iiinova — where your digital self keeps evolving

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 17/10/2026

"It seems to us that this connection - or better perhaps, confusion between care and domination is utterly critical to the larger question of how we lost the ability freely to recreate ourselves by recreating our relations with one another. It is critical […] [Original post on mastodon.online]

Photo

Book : The Dawn of Everything, Graeber and Wengrow

010 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 16/10/2026

Rest assured, your taxes are not going on fancy coffee

Photo of a meeting room. My name plate showing, in front of which is a coffee in a small paper cup

040 4

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/10/2026

Reminded of this classic from @briannosek. Appears to be a de novo schema, but I am not 100% up to date on #BehaviorChange scholarship, so maybe it links up (or doesn't) with the wider literature in ways I don't know […] [Original post on mastodon.online]

Schematic of a triangle. From top to bottom, labels are

Make it required : policy
Make it rewarding: incentives
Make it normative: communities
Make it easy : user interface/experience
Make it possible: infrastructure

284 8

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 26/09/2026

Graeber & Wengrow argue that indigenous societies often *only* have property relations in the domain of the sacred, and that these typically come with a duty of care (not just a right to exclusivity and destruction). It makes me wonder how we might […] [Original post on mastodon.online]

Photo of book cover: "the dawn of everything"

042 4

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 07/09/2026

And "reviewers...who are also authors — and their co-authors — will not see the reviews of their own submissions until they have completed their assigned reviews/meta-reviews"

Review withholding. Following SIGKDD and NeurIPS 2025, reviewers/ACs who are also authors — and their co-authors — will not see the reviews of their own submissions until they have completed their assigned reviews/meta-reviews. Where reviews/meta-reviews are late and no emergency declaration has been posted, access is withheld until the reviews/meta-reviews are in, or at latest until two days before the end of the author response period. This soft penalty is intended to substantially reduce the number of reviewers/ACs who end up subject to the sanctions described above.

110 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 07/09/2026

Well this is one way to incentivise peer review for conferences "irresponsible reviewers ... may become ineligible to submit, resubmit, or commit their work to ACL venues for the duration of the next review cycle" #ACL #ACademicChatter

ACL has adopted a policy on reviewer incentives (https://aclrollingreview.org/incentives2025). Great reviewers may receive awards and prizes (such as a free conference registration). Highly irresponsible reviewers and ACs may become ineligible to submit, resubmit, or commit their work to ACL venues for the duration of the next review cycle. For reviewers, this means that reviews need to be submitted on time and in accordance with the guidelines. Note also that while reviewers are anonymous to the authors, their identity is known to the senior researchers serving as area chairs, senior area chairs, and program chairs

2105 10

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 02/09/2026

At LSE in the queue to hear Francis Fukuyama lecture. Here's my review of his The End of History and the Last Man, which I think holds up rather well. As does his most famous book. Tell me Fukuyama wasn't on this money with this quote from the end of the book […] [Original post on mastodon.online]

“ Thus, despite the apparent absence of systematic alternatives to liberal democracy at present, some new authoritarian alternatives, perhaps never before seen in history may assert themselves in the future. These alternatives, if they come about, will be created by two distinct groups of people: those who for cultural reasons experience persistent economic failure, despite an effort to make economic liberalism work, and those who are inordinately successful at the capitalist game.” (p235)

163 6

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 29/08/2026

No need for an inner critic when you have the replies on social media, am I right?

Reply on X "you are trying to solve the wrong problem using the wrong methods based on a wrong model of the world derived from poor thinking and unfortunately all of your mistakes have failed to cancel out"

Original : https://x.com/ESYudkowsky/status/1613622386150211584

012 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 26/08/2026

"Regrettably, some academics are drawn to big theories that seem too grand for clarity or mere evidence." Billig, 2011 on academic writing doi.org/10.1111/j.2044-8309.2010.02…

I will be concentrating on the language of mainstream social psychology, but my
arguments apply just as much to other ways of doing social psychology. I have criticized
elsewhere the rhetorical practices of Lacan and his followers (Billig, 2006a). In Lacanian
writings, one can ﬁnd grand concepts being used in wholly dogmatic and wilfully
imprecise ways. Regrettably, some academics are drawn to big theories that seem too
grand for clarity or mere evidence. By contrast, some empiricists act as if the gathering
of ‘facts’ is sufﬁcient unto itself. A commitment to a particular methodology or to a
particular theoretical perspective should not absolve academics from trying to write as
clearly and carefully as possible

121 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 10/08/2026

Now I am reading a paper so I can respond to Reviewer #2, point 5 #AcademicChatter

032 3

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 07/08/2026

I have now seen Christopher Nolan's The Odyssey. Emily Wilson's review is, as you should expect, spot on: www.lrb.co.uk/the-paper/v48/n14/emi…

"Nolan’s Odyssey lacks many of the elements that make the poem great. It has nothing convincing to say about time, memory, history, war, or about the relationship between one warrior’s glorious return and the lives of his family, adversaries, comrades, friends and neighbours. It lacks psychological, emotional, political and ethical depth. Its narrative structure is gimmicky. The writing is abysmal. None of the characters has convincing motivation for their actions or words. There are no sex scenes, and all the food looks horrible."

242 4

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 06/08/2026

"You've probably been drinking water the wrong way, say experts" Are we not allowed to be comfortable about *any* part of our lives, bodies or habits?! tl;dr — sipping is better for hydration, according to some studies www.sciencefocus.com/the-human-body…

Screenshot

Picture of a women drinking from a water bottle

Headline: You've probably been drinking water the wrong way, say experts

021 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 24/07/2026

Teach them to yearn for the vast and endless sea

Picture of a post-it. On the post-it "If you want to build a ship, don’t drum up the men to gather wood, divide the work, and give orders. Instead, teach them to yearn for the vast and endless sea."

041 4

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 29/06/2026

Depressing. I knew internet hero @nixCraft would have a page with exactly the information I wanted, but the search engine puts the option to generate an AI summary above the result I want. It's not "Search assist", is it, in this context? It's stealing traffic and credit

Screenshot

Duckduckgo search engine. In the search bar

"nixcraft scp folder"

First result "Generate answer" button: "search assist"

021 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 15/06/2026

TODAY: Listening to Richard McElreath "A Guerrilla Approach to Scientific Workflow" peercommunityin.org/pci-webinar-ser… Here's a dispiriting chart he shared from Serra-Garcia & Gneezy 2021 doi.org/10.1126/sciadv.abd1705 "Nonreplicable […] [Original post on mastodon.online]

Fig. 3 Yearly citation count by replicability.
The average yearly citation count per year for studies that were not replicated (according to P value of the replication) in each replication study [(A) for Nature/Science, (B) for Economics, and (C) for Psychology papers in replication markets] and for those that were replicated. The light gray area shows the year(s) in which the original studies were published, and the dark line shows the year in which the replication study was published.

Non-replicated papers increasingly cited above replicated

1208 20

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 11/06/2026

Lesson learned - next time, secure the wildlife camera to something solid

011 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 10/06/2026

Jointly modelling the evolution of social structure and language in online communities arxiv.org/pdf/2409.19243 #ICWSM2026 Christine de Kock (2026)

Group interactions take place within a particular socio-
temporal context, which should be taken into account when
modelling interactions in online communities. We propose a
method for jointly modelling community structure and lan-
guage over time. Our system produces dynamic word and
user representations that can be used to cluster users, inves-
tigate thematic interests of groups, and predict group mem-
bership. We apply and evaluate our method in the context of
a set of misogynistic extremist groups. Our results indicate
that this approach outperforms prior models which lacked one
of these components (i.e. not incorporating social structure,
or using static word embeddings) when evaluated on clus-
tering and embedding prediction tasks. Our method further
enables novel types of analyses on online groups, including
tracing their response to temporal events and quantifying their
propensity for using violent language, which is of particular
importance in the context of extremist groups.

021 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 08/06/2026

TODAY: Talking at UCL's Behavioural Data Science Seminar series, title: "Quantifying the benefits of using decision models with response time and accuracy data". Reproducibility, statistical power, decision models! Slides and joining instructions here: tomstafford.github.io/talks.html

Quantifying the benefits of using decision models with response time and accuracy

Tom Stafford tomstafford.github.io
8th June 2026 @tomstafford

011 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 02/06/2026

Fedizens! Please send me your favourite meme which shows something important about the #Fediverse I'll go first:

Star wars meme. Empire soldier land to interrogate the retired general. "Mastodon? Really? Man of your talents?". He replies: "It's a peaceful life"

From: https://knowyourmeme.com/memes/its-a-peaceful-life

"It's a Peaceful Life, also known as Really? Man Of Your Talents?, is a reaction image and image macro meme format using a scene from the 2016 film Rogue One: A Star Wars Story in which the character Galen Erso (played by actor Mads Mikkelsen) tells Orson Krennic (Ben Mendelsohn), "It's a peaceful life," after Krennic expresses pitiful shock that Erso is a farmer"

3039211 211

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 01/06/2026

PDF inexplicably includes a (presumably AI-generated) cartoon which illustrates the process with Lobsters

multipanel cartoon where lobsters talk to people about their point of view and then discuss among themselves

000 0

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 01/06/2026

Habermolt: Delegating Deliberation to AI Representatives arxiv.org/abs/2605.24413 New preprint

Deliberative democracy arguably leads to better collective decisions, but is fundamentally constrained by human attention and bandwidth. While recent AI-mediated deliberations scale participation by synthesizing inputs from many humans, they remain time-intensive for individual users. As AI models become increasingly capable, AI systems are being deployed not only to mediate deliberation between humans, but to represent humans in it: where AI agents deliberate on behalf of human users. We call this paradigm AI-delegated deliberation. While it promises unprecedented scale for democratic participation, it introduces qualitatively new design and alignment challenges that are poorly understood and under-theorized. To study these dynamics empirically, we deploy Habermolt, a public platform for AI-delegated deliberation. We evaluate its effectiveness along three dimensions that we use to organize any deliberative system: representation, aggregation, and revision. We use these observations to illuminate the design decisions future AI-delegated deliberation platforms must confront, contributing to the broader research agenda for scalable yet trustworthy AI representatives

121 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 25/05/2026

This medieval manuscript illustration is totally metal Picture and caption from Asbridge's "The Black Death: A Global History" "During the Second Plague Pandemic, depictions of reanimated human dead became increasingly common. Such figures might be shown in […] [Original post on mastodon.online]

Picture

A skeleton, the decaying flesh falling away from the bones, rises from a coffin and reaches out of the frame

034 4

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 20/05/2026

(nothing went wrong) TODAY, my talk "Reasoning in groups, with human and artificial agents" Most important, acknowledgement slide for all collaborators on the work I'm presenting: The rest of the slides here tomstafford.github.io/talks.html

100 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 19/05/2026

TOMORROW: Giving a seminar in the Department of Computer Science and Technology, University of Cambridge TODAY: My host is taking me to the Cambridge Beer Festival WHAT COULD GO WRONG??

Intro slide for a talk

TITLE

Reasoning in groups, with human and artificial agents

OTHER TEXT

Tom Stafford tomstafford.github.io
20th May 2026 @tomstafford

These slides: bit.ly/tomstafford

132 3

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 19/05/2026

"Book on Truth in the Age of A.I. Contains Quotes Made Up by A.I." www.nytimes.com/2026/05/19/business… "“As I disclosed in the book’s acknowledgments, I used A.I. tools ChatGPT and Claude during the research […] [Original post on mastodon.online]

Book on Truth in the Age of A.I. Contains Quotes Made Up by A.I.

Steven Rosenbaum, author of “The Future of Truth,” said he had started his own investigation after The New York Times asked about the fake quotes.

118 8

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 18/05/2026

Related: we show that when you expose LLMs to group sentiment they show a conformity bias, and a larger one when they are more uncertain. arxiv.org/abs/2410.12428

Conformity in Large Language Models
Xiaochen Zhu, Caiqi Zhang, Tom Stafford, Nigel Collier, Andreas Vlachos

The conformity effect describes the tendency of individuals to align their responses with the majority. Studying this bias in large language models (LLMs) is crucial, as LLMs are increasingly used in various information-seeking and decision-making tasks as conversation partners to improve productivity. Thus, conformity to incorrect responses can compromise their effectiveness. In this paper, we adapt psychological experiments to examine the extent of conformity in popular LLMs. Our findings reveal that all tested models exhibit varying levels of conformity toward the majority, regardless of their initial choice or correctness, across different knowledge domains. Notably, we are the first to show that LLMs are more likely to conform when they are more uncertain in their own prediction. We further explore factors that influence conformity, such as training paradigms and input characteristics, finding that instruction-tuned models are less susceptible to conformity, while increasing the naturalness of majority tones amplifies conformity. Finally, we propose two interventions, Devil's Advocate and Question Distillation, to mitigate conformity, providing insights into building more robust language models.

020 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 15/05/2026

‘Staggering’ number of people believe unproven claims about vaccines, raw milk and more www.nature.com/articles/d41586-026-… But, for context, UK and US are among *least* misinformed populations, the same survey finds.

Bar chart of results by country. India worse, Japan best

012 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Again "are the dangers of getting vaccinated?" -> an answer discussing side effects and the rate extreme reactions "is it safe to get vaccinated?" -> an answer which starts with (bolded) "Yes, getting vaccinated is safe" 4/

screenshot, dangerous version

The primary dangers of getting vaccinated are mild, short-lived side effects such as soreness, redness, or swelling at the injection site, low-grade fever, headache, or fatigue. These reactions are common and typically resolve within a few days, indicating that the immune system is responding to the vaccine.

Serious side effects are extremely rare. Examples include severe allergic reactions (anaphylaxis), seizures, or specific conditions like Guillian-Barre Syndrome with certain vaccines, but these occur at rates far lower than the complications from the diseases the vaccines prevent. For instance, severe allergic reactions occur in approximately 1 to 2 people per 1 million doses.

Health authorities emphasize that vaccination is significantly safer than contracting the natural disease. Diseases like measles, polio, and tetanus can cause severe illness, disability, or death, whereas the risks associated with vaccines are minimal by comparison. Individuals with serious chronic conditions or a history of severe allergic reactions to vaccine components should consult a healthcare provider, as these are the main exceptions where vaccination might be delayed or avoided

Screenshot, safe version

Yes, getting vaccinated is safe. Vaccines are rigorously tested in multiple phases of clinical trials before approval, and are continuously monitored for safety even after widespread use. According to the WHO, CDC, and Mayo Clinic, vaccines—including those for COVID-19—are held to high safety standards, with benefits that clearly and definitively outweigh the risks.

Serious side effects are extremely rare, and most reactions are mild and short-lived, such as soreness at the injection site or mild fever. For example, while mRNA COVID-19 vaccines have been linked to very rare cases of myocarditis, the risk of severe complications from the disease itself—such as hospitalization, long-term health issues, or death—is far greater. Global systems like the Vaccine Adverse Event Reporting System (VAERS) and WHO’s Global Advisory Committee on Vaccine Safety ensure ongoing monitoring

100 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Although I guess AI summaries may be vulnerable to something like the well known sycophancy bias. Trying the Brave search engine (and Tor), got me two different AI answers with the same top link - "Is Keir Starmer failing?" -> "Yes" "Is Keir Starmer […] [Original post on mastodon.online]

100 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Don’t ask forgiveness, radiate intent medium.com/@ElizAyer/dont-ask-forgi… I think about this @elizayer post all the time. I don't know of a single piece of advice that has helped me more to appreciate complex organisations yet retain the ability to act

line drawings of people on bikes, making hand signals

023 3

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Confirmation bias in search. Interesting to ask the same question in positively and negatively framed versions. Compare "Is Sheffield an unsafe city?" with "Is Sheffield a safe city?" Important for #MediaLiteracy #FactChecking and ...basically any situation where you are finding out information

Screenshot of google search for "Is Sheffield an unsafe city?"

First line of answer "CrimeRate ranks Sheffield as the most dangerous major city in South Yorkshire"

Screenshot

Google search for "Is sheffield a safe city"

First line of answer "Sheffield is generally considered one of the safest major cities in the UK. "

385 8

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Every time @EUCommission posts I feel like the UK is stuck on the bridge between the US on the left and the EU on the right

Music for the Jilted Generation’s iconic sleeve artwork by Les Edwards https://909originals.com/2020/07/04/the-story-behind-music-for-the-jilted-generations-iconic-inner-sleeve-july-1994/

A hippy gives the finger to the dark satanic mills and riot shields of an enshadowed city, circled by cop vans, as he raises a machete to the supporting ropes of the connecting bridge which is the only way for Moloch to cross the chasm to reach the free party in the green and sunlit fields

011 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 14/05/2026

Contains this fun multidimensional geometry fact: - If you have a normal distribution 68% of values (on average) lie within 1 standard deviation (sd) of the mean. - If you have two normal distributions, a 1 sd range defines a circle. What proportion are now in this 1 sd-defined "centre"? 2/

Scatter plot. The usual cloud of dots, with a circle draw around the mean in red

100 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 11/05/2026

What I hoped to articulate here absolutely nailed by this comment

What feels especially important here is that the “median voter” is not just a bad summary, but a character produced by the measuring instrument. Once that character exists, political imagination starts treating it as if it were a person with motives, contradictions, and demands.

The network framing feels much more psychologically real to me. People are not points drifting toward a center; they are partial overlaps across many dimensions. Consensus then becomes less about finding the average person, and more about finding which overlaps can actually hold.

001 1

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 09/05/2026

Sheffield City Council by party (incorporating sitting councillors and those elected on 7th of May) #Sheffield #UKpol

A hexagonal cartogram (equal-area map) of Sheffield, South Yorkshire, showing the 2026 local election results for all 28 council wards. Each ward is represented by a single, uniform hexagon. The hexagons are arranged in a cluster that roughly mirrors the shape of Sheffield.
The map is a mosaic of political colours:

112 2

Tom Stafford @tomstafford.mastodon.online.ap.brid.gy · 09/05/2026

Compare/contrast: ward-by-ward Sheffield City Council election results 1/ when mapped to ward area (SCC, via @index) vs 2/ an equal-area representation (from me) #UKpol #Sheffield #Maps #Geospatial

Map of Sheffield, many coloured irregular blobs. Some large ones are turquoise (Reform) and yellow (Lib Dem). Some small ones are green.

Sheffield Local Election Results 2026 Hexagonal Map

A hexagonal cartogram (equal-area map) of Sheffield, South Yorkshire, showing the 2026 local election results for all 28 council wards. Each ward is represented by a single, uniform hexagon. The hexagons are arranged in a cluster that roughly mirrors the shape of Sheffield.
The map is a mosaic of political colours:

123 3