Lightnews — Scholar-powered news

Uri Simonsohn

@urisohn.bsky.social

3.2K followers 570 following 810 posts

Pragmatic methodologist and behavioral scientist at Esade business school. Barcelona website: https://urisohn.com Blog: https://DataColada.org

urisohn.com

Posts Media Videos Starter Packs

Pinned

Uri Simonsohn @urisohn.bsky.social · Nov 7

With statistics prediction is easier than causal inference.
With punditry it's the opposite.

1 1 9

Uri Simonsohn @urisohn.bsky.social · 6d

Which restrictions do you have in mind?

Uri Simonsohn @urisohn.bsky.social · 10d

why may chatgpt give better answers than Cursor, even when i set gpt as the engine for the latter and it has the whole project as background instead of just one script?

Reposted by Uri Simonsohn

Ni-cola Ga-stan @nicgaston.bsky.social · 13d

First time I’ve seen the rabbit/bird illusion with an actual photo. Took me just a half second to process the word in all caps…

Matt @mfyfyr.bsky.social · 13d

whomst among us etc

25 80

Uri Simonsohn @urisohn.bsky.social · 15d

I don't think it is fair to characterize this as a "both sides' situation. I tried *for months* to engage in open dialogue

They lacked the common-decency to reply to the emails asking for the paper, after I answered emails from them for years

If they write a reply, happy to have a follow-up post

Uri Simonsohn @urisohn.bsky.social · 15d

2) Clintin & Richard say we claim to test if p-curve is skewed but we don't do that

I see this as semantics.

Indeed. We don't test whether *observed* p-values are skewed.
Rather. We test whether data generator produces skewed ps

e.g. p=.0001, p=.0002 are not skewed but suggest skewed generator

Uri Simonsohn @urisohn.bsky.social · 15d

They say, I think 2 things that fall in this category.

1 ) Clintin & Richard say "evidential value" is not a stats concept.

I agree, but this is because we invented it. We needed a term for "significant after accounting for p-hacking", It didn't exist, so we came up with evidential value.

Uri Simonsohn @urisohn.bsky.social · Aug 25

A feature i really like is custom intro and "outro" skipping so you skip the yada yada and get to the actual episode

1 1

Uri Simonsohn @urisohn.bsky.social · Aug 25

Podcast addict

1 1

Reposted by Uri Simonsohn

Cloudflare @cloudflare.social · Aug 4

Perplexity is repeatedly modifying their user agent and changing IPs and ASNs to hide their crawling activity, in direct conflict with explicit no-crawl preferences expressed by websites. blog.cloudflare.com/perplexity-i...

Perplexity is using stealth, undeclared crawlers to evade website no-crawl directives

Perplexity is repeatedly modifying their user agent and changing IPs and ASNs to hide their crawling activity, in direct conflict with explicit no-crawl preferences expressed by websites.

blog.cloudflare.com

3 44 70

Uri Simonsohn @urisohn.bsky.social · Aug 4

Victorian Internet by Sandage

History of the Telegraph

Uri Simonsohn @urisohn.bsky.social · Jul 25

yeah, i never got in portrait mode. but i can see the upside of 2 screens if you like that

Uri Simonsohn @urisohn.bsky.social · Jul 25

Been using two monitors for many, many years.
Just tried a single wide (curved 34") monitor instead.

It's slightly less wide total a bit taller.
It's way better.

Looing at center of screen instead of to the side, no dead zone in the middle.

1 5

Reposted by Uri Simonsohn

Ehud @duhe.bsky.social · Jul 7

While in many respects the situation is dire I think it’s worth pointing out that there are multiple scientific ecosystems. Being scandalized shows that we normally live in a better functioning, privileged part of the scientific community.

Gautam Kamath @gautamkamath.com · Jul 7

System is so broken:
- researchers write papers no one reads
- reviewers don't have time to review, shamed to coauthors, use LLMs instead of reading
- authors try to fool said LLMs with prompt injection
- evaling researchers based on # of papers (no time to read)

Dystopic.

1 5

Uri Simonsohn @urisohn.bsky.social · Jun 24

With their approach this seems easy actually. What we call 'treated stimulus design'. Draw one man, black or white, algorithmically generate the twin of the other race. And There is still room for stratified sampling

1 2

Reposted by Uri Simonsohn

Data Colada @datacolada.bsky.social · Jun 23

Colada[128] The Best Audit Study and its interesting shortcoming
datacolada.org/128

10 18

Uri Simonsohn @urisohn.bsky.social · Jun 23

Interesting that 10% of people don't answer leading poll questions.

Uri Simonsohn @urisohn.bsky.social · Jun 15

It's what we choose to focus on actually.

1 1

Uri Simonsohn @urisohn.bsky.social · Jun 15

I dont think thats a realistic concern if you just browse the home page. It's a funny headline, but not super diagnostic

1 3

Reposted by Uri Simonsohn

John Bull @garius.bsky.social · Jun 15

Scene from All the Presidents Men where the journos all stand looking disappointed

9 54 270

Uri Simonsohn @urisohn.bsky.social · Jun 13

Fellow European researchers, if at all in your power, try to get your school to consider late applicants who were admitted to US programs.

Hearing from many cases of admitted to top US schools who last minute realize they won't be allowed in the country.

1 6 18

Uri Simonsohn @urisohn.bsky.social · Jun 5

This is the kind of reporting i had in mind as necessary.
It does not look good
edition.cnn.com/2025/06/04/m...

Videos, expert analysis and witnesses point to Israeli gunfire in deadly Gaza aid site shooting | CNN

A CNN investigation into a deadly incident near an aid distribution site in southern Gaza on Sunday points to the Israeli military opening fire on crowds of Palestinians as they tried to make their wa...

edition.cnn.com

Uri Simonsohn @urisohn.bsky.social · Jun 4

Still seems like no footage, no witness accounts, denials by idf and agency, and only source is the terrorist group who has an incentive to make stuff up...

Uri Simonsohn @urisohn.bsky.social · Jun 3

Hm, Israel and the aid agency deny the shooting.
We can't adjudicate the truth here, but if the journalists don't tell you that it's disputed, I think you should not trust the journalists.

Reposted by Uri Simonsohn

Data Colada @datacolada.bsky.social · Jun 2

Don't assume, plot
datacolada.org/126

1 14 34

Uri Simonsohn @urisohn.bsky.social · May 29

I think the best comparison would be something like
X1: how interesting is the (non)finding?
X2: how credible is the finding?

Compare x1*x2 across formats

Know of any attempts to do that?

(though even that is just a correlation)

2 1 2