Pieces of the puzzle
Reasonable People #50: snippets on reason, bias and persuasion. And Tetris
Hello new readers. Normally I write an essay, but today it is just a collection of things I’ve been reading, writing and thinking about.
Tetris: How a US teenager achieved the 'impossible' and what his feat tells us about human capabilities
From me, on BBC Future, published on Wednesday.
It's about how a teenage achieved the impossible in an arcade game nearly three times older than him, about how communities support learning, and about why I'm optimistic about human flourishing in the digital age.
Things that didn't make it into the article
- quote from Alexey Pajitnov, the creator of the original Tetris: "For me, Tetris is some song which you sing and sing inside yourself and can't stop."
- Blue Scuti took his handle from one of the largest stars in the known universe: UY Scuti, which is 5 billion times bigger than the sun and 5900 light years away
- Link to article in 404media: 13-Year-Old Becomes First Person to Ever Beat Tetris
- The psychology of Tetris Why Tetris is compelling (“it takes advantage of the mind's basic pleasure in tidying up - and uses it against us”, me, from 2012)
- Thanks to Richard Fisher for being enthusiastic and a great editor
BBC.co.uk : Tetris: How a US teenager achieved the 'impossible' and what his feat tells us about human capabilities
PAPER: Political Fact-Checking on Twitter: When Do Corrections Have an Effect?
Warning: just to highlight that this is from 2018 - I include thing based on when I notice them, so there’s a mix of old and new.
Margolin & Hannock report:
“we find that individuals who follow and are followed by the people who correct them are significantly more likely to accept the correction than individuals confronted by strangers. … These findings suggest that the underlying social structure is an important factor in the correction of misinformation.”
The difference is large - people are twice as likely to accept a correction from a connection rather than a stranger.
Poytner summarised the work as On Twitter, you’re better off fact-checking your crazy uncle than a complete stranger and quote lead author
“The idea that it’s actual people who could have a relationship with you, instead of just some sort of machine, is really important,” Margolin said. “That suggests, ‘What is the goal or intent of this correction? Who is behind this, why are they doing it?’”
As we see again and again, there is no realm of pure reason. How things like fact-checking are socially embedded really matters
Margolin, D. B., Hannak, A., & Weber, I. (2018). Political fact-checking on Twitter: When do corrections have an effect?. Political Communication, 35(2), 196-219.
NICK BYRD: Which matters more: The size of language models or the psychology of language models?
What he means by “psychology” is architecture. Nick argues that the evidence shows that dual systems language models, one’s that temper their “intuitive” responses with reflection/checking outperform single system language models.
These results not only confirm the diminishing returns on increasing the size of large language models. The results also suggest that before language models became as large as they are today, their size may have mattered less than their psychology. This means that AI companies may need to compete on psychological architecture rather than size. For psychology, this serves as a reminder that our reasoning strategies may be at least as important as other cognitive factors like cognitive capacity
Psychology Today: Which matters more: The size of language models or the psychology of language models?
TALK: Chatbots for Good and Evil
Now open, Kevin Munger’s keynote at #EACL2023. Full of bon mots, and a critical take on the rush to develop human-like language models. What’s not to like??
Kevin Munger: https://underline.io/lecture/72154-chatbots-for-good-and-evil
PAPER: Conversations with a concern-addressing chatbot increase COVID-19 vaccination intentions among social media users in Kenya and Nigeria
During mass vaccination campaigns, social media platforms can facilitate the dissemination of public health information but may also contribute to vaccine hesitancy by serving as a vehicle for the spread of false and misleading information. Although talking with health professionals is an important avenue to address individuals’ concerns, one-on-one conversations with healthcare providers are challenging to scale. Can automated, personalized messaging delivered by a chatbot address individuals’ concerns and increase vaccine acceptance? To answer this question, we designed and deployed a Facebook Messenger chatbot to address questions and concerns social media users in Kenya and Nigeria had about the COVID-19 vaccine. After optimizing messaging using an adaptive experimental design on 3,905 respondents, we compare the interactive concern-addressing chatbot to a chatbot that delivers a non-interactive public service announcement (PSA), as well as to a control, no information, chatbot condition. We find that the concern-addressing chatbot increases COVID-19 vaccine intentions and willingness by 4-5% compared to the control condition, and by 3-4% compared to the PSA intervention. Among the 22,052 respondents in our evaluation sample, who at the time of the survey in early 2022 had not yet received a single COVID-19 vaccine, we observe the largest treatment effects among those most hesitant at baseline. With advertising costs as low as $0.21 per person engaged and $4.33 per person influenced, policymakers may want to consider using personalized messaging on digital platforms to quickly and cheaply reach many people to encourage compliance with public health programs during disease outbreaks
Rosenzweig, L. R., & Offer-Westort, M. (2022, July 14). Conversations with a concern-addressing chatbot increase COVID-19 vaccination intentions among social media users in Kenya and Nigeria. https://doi.org/10.31219/osf.io/mgyxu
Catch-up service
My last newsletter, from between Christmas and New Year …
Which I summarised on mastodon as:
Fresh from the refrigerator of cool takes, I consider that the real danger of conspiracy theories is the harmful way of thinking they ask you to buy into, not the wrong-headed conclusions. The flip side of this - that you can't get away from - is that dismissing every fringe belief as obviously false is also a harmful way of thinking.
Essentially, I am imploring us, like Warpaint, to
Keep It [Epistemically] Healthy
Read: Complex problems and the rush to judgement
TALK: Adventures in the Multiverse
This is me, details below. Outside the talk, if you’re in Bristol around that date and would like to talk about cognitive science let me know.
9th February 2024, Adventures in the Multiverse at TARG, University of Bristol
Complex data and complex methods mean multiple statistical analyses are potentially legitimate. Multiverse analysis offers a tool to interrogate the stability of results under different analysis choices. After introducing the general history and rationale of multiverse analysis, I will talk about two multiverse projects I have been involved in, both in my research area of bias and decision making, and what we have learned from them
Slides: "Adventures in the Multiverse"
So what about those Nazis?
Substack - who host this newsletter - are facing growing criticism for allowing Nazis to make money on their platform. Here’s where I’m at with this:
This is not hyperbole, it is not people who are only being called Nazis by the hysterical. I checked. These newsletters are by/for actual Nazis - swastikas and kill-the-Jews Nazis.
There’s no one better than Casey Newton to articulate why Substack’s position is untenable.
I’m expecting Substack to change course and get their act together around moderation. If they don’t, I’ll need to make a plan. Not just because I don’t want to encourage the Nazis monetising hate speech, but because it would signal clearly the direction Substack was taking and its better to get out before the enshittification hits the fan
Dave Karpf is in a similar position to me. I don’t charge subscriptions, and I know that hosting my own newsletter would incur costs (either server fees, or configuration time).
Substack VC subsidises a newsletter platform which I can use for free. The choice isn’t “Substack” vs “another platform”, it is really “Substack” vs “stop writing newletters” vs “another platform and ask for subscriptions”. If you have any thoughts on what I should do, let me know in the comments.
And finally…
Comments? Feedback? Ironic points of light? I am tom@idiolect.org.uk and on Mastodon at @tomstafford@mastodon.online
END
Read Jesse Singal for real investigative reporting on Substack's nazis.