Jon Mellon
@jonmellon.bsky.social
3K followers 2.6K following 660 posts
Co-director British Election Study. Political Scientist and Data Scientist. Political science methods/political behavior/causal inference. Posts do not represent employer.
Posts Media Videos Starter Packs
Pinned
jonmellon.bsky.social
Out now open access at
@ajpseditor.bsky.social. 194 potential exclusion-restriction violations for studies using weather as an instrumental variable onlinelibrary.wiley.com/doi/full/10....
jonmellon.bsky.social
Has anyone had success stories (or can point me to writeups) with fine tuning frontier LLMs?
Reposted by Jon Mellon
jamesbreckwoldt.bsky.social
Education? Completed it, mate.

Last week, I defended my PhD thesis and am now a doctor!

Thank you to @robfordmancs.bsky.social and @jonmellon.bsky.social for being great supervisors for 4 years

And thanks to @sarahobolt.bsky.social and Ed Fieldhouse for their interesting feedback and discussion
jonmellon.bsky.social
These are continuous not discrete variables and every commentary on this is assuming different cutpoints
jonmellon.bsky.social
I would say that a natural parsing of “is it acceptable to rob a store for food?” either implies that you cannot get food in another way or is answered with an easy no
jonmellon.bsky.social
Right but the question is whether we’re faced with the choice between (theft or starvation) or the choice between (theft or starvation or buying food). The urgency of an issue only necessitates support for extreme measures if less extreme measures won’t work
jonmellon.bsky.social
I actually don’t know which particular existential discussion you’re referring to, so this is purely reacting to the abstract point
jonmellon.bsky.social
Not eating food is existential but there’s a lot of limits on acceptable ways to get food if an option on the table is going to the store and buying it.
Reposted by Jon Mellon
rexdouglass.bsky.social
Rex W. Douglas PhD Applied Scientist (Remote/Austin)

Looking for full time and freelance projects.

Hoping for somewhere stable. I've never been more productive in my life, but mass layoffs and funding collapses have been endemic.
Reposted by Jon Mellon
rexdouglass.bsky.social
Help! ─ 𝗔𝗽𝗽𝗹𝗶𝗲𝗱 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁 open to work ─ Remote/Austin → www.rexdouglass.com
With new tools, I've been cooking:
▸ Machine-vision pipelines — RIOS Intelligent Machines
▸ Information-extraction pipelines — Microsoft
▸ Interactive SMS surveys — Pantheon Insights
Reposted by Jon Mellon
phil.gyford.com
I remember around 1995 a woman knocked on the door, doing a survey about air fresheners. I never bought any so it didn't last long but I was quite excited that someone was asking my opinion about a product or service. At last! Finally, Gen X was being consulted!
jonmellon.bsky.social
Social scientists often bemoan falling survey response rates. A big part of the story has to be the proliferation of survey requests.

I took every survey I was invited to in an 8 day period and took 27 surveys totaling 33 minutes.
jonmellon.bsky.social
Honestly the short ones seemed fine to me. The people I felt annoyed by were the ones making the giant tickbox grids (I have been one of those people before)
jonmellon.bsky.social
As a survey professional, I highly recommend that everyone takes all the surveys they’re shown
jonmellon.bsky.social
Yes those are included. There were 5 longer surveys (that account for the bulk of the time) and lots of micro ones
jonmellon.bsky.social
If I expand this out I would be taking 25 hours of surveys a year if I answered all of them
jonmellon.bsky.social
Most surveys are literally one question along the lines of: was this experience good? Yes/no

I took 5 surveys longer than a minute
jonmellon.bsky.social
I gave up on the experiment after a long jstor survey
jonmellon.bsky.social
Many of these were 1 click in-app surveys but there were multiple chunky qualtrics surveys in there too
jonmellon.bsky.social
Social scientists often bemoan falling survey response rates. A big part of the story has to be the proliferation of survey requests.

I took every survey I was invited to in an 8 day period and took 27 surveys totaling 33 minutes.
jonmellon.bsky.social
I would argue this is actually about Bayesian inference
jonmellon.bsky.social
Come see our work on using LLMs to track political science research methods at APSA at 4pm today!

We find the latest LLMs beat graduate student RAs at extracting information from academic papers
jonmellon.bsky.social
It could be something weird like they measured it in the wrong scale and had to do a weird fix. It all seems very odd
jonmellon.bsky.social
Not sure if this could be the case for some but not all (maybe if done manually)
jonmellon.bsky.social
Suppose the raw coefficients are 1.12, 2.67, and 3.13 and they round them to 1, 3,3. They then apply some scaling to that (maybe to standardize effects) by multiplying by 0.008.
jonmellon.bsky.social
One possibility is that there’s some rounding applied and then a rescaling on top of it. Eg if the original estimates were rounded to integers and then multiplied by an SD of 0.08
jonmellon.bsky.social
This varies a ton across space and time. Russia and Ukraine’s strategic positions rely far more on their Armies than Navies for instance.