Ron Garcia

screenshot of the previous post, with "a murder of subtweets" circled in bold

Your most important contribution to the discourse to date:

Introducing ARIA’s next CEO Kathleen Fisher will join us in February 2026.

Kathleen Fisher is an inspired choice as the next ARIA CEO. And I was *just* reading her papers on PADS to send to @patrick.sirref.org as ideas for our time travelling shell! ariaresearch.substack.com/p/introducin...

Have you tried discarding the computer and getting another?

This is very cool! Totally on board with statement 1, but I'm having trouble fully grasping statement 2, since the contrast of interest does not involve conditioning on G, and S has no backdoor paths to W. Maybe the pipe from C to Ds could be a salient information flow?

So is the relevant observation to make of this model
P(W=yes | do(S = yes) ) vs. P(W = yes | do(S = no))?

An excerpt from McDonald's "Test Theory" book where he grouses about the term "difficulty".

I got a good chuckle out of this rather formal rant about difficulty the other week:

(FWIW I've often heard (and agree) that the hardest problem in computer science is naming things, so pretty sure we suck just as much!)

Obituary: Brian Cantwell Smith (1950 to 2025) - Faculty of Information Former dean wove philosophy and computation into a lifelong inquiry into the human dimensions of intelligence, judgment, and meaning Born into a prominent Canadian family known for both its intellectu...

Beautiful remembrance of Brian Cantwell Smith, from Faculty of Information, University of Toronto: ischool.utoronto.ca/news/obituar...

New on my blog: "Why don't you use dependent types?"
lawrencecpaulson.github.io/2025/11/02/W...

Is that leftmost obscured book a methods book on writing memes? Please say yes, even if it's a lie!

UBC Computer Science is hiring up to 2 full-time tenure-track Assistant Professor positions. Applications due Dec 10 2025

UBC Computer Science invites applications for up to two full-time tenure-track positions with the following priority areas: visualization, robotics, reinforcement learning, data management, and data mining. Applications are due Wed Dec 10, 2025. Learn more: www.cs.ubc.ca/our-departme...

I enjoy this sport (having played as a young'un), but nowadays I call it Helmetball to disambiguate.

New on my blog: Everything you know is wrong
lawrencecpaulson.github.io/2025/09/20/W...

There's a Timsort, why not a Bernasort?

Interesting! What's the context, somehow I missed it while I was in Indiana 😂

Ahh! Thanks. I would have expected "second-class" to contrast and first-class (and match Strachey's discussion of the terminology).

What does it mean for a variable to be first *order*? Is it that there are no variable references whose referent can be changed after initialization?

While technically true, I look forward to hearing about how many people receive the signal 😂 (a shame tbh).

Some gnarly example due to Lévy, where conditional probability according to Kolmogorov's axioms gives answers that make no one happy.

Recently I learned that sometimes the axioms aren't enough 😢

(from "When is conditional probability meaningful?" in arxiv.org/pdf/1802.06071)

Summer Undergraduate Internship - reposts welcome!

Are you a senior undergrad, interested in Programming Languages? Do you want to visit Canada for a paid 12-week internship?

ONE OF US! ONE OF US! ONE OF US!

A history of Clojure | Proceedings of the ACM on Programming Languages Clojure was designed to be a general-purpose, practical functional language, suitable for use by professionals wherever its host language, e.g., Java, would be. Initially designed in 2005 and released...

Not "can't", but "didn't": so the question is "why not?". In the case of Clojure, you can find some explanation in the HOPL paper:
dl.acm.org/doi/10.1145/...

Probably the most relevant part is Section 3.5 (Clojure and the Host), but tail calls are specifically addressed briefly in Section 3.2.3.

I wonder if there is a connection between these ideas and Default Logic :
en.wikipedia.org/wiki/Default...

(whenever I see some logic called "non-monotonic" I wonder if there is some structure missing from its formulation that could be judgmentally reconstructed)

Caricature of Kuhn by David Levine for The New York Review of Books.

Thomas Kuhn was born OTD in 1922.

Historicity—over revolutions and paradigms—was core to his thought.

“History, if viewed as a repository for more than anecdote or chronology, could produce a decisive transformation in the image of science by which we are now possessed.”

🧠🗃️ #HistSTM #PhilSci 🦋🦫

Also it should be possible to relate the high-level (cyclic) model to a lower-level (acyclic) model. Recent theoretical work along these lines went by here recently:
bsky.app/profile/p-hu...

While this isn't their point, it helped me to think of adding TC to the "micro" model as well. There it would be a descendant of LDL and HDL, but have no arrow to HD. So if you treat LDL and HDL as unmeasured, you get two common causes of TC and HD, i.e. backdoor paths galore :).

Did you hear that I'll be teaching four brand new remote courses this year covering Bayesian mixture modeling, survival modeling, pairwise comparison modeling, and ordinal modeling, www.eventzilla.net/e/advanced-b...?

Reskeets and sharing with your friends and colleges appreciated!

cue my standard mantra 😉:
bsky.app/profile/did:...

Thanks! I likely don't know enough about DAGs, but I had imagined that it would suffice to add a finite number of extra nodes (some function of the nodes one wrote down) to summarize all of the potential threats to inference (as aggregate causal forces). Is that wrong-headed?

Makes sense! For a while I did not understand that the *absent* arrows in a DAG are the assumptions, not the arrows that are visible. Then there is also all those absent "unobserved common causes". Might be educational to draw out the "negative space" of a DAG and then argue why they're absent 🤔

(oh also curious about how these assumptions end up being invisible, since I thought the whole point of the framework was to make the assumptions explicit)