the impact of dairy milk (Almond milk uses ~370 liters of water per liter of milk, to dairy milk's ~630). In every other case, every other milk has a fraction of the impact
July 6, 2025 at 7:11 PM
bennettmcintosh.com
BTW you probably aren't allergic to penicillin, and if your chart says you are, ask your doc about delabeling for your own sake and the sake of antibiotic stewardship. Test might be 1hr; $0 out-of-pocket

Helps everyone if front-line antibiotics are available!
onlinelibrary.wiley.com/doi/full/10....
Abstract Background Even though 8%-25% of most populations studied globally are labeled as penicillin allergic, most diagnoses of penicillin allergy are made in childhood and relate to events that are either not allergic in nature, are low risk for immediate hypersensitivity, or are a potential true allergy that has waned over time. Penicillin allergy labels directly impact antimicrobial stewardship by leading to use of less effective and broader spectrum antimicrobials and are associated with antimicrobial resistance. They may also delay appropriate antimicrobial therapy and lead to increased risk of specific adverse healthcare outcomes. Operationalizing penicillin allergy de-labeling into a new arm of antimicrobial stewardship programs (ASPs) has become an increasing global focus.
June 19, 2025 at 1:14 AM
bennettmcintosh.com
i personally think viewing each other as flat dossiers has done wonderful things for our social lives & especially dating and that the only thing that could improve on it would be if it were also how parents picked their children

VCs pls give me $53,800,000 thx

techcrunch.com/2025/06/06/g...
Demo of a "compare your embryos" dossier from Nucleus Genomics, allowing prospective parents to select embryos predicted to have their favored appearance, disease risk, psychological profile, and IQ
June 6, 2025 at 11:33 PM
bennettmcintosh.com
Nice exploration by @ericlevitz.bsky.social al challenging the idea that digital orality threatens literary liberalism (an idea that, ironically, gains adherents as much through repetition as through rhetorical validity)

www.vox.com/politics/414...

But I think AI's role will be... weirder 🧵
Snippet from article: For much of that interval, books enjoyed a cultural relevance that they currently lack. And generative AI appears to be accelerating deep literacy’s marginalization, enabling even humanities majors to eschew the cognitive burdens of reading and writing. It is reasonable to worry that the mental habits instilled by TikTok and ChatGPT might prove less conducive to liberal democracy than those promoted by the written word. In particular, Garfinkle and Mir’s concerns about social media’s corrosive influence on independent thought, and tendency to promote polarization, seem well-founded. And yet, even as Mir and Garfinkle lament the post-literate tendency to substitute “lived experience” for objective evidence, they themselves often make very strong claims on the basis of little more than intuition.
June 3, 2025 at 5:05 PM
bennettmcintosh.com
INWORKS data availability statement from their most recent pub:

www.bmj.com/content/382/...
Data availability statement For reasons of ethics and permissions from different agencies, the data are maintained at the International Agency for Research on Cancer (Lyon, France) and cannot be made available outside of the agency.
May 29, 2025 at 10:11 PM
bennettmcintosh.com
Follow up: in another EO on Friday, the White House apparently ordered the Nuclear Regulatory Commission to reconsider the Linear No Threshold dose response model discussed in our paper, at least in regards to radiation

www.whitehouse.gov/presidential...
Screenshot from White House site: (b) Adopt science-based radiation limits. In particular, the NRC shall reconsider reliance on the linear no-threshold (LNT) model for radiation exposure and the “as low as reasonably achievable” standard, which is predicated on LNT. Those models are flawed, as discussed in section 1 of this order. In reconsidering those limits, the NRC shall specifically consider adopting determinate radiation limits, and in doing so shall consult with the Department of Defense (DOD), the Department of Energy (DOE), and the Environmental Protection Agency.
May 29, 2025 at 2:11 AM
bennettmcintosh.com
Such science – even in less-partisan contexts – can be difficult & time-consuming to adjudicate properly. There's no objectively best analytical decision, & reputational markers are imperfect at best & counterproductive at worst.

We don't have a lot of solutions, but hoping this prompts discussion!
When they do gain access, the debate necessarily shifts away from the quality of the data they hold and towards the quality of the analysis they have conducted. This is typically more difficult and time consuming to adjudicate. Even in cases not already mired in political controversy, disagreements about whether to use dichotomized or continuous variables or which participants to exclude can consume hours of time from dozens of editors and reviewers. A recourse to reputational markers may further exacerbate the problem by feeding a politics of resentment that provides additional motivation for dissident scientists to engage in these disputes
May 26, 2025 at 4:20 PM
bennettmcintosh.com
Anyway, we see these political actions -- subpoenas, legislation, regulations -- as continuous with attempts by folks we call "dissident scientists" to gain access to regulatory data. They were often stymied by data holders, but when they did gain access they did do and publish science on these data
that this is a case where the cigar may just be a cigar—there is ample historical evidence to show that proponents of the Transparency Rule did want access to data to perform their own reanalyses. Moreover, taking these efforts at face value is analytically useful because it demonstrates how open data polices, regardless of motive, could promote adversarial reanalysis and produce potentially paralyzing challenges to existing scientific and regulatory infrastructures.
May 26, 2025 at 4:20 PM
bennettmcintosh.com
As @danhicks.bsky.social notes, this order has a lot of continuity with the Pruitt/Wheeler EPA's "Strengthening Transparency" rules – and we trace that history back as well to legislation aimed at "secret science" in the early 2010s and to attempts as early as the 1990s to access regulatory data
The 2013 subpoena that Christakis and Zimmerman cited is one plot point in a much longer story about data access, reanalysis, and environmental regulation. It begins in the 1990s when analyses of the two data sets in question first demonstrated links between air pollution and mortality. These studies prompted the EPA to introduce new National Ambient Air Quality Standards for fine particles in 1997, and the cost of implementing these new regulations in turn prompted intense scrutiny of the underlying research. Opponents of the regulation had demanded that the authors release their data, and the authors had refused, calling these access requests ‘researcher harassment’ (Allen, 1997). As one Harvard scientist involved in the controversy put it, ‘We knew that if we released the data, it would be endless aggravation and defending against attacks’ (Grant, 2012). The leaders of both studies eventually agreed to an independent reanalysis of their data (Health Effects Institute, 2000), but this did little to quell the controversy. When Christakis and Zimmerman were writing, more than a decade after the independent reanalysis, dissident scientists were still attempting to gain access to the data.
May 26, 2025 at 4:01 PM
bennettmcintosh.com
First, what's the EO? The order defines "Gold Standard Science" thusly

@danhicks.bsky.social has a good thread here, on some of the weirdness and contradictions in those definitions: bsky.app/profile/danh...

And you can read the full order here: www.whitehouse.gov/presidential...
For the purposes of this order, Gold Standard Science means science conducted in a manner that is: (i) reproducible; (ii) transparent; (iii) communicative of error and uncertainty; (iv) collaborative and interdisciplinary; (v) skeptical of its findings and assumptions; (vi) structured for falsifiability of hypotheses; (vii) subject to unbiased peer review; (viii) accepting of negative results as positive outcomes; and (ix) without conflicts of interest.
May 26, 2025 at 3:52 PM
bennettmcintosh.com
I have a new preprint out w/ @nicolecnelson.bsky.social & Kelsey Ichikawa!

We focus on EPA regulatory transparency policies between '17 and '21 but wow is it relevant to current policies on #openscience, #opendata, including this weekend's "gold standard #science" EO

📄 osf.io/preprints/me...

🧵⤵️
MetaArXiv Preprint title: "Adversarial reanalysis and the challenge of open data in regulatory science" by Bennett McIntosh, Nicole Nelson, and Kelsey Ichikawa
May 26, 2025 at 3:41 PM