Jon Mellon @jonmellon.bsky.social · 18/09/2026Someone should really teach the devs that a world war starting because of a series of bilateral alliances is unrealistic 130
Reposted by Jon MellonWeRateDAGs @weratedags.com · 10/09/2026Economists would declare that rainfall is already random, then go play outside their discipline. Which is a great moment to give 13/10 to the DAG o'Mellon for its child-friendly iv-assumptions & huge cast of possible treatment variables From @jonmellon.bsky.social (2024) 'Rain, Rain, Go away' AJPS 021
Jon Mellon @jonmellon.bsky.social · 01/09/2026would love to see a working paper if it sees the light of day 020
Reposted by Jon MellonJohn V. Kane @uptonorwell.bsky.social · 31/08/2026Compared to "significant effects," null results lead citizens to: --View study as less important --Ask for less info --Be less willing to share it --Think journalists would be less likely to cover the study --See the study as being lower quality & researchers as less competent (😬) 284
Reposted by Jon MellonJohn V. Kane @uptonorwell.bsky.social · 31/08/2026🧐 What do citizens think about "null results" in scientific research? To find out, @johnholbein1.bsky.social & I fielded a large-scale online experiment. We gave respondents straightforward info about a study, asked them some Qs, showed them the results, then asked them Qs again. 53617
Reposted by Jon MellonThe British Election Study @britishelectionstudy.com · 20/08/2026📢 New Data 📢 We are pleased to announce the release of Wave 31 of the British Election Study Internet Panel. You can find out more information, including where to download the data, here: www.britishelectionstudy.com/uncategorize...britishelectionstudy.com Release note for Wave 31 of the British Election Study Internet Panel - The British Election Study 287
Jon Mellon @jonmellon.bsky.social · 19/08/2026being part of a community of people who are collectively producing science is a big perk. 010
Jon Mellon @jonmellon.bsky.social · 15/08/2026People have asked whether the 1.4 pt average overestimate of left wing parties that we find matters. Well that enough to move the democrats senate chances from realistic to long shot according to Nate Silver’s latest model 110
Jon Mellon @jonmellon.bsky.social · 13/08/2026You have to be so paranoid with any systematic errors in industry because even tiny ones will be significant when your n is in the hundreds of millions (and substantively meaningful simply because even tiny effects sum up to meaningful dollars) 180
Jon Mellon @jonmellon.bsky.social · 05/08/2026Wait. the reviewers are likely halfway through the review. 010
Reposted by Jon MellonChris Prosser @caprosser.com · 05/08/2026Lots more analysis in the paper, breaking it down over time and by country, so take a look! osf.io/preprints/so...osf.ioOSF 121
Reposted by Jon MellonChris Prosser @caprosser.com · 05/08/2026Obviously, 'on average' doesn't mean polls are always-and-everywhere biased to the left - there is a lot of heterogeneity between elections - but elections are much more likely to have a leftward polling bias than a rightward one. 121
Reposted by Jon MellonChris Prosser @caprosser.com · 05/08/2026Using ~8.5k election-poll-party observations from 372 elections in 32 countries, we find a persistent partisan asymmetry: polls tend to overestimate the left relative to the right. On average across elections, polls the month before an election overstated the left by about 1.4 points 3128
Reposted by Jon MellonChris Prosser @caprosser.com · 05/08/2026New working paper from Stuart Perrett, @drjennings.bsky.social , @jonmellon.bsky.social , @cbwlezien.bsky.social and me: Do polls underestimate support for right-wing parties? Assessing variation in polling error by party family osf.io/preprints/so...osf.ioOSF 33221
Reposted by Jon MellonChris Prosser @caprosser.com · 29/07/2026No doubt you’ll all be rushing to pre-order the hardcover for the bargain price of £95 (😬), but should that prove a little steep, we’re pleased to say that the electronic version will be available open access, i.e. free! 183
Reposted by Jon MellonChris Prosser @caprosser.com · 29/07/2026It’s not out for another six months or so, but *Electoral Realignment: How Brexit Reshaped British Voting Behaviour*, my book with Ed Fieldhouse, @profjanegreen.bsky.social , Geoff Evans, @jack-bailey.co.uk and @jonmellon.bsky.social, is up on the OUP website! global.oup.com/academic/pro... 13013
Jon Mellon @jonmellon.bsky.social · 25/07/2026And if you really want to seal it, I have a very mild distaste of gambling on top of that 000
Jon Mellon @jonmellon.bsky.social · 25/07/2026If I’m also due to inherit 10 billion dollars in a month anyway but it won’t come in time to buy the medicine, then I think there’s no n that makes it worth it. 100
Jon Mellon @jonmellon.bsky.social · 25/07/2026If I need exactly $50,000 to afford my and my family’s life saving medicine (that is only taken once), then I think the answer is no for any finite n. 100
Jon Mellon @jonmellon.bsky.social · 15/07/2026I don’t know if you address it but the small sample size issue bites very differently for descriptive (where finite population corrections apply) and causal questions 100
Jon Mellon @jonmellon.bsky.social · 15/07/2026And to be clear I’m reacting to the framing in the thread/abstract. I’m sure the paper covers much of this 110
Jon Mellon @jonmellon.bsky.social · 15/07/2026At the most extreme, understanding the economics of alien societies would be fascinating, but the available data is zero. I think most social scientists would say the correct response is not to study it (at least empirically) 100
Jon Mellon @jonmellon.bsky.social · 15/07/2026Ie even if you’re correct that “there’s not a feasible way to collect data that would be more informative” that doesn’t mean the design succeeded 210
Jon Mellon @jonmellon.bsky.social · 15/07/2026My gut reaction is Tukey’s “The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.” 110
Jon Mellon @jonmellon.bsky.social · 15/07/2026Really interesting paper. Are you actually saying we should lower statistical standards for fields where large samples are infeasible or is the argument more subtle than that? 200
Jon Mellon @jonmellon.bsky.social · 15/07/2026My impression is that the ideological post-training for Chinese models is not that intense and mostly surface level. A more ideological government might try harder 020
Jon Mellon @jonmellon.bsky.social · 14/07/2026Or do they risk training on western sources and do industrial scale Soviet post training with political advisors monitoring closely 100
Jon Mellon @jonmellon.bsky.social · 14/07/2026It’s hard to avoid that because their reaction has to involve whether they have sufficient domestic training data to avoid training on primarily western sources 110
Jon Mellon @jonmellon.bsky.social · 14/07/2026I think you have to settle on a theory of how the USSR has reacted to the internet first (I’m assuming the hypothetical is the USSR never fell and LLMs developed on same schedule?) 120
Jon Mellon @jonmellon.bsky.social · 14/07/2026Do you see their incentives and priorities being very different to China’s on this? 120
Jon Mellon @jonmellon.bsky.social · 11/07/2026Followup question from my 3 year old: how many windows are there in the world? 080
Jon Mellon @jonmellon.bsky.social · 11/07/2026Is it easy to define how overconstrained a system is generically? Even in the sudoku case, you could presumably have situations where you could remove 1 or remove 8 clues to make the puzzle non-unique? 020
Jon Mellon @jonmellon.bsky.social · 06/07/2026There’s going to be 10 minutes of additional time isn’t there? 210
Jon Mellon @jonmellon.bsky.social · 06/07/2026Sure I’m saying there’s an alternate theory here that’s compatible with the evidence and that that’s grounded in my general scepticism of interaction effects (which are usually not that large in well powered studies) 100
Jon Mellon @jonmellon.bsky.social · 06/07/2026I’m making a claim about interaction effects in social science in general (which is definitely true) not in this one specific field to be clear with that statement 100
Jon Mellon @jonmellon.bsky.social · 06/07/2026I think you’re reading me as making a far larger claim than I am. My claim is just that this study does not provide good evidence for its claims and that we should not be relying on it. 100
Jon Mellon @jonmellon.bsky.social · 06/07/2026I’m open to an effect existing but this specific study is getting a lot of play and it isn’t well designed. It wouldn’t be that much effort for someone to run a good version of this study and I think someone should. 100
Jon Mellon @jonmellon.bsky.social · 06/07/2026This kind of analysis is specifically what I’m thinking of. They’ve already highlighted that gender predicts AI suspicion so it’s clearly post-treatment conceptually 110
Jon Mellon @jonmellon.bsky.social · 05/07/2026Interesting but seems like a lot of conditioning on colliders going on in there 110
Jon Mellon @jonmellon.bsky.social · 05/07/2026getting external validity right is a lot harder. Also at that point you might want to compare more than just resumes (a linkedin post, email, literary prize submission etc) 110
Jon Mellon @jonmellon.bsky.social · 05/07/2026The methods table in the report for anyone who doesn’t want to dig it out 030
Jon Mellon @jonmellon.bsky.social · 05/07/2026Maybe, but the good design seems quite simple. Constant Resume (or ideally multiple resumes) Vary name X telling respondent about AI use dv: perceived competence, hiring decision yes/no Comp~female+AI+ femaleXAI 110
Jon Mellon @jonmellon.bsky.social · 05/07/2026Also based on my prior that most interactions are small 120
Jon Mellon @jonmellon.bsky.social · 05/07/2026My theory is that people are generically sexist and manufacture reasons to justify it. My bet is that the interaction effect size is trivial 160
