Sign in

Joshua Loftus

@joft.bsky.social
1.9K followers 385 following 447 posts

Random wanderer, toddler parent, famous fitness influencer Statistics & Data Science Asst Prof at LSE (machine learning, causality, fairness, interpretability, ethics). First gen grad. Unschooled -> math @ WMich & Rutgers -> stats PhD @ Stanford

PostsRepliesMedia
Joshua Loftus @joft.bsky.social · 22h
Other people: ideologues Me: pure objectivity
020
Joshua Loftus @joft.bsky.social · 29/09/2026
My conspiracy theory for why the "AI" CEOs are suddenly all in agreement about slowing down is that they already expect model improvement to plateau and want to say it's voluntary rather than hitting hard technological/statistical limits
0121
Reposted by Joshua Loftus
Ben Recht @beenwrekt.bsky.social · 29/09/2026
The illusion and politics of objectivity in cost-benefit analysis, a decision-making institution so ingrained that Americans can’t imagine life without it.
argmin.net
Pricing Commensurability
On the origins of cost-benefit analyses in governmental decision making
2269
Joshua Loftus @joft.bsky.social · 29/09/2026
The old world is dying, and the new world struggles to be born: now is the time of monsters… extremely cute monsters that culturally evolved to hijack our baby-preserving psychology
0289
Reposted by Joshua Loftus
the fool @agnoster.net · 29/09/2026
the future could be magnificent, and you deserve a share of it
Retro-futurist propaganda poster overlooking a radiant, green utopian city of rivers, bridges, gardens, and gleaming towers at sunrise. Four people and a shaggy dog sit on a flower-covered hillside, gazing toward the city. Across the sky in huge block lettering: “the future could be magnificent, and you deserve a share of it.”
316727
Joshua Loftus @joft.bsky.social · 29/09/2026
MTG might be one of the most “median voter” representatives to ever be elected
110
Joshua Loftus @joft.bsky.social · 28/09/2026
Always Sunny title meme: The gang goes out to eat
000
Reposted by Joshua Loftus
James O’Loughlin @jamesoloughlin.bsky.social · 28/09/2026
02517338
Joshua Loftus @joft.bsky.social · 27/09/2026
This is a good explanation for why my toddler is able to say "dowwwwnnn" with so much feeling
000
Joshua Loftus @joft.bsky.social · 26/09/2026
Rather disappointing to see such sloppiness from a scholar whose actual peer reviewed work and even(!) popular books were pretty respectable
120
Joshua Loftus @joft.bsky.social · 24/09/2026
We put a random number generator in our computer program, so, actually your honor, nobody can blame us for what it does
001
Joshua Loftus @joft.bsky.social · 23/09/2026
yes please 🥹
Email subject: Training with tendon injuries
Sent by: strongerbyscience.com newsletter
000
Joshua Loftus @joft.bsky.social · 23/09/2026
This is how we can pretend every problem in the world is exactly like enterprise software development
000
Joshua Loftus @joft.bsky.social · 19/09/2026
If you tell them I sent you you will automatically be hired, guaranteed
010
Reposted by Joshua Loftus
Mattan S. Ben-Shachar @mattansb.msbstats.info · 31/05/2026
Psychologists doing clustering be like: #statsmeme
940663
Joshua Loftus @joft.bsky.social · 18/09/2026
Yes
000
Reposted by Joshua Loftus
Julia M. Rohrer @dingdingpeng.the100.ci · 15/09/2026
The question that’s dominating the field is not a causality question. It’s really, how can one sell a causal relationship while crafting the language so that there’s plausible deniability when somebody criticizes anything.
56512
Joshua Loftus @joft.bsky.social · 16/09/2026
Statisticians, data scientists, or similar nerds- come join my department! It's a great place to work, and it's in London! www.lse.ac.uk/statistics/a...
lse.ac.uk
Recruitment - Join us in the Department of Statistics!
000
Joshua Loftus @joft.bsky.social · 15/09/2026
The President of the United States is a… President… of the United States
000
Joshua Loftus @joft.bsky.social · 13/09/2026
Bsky libs competing over “The President of the United States is a [various criminal and other epithets]” But morally, just being POTUS is worse (more harmful) than any of the other things they say. The next person will also be a POTUS and also oversee an empire of extraction and planet destruction
000
Joshua Loftus @joft.bsky.social · 11/09/2026
It’s common around 9/11 for people to say the attack succeeded, that the US empire has been undermining itself and declining ever since I think this is wrong. The global empire of capital (which the US government mostly serves and protects) is stronger and less accountable than ever
001
Reposted by Joshua Loftus
Ted McCormick @tedmccormick.bsky.social · 10/09/2026
Wikipedia is not at all like generative AI slop, and it is *most specifically* unlike slop in the very features that make it useful and reliable — that it is authored by people who take responsibility for what they do and who both credit and point readers towards the sources they have in fact used.
116123
Reposted by Joshua Loftus
Boston Review @bostonreview.bsky.social · 09/09/2026
“Mathematics only makes sense as a gift economy. It has no trade secrets. Substituting a monetizable ersatz culture for the one evolved through historical experience is a particularly insidious form of knowledge collapse.” Mathematician Michael Harris on the AI companies coming for math:
bostonreview.net
Knowledge Collapse
AI companies are racing to mechanize mathematics. Where does that leave human understanding?
1355121
Reposted by Joshua Loftus
will pooley @willpooley.bsky.social · 08/09/2026
it’s plagiarism all the way down
0151
Joshua Loftus @joft.bsky.social · 05/09/2026
how to tell if a mammal is actually smart
020
Joshua Loftus @joft.bsky.social · 04/09/2026
For years, (some) people have been telling me that counterfactual fairness is debunked or flawed or wrong or something because "race cannot be a cause," and I've mainly shrugged ("all models are wrong," etc), but now I can also cite a Leverhulme award winning philosopher who agrees with me
0151
Joshua Loftus @joft.bsky.social · 04/09/2026
I saw the first graph and was like yeah yeah ok, it's probably just a sample composition changing over time. Then I saw the second graph 😬 www.glassdoor.com/blog/how-wor...
Graph shows portion of Glassdoor reviews that mention "AI" by sentiment, positive vs negative, with percent negative rising from 15 in 2019 to 53 in 2026Graph shows share of Glassdoor reviews that mention AI or LLMs, it's up from less than 0.1 percent of reviews in 2022 to around 2.5 percent of reviews in 2026
120
Joshua Loftus @joft.bsky.social · 04/09/2026
Imagine you sent an email to someone years, maybe even decades ago, and now they're giving OpenAI access to its contents
131
Joshua Loftus @joft.bsky.social · 04/09/2026
No surprise, he's clearly just another utilitarian who doesn't think rules apply to them
000
Joshua Loftus @joft.bsky.social · 03/09/2026
The new Claude Fable model found an error in my 2024 NeurIPS paper (jk) by constructing a counterexample DAG with a "direct edge from A to a grandchild" -- which my wife immediately named the Woody Allen DAG
Claude acknowledging its own error: "Freezing the children or all descendants gives the same counterfactual for both the NDDP and the NIDP: every directed path from A to a later descendant passes through a child, so the rest recompute identically. My "direct edge from A to a grandchild" case is incoherent — such a node is a child."
020
Joshua Loftus @joft.bsky.social · 31/08/2026
Over on X (The Everything App) a podcaster wrote this up as a scifi story about several "civilizations" of "AI agents" rising and falling over a few days, and the people cooking their brains on that site are resharing it with the awe that my toddler has when he sees a new kind of toy
050
Joshua Loftus @joft.bsky.social · 29/08/2026
On the one hand the USA has accelerated climate change more than anywhere else; bombed, invaded, sanctioned, and tariffed most of the world; and is leading the way in replacing democracy and science with kleptocracy and antivax fundamentalism But on the other hand, France created the Minions...
110
Reposted by Joshua Loftus
Paul Hünermund @p-hunermund.com · 27/08/2026
📢 The Causal Data Science Meeting 2026 is back! Join us virtually on Nov 4–5 for two days of discussion on #causality in ML & AI, with Teppo Felin as our first keynote speaker. No fees, no proceedings—just research, feedback & exchange. 📅 Submit by Sept 30 ✉️ submission@causalscience.org
0104
Joshua Loftus @joft.bsky.social · 27/08/2026
I was cycling in the Netherlands recently. Practically nobody wears helmets
100
Joshua Loftus @joft.bsky.social · 25/08/2026
Choose almost any (knowledge) learning goal in advance *that isn't about LLMs* Design a test of that learning Now tell me how the friction of *using an LLM to pass the test for you* is going to help you learn something about the original goal
260
Joshua Loftus @joft.bsky.social · 25/08/2026
Nothing has influenced any of my estimates since I was ~25 and had already converged on oracle level accuracy
000
Joshua Loftus @joft.bsky.social · 24/08/2026
The entire US is ~5% of the world population and ~20% of world GDP. I still don’t understand why the rest of the world isn’t just doing a little coordination and stomping trump like the soggy worm he is
000
Reposted by Joshua Loftus
jon ben-menachem @jbenmenachem.com · 23/08/2026
I think the class of school administrators (not just in higher ed) who act fast to insulate core pedagogical functions from AI will reap major rewards for their trouble. In a few years, there will be an extremely stark difference between people who used AI in school and people who are competent
2395113
Reposted by Joshua Loftus
Maria Farrell @mariafarrell.eurosky.social · 21/08/2026
My standout favourite edited book chapter is: @juliapowles.bsky.social: The Corporate Culpability of Big Tech, in The Culpable Corporate Mind, edited by Elise Bant, Hart Publishing, Oxford, 2023. Points the way to criminal culpability for Big Tech executives who know exactly what they're doing.
papers.ssrn.com
The Corporate Culpability of Big Tech
This contribution cracks open and lays bare the paradox of Big Tech's small liabilities for systemic misconduct, and provides a richly expansive pathway for bri
0122
Joshua Loftus @joft.bsky.social · 19/08/2026
Sorry, but this is the market predicting that this cure will cost more than chemo
010
Joshua Loftus @joft.bsky.social · 19/08/2026
This is my actual job (teaching machine learning)
010
Joshua Loftus @joft.bsky.social · 17/08/2026
Why are manufacturers legally allowed to sell cars that can go 120mph+? Regulate the industries and companies profiting from mass death (tobacco, guns, cars) before you tell me we need stricter policing and individual fines and should create a (private and for-profit!) panopticon
100
Joshua Loftus @joft.bsky.social · 17/08/2026
Statistician here. Harper is mostly correct, but with two deadly errors First, it’s not limited to social sciences. All academia is a drop in the bucket compared to “fake” analysis/research in the business world And second, it mostly doesn’t matter. Most research (and thought) is never acted upon
110
Joshua Loftus @joft.bsky.social · 16/08/2026
There is a massive UK-US cultural difference in medicine In the US, people self-select out of seeing a doctor because they know it will cost them money (I did recently while visiting) In the UK, people pay via waiting times and, basically, being told to "walk it off" as a default treatment
130
Joshua Loftus @joft.bsky.social · 15/08/2026
Not sure if it counts as an intuition, but I do "feel" that the following things are not mutually contradictory but actually hold or are real: - Actualism (or modal antirealism?) - Determinism - Free will
020
Joshua Loftus @joft.bsky.social · 14/08/2026
In the UK, private individuals with random anonymous social media accounts can be visited by the police for "abuse and harassment," while the media can do this legally and be rewarded for it
010
Joshua Loftus @joft.bsky.social · 13/08/2026
Some especially wealthy, well connected, and incurious people
000
Joshua Loftus @joft.bsky.social · 13/08/2026
Yes. And yet, (like democracy), it is still one of the best things humanity has going
1203
Reposted by Joshua Loftus
Julia M. Rohrer @dingdingpeng.the100.ci · 13/08/2026
Today I learned… en.wikipedia.org/wiki/Garrett...
Screenshot from Wikipedia:

Garrett James Hardin (April 21, 1915 – September 14, 2003) was an American ecologist, microbiologist, and white nationalist best known for the concept of the tragedy of the commons.
1212333
Joshua Loftus @joft.bsky.social · 12/08/2026
Wearing eclipse sunglasses is woke. ISO 12312-2 is basically socialism, to me
010