Graham Walker
@grahamwalkernrg.bsky.social
Research @ VaasaETT
Previously @ Petrologica
Alum University of Essex
Views inevitably my own.
Previously @ Petrologica
Alum University of Essex
Views inevitably my own.
In Lancashire, it's black pudding Friday.
November 28, 2025 at 10:02 AM
In Lancashire, it's black pudding Friday.
Of course, applies to the Johnson fracking ban, too. And also applies in the reverse - e.g. Trump opens up Atlantic (and to lesser extent Arctic) where no one will drill as a culture war signal.
November 27, 2025 at 9:28 AM
Of course, applies to the Johnson fracking ban, too. And also applies in the reverse - e.g. Trump opens up Atlantic (and to lesser extent Arctic) where no one will drill as a culture war signal.
Applies also to other bans worldwide (Ireland, Denmark, NZ when the ban came in) - easier to ban exploration when you have minimal prospectivity.
November 27, 2025 at 9:26 AM
Applies also to other bans worldwide (Ireland, Denmark, NZ when the ban came in) - easier to ban exploration when you have minimal prospectivity.
I think you have hit upon why it is controversial, which is that it is a load-bearing ideological pillar of (a particular brand of) Christian social thought, and as such can be a major point of differentiation for other ideologies. The Sixties involved a major reaction against it, for example.
November 25, 2025 at 10:35 AM
I think you have hit upon why it is controversial, which is that it is a load-bearing ideological pillar of (a particular brand of) Christian social thought, and as such can be a major point of differentiation for other ideologies. The Sixties involved a major reaction against it, for example.
If you want a picture of the future, imagine a boot missing an American football - forever.
November 24, 2025 at 10:18 AM
If you want a picture of the future, imagine a boot missing an American football - forever.
There's surely a paper to be written replicating studies like these (even specifically this one), first with the same prompt and data, then with different prompt, then different data. Multiple runs - do you get consistent results with same prompt/data, what are the differences when you vary?
November 21, 2025 at 9:04 AM
There's surely a paper to be written replicating studies like these (even specifically this one), first with the same prompt and data, then with different prompt, then different data. Multiple runs - do you get consistent results with same prompt/data, what are the differences when you vary?
The piece is great, but the methodological issues with using an LLM for qualitative work are surely huge. Bias is mentioned as a reason to use it, but it's a black box. As far as I can tell, you don't know that any "analysis" has been done by the LLM, just that the results "look right".
November 21, 2025 at 8:52 AM
The piece is great, but the methodological issues with using an LLM for qualitative work are surely huge. Bias is mentioned as a reason to use it, but it's a black box. As far as I can tell, you don't know that any "analysis" has been done by the LLM, just that the results "look right".
I also note their objection to the low wages for MTurk works is "this means there is an incentive to lower quality, and maybe use an LLM themselves", not "maybe it is unethical for us to not pay fair wages for this work."
November 20, 2025 at 3:23 PM
I also note their objection to the low wages for MTurk works is "this means there is an incentive to lower quality, and maybe use an LLM themselves", not "maybe it is unethical for us to not pay fair wages for this work."
There is also an issue of getting the probabilities out of the LLM - how do the authors know these are robust and accurately represent the probabilities the model uses? Would also be interesting to replicate the study with and without the few-shot approach.
November 20, 2025 at 3:21 PM
There is also an issue of getting the probabilities out of the LLM - how do the authors know these are robust and accurately represent the probabilities the model uses? Would also be interesting to replicate the study with and without the few-shot approach.
Curious about the methodological implications of this, though. In both pieces, the assumption is that the LLM will be less biased and "better", but that is really taken on faith. In the linked piece, the authors seem to say "it cost a lot, and has a lot of parameters, so it must be good!"
November 20, 2025 at 3:14 PM
Curious about the methodological implications of this, though. In both pieces, the assumption is that the LLM will be less biased and "better", but that is really taken on faith. In the linked piece, the authors seem to say "it cost a lot, and has a lot of parameters, so it must be good!"
It was Blair's realism on climate that led him to switch tack in 2004-5 from renewables to nuclear - 20 years later, the UK does not have a new nuclear power station built.
November 20, 2025 at 1:59 PM
It was Blair's realism on climate that led him to switch tack in 2004-5 from renewables to nuclear - 20 years later, the UK does not have a new nuclear power station built.
To be fair, landed racial aristocracies voting for their leaders aren't democracy either.
November 17, 2025 at 7:47 AM
To be fair, landed racial aristocracies voting for their leaders aren't democracy either.
Absolute instrumentalism, with not a single thought of "what is the purpose of this system in the first place?"
November 14, 2025 at 12:33 PM
Absolute instrumentalism, with not a single thought of "what is the purpose of this system in the first place?"
Brilliant that in the piece, objections are automated, and the government is automating reading and responding to consultations. So if the applicants are automating the applications (and they are - they are trying to automate nuclear license applications, for example), the circle is complete.
November 14, 2025 at 12:26 PM
Brilliant that in the piece, objections are automated, and the government is automating reading and responding to consultations. So if the applicants are automating the applications (and they are - they are trying to automate nuclear license applications, for example), the circle is complete.
Not too dissimilar from current HMRC practice.
November 14, 2025 at 12:12 PM
Not too dissimilar from current HMRC practice.
Many media careers have been made out of expressing "actually it's complicated!" answers so that people feel clever by understanding them. That's affect, too, btw! And it's the job political candidates and leaders. It's what separates the Obamas from the Starmers, if we're talking centrists.
November 13, 2025 at 9:07 AM
Many media careers have been made out of expressing "actually it's complicated!" answers so that people feel clever by understanding them. That's affect, too, btw! And it's the job political candidates and leaders. It's what separates the Obamas from the Starmers, if we're talking centrists.
The piece makes the claim that complicated answers can't win on TikTok, but that is manifestly not true. "This one weird trick" and "ironically, it's the opposite!" are all social media tropes that lean in to projecting the image of a complex, counterintuitive explanation (regardless of content).
November 13, 2025 at 9:03 AM
The piece makes the claim that complicated answers can't win on TikTok, but that is manifestly not true. "This one weird trick" and "ironically, it's the opposite!" are all social media tropes that lean in to projecting the image of a complex, counterintuitive explanation (regardless of content).
Said it elsewhere, but entire branches of political theory treat politics as a contest over what "common sense" is. And that content can be more "rational" or less so. Presenting as the "rational adult in the room" is as much an image choice as "man of the people/common sense".
November 13, 2025 at 8:59 AM
Said it elsewhere, but entire branches of political theory treat politics as a contest over what "common sense" is. And that content can be more "rational" or less so. Presenting as the "rational adult in the room" is as much an image choice as "man of the people/common sense".
It's been 11 years since I submitted a PhD on how the Labour government got to "nuclear: back with a vengeance", and 20-odd years since they did that.
November 13, 2025 at 8:10 AM
It's been 11 years since I submitted a PhD on how the Labour government got to "nuclear: back with a vengeance", and 20-odd years since they did that.
Rather the mirror image of a common attitude among engineers of "why can't we just get the politics out of it and do my preferred solution, which is obviously best"
November 13, 2025 at 8:07 AM
Rather the mirror image of a common attitude among engineers of "why can't we just get the politics out of it and do my preferred solution, which is obviously best"