Wherever you’re reading this, I appreciate you! I’m travelling a lot this month and the pace of posts might be a little slower than usual, but I hope for it to pick up by month’s end. For now, some items of interest:
AI Ups and Downs
In links for April I talked about a woman who was fooled when criminals used an AI-generated clone of her daughter’s voice to fake a kidnapping. Another way to use AI technology for ill-gotten gain is to manipulate the market with fake news of disasters befalling a country, region, or a particular company. Consider this example by way of Zvi: A faked photo of an explosion at the Pentagon seems to have led to a substantial drop in the S&P 500 that reversed when word got around the image was bogus. Unscrupulous short sellers could have a field day with this sort of thing.
Arnold Kling discusses a more honest application of AI generated sound and picture: AI generated house concerts. The basic idea is that a band might license AI generated versions of themselves that you could engage to play private concerts right there in your living room, maybe with some sort of customizability like you get to do guest vocals on one of the songs. (Now picturing myself doing the Henry Rollins bit in Tool’s “Bottom.”)
An application closer to my day job is to use AI for coding data. People have been using data scrapers for a while now, but as far as I know they’re limited to pulling specific terms or numbers that are already in the sources — they can’t actually make determinations like whether the defendant and victim in a court case were intimates or strangers.
It looks like the new language models are moving beyond that, though. Thus David Rozado and colleagues recently published a paper based on using such a model to classify different sentiments in news headlines. Link above is to his Substack; here’s the abstract from the published version:
This work describes a chronological (2000–2019) analysis of sentiment and emotion in 23 million headlines from 47 news media outlets popular in the United States. We use Transformer language models fine-tuned for detection of sentiment (positive, negative) and Ekman’s six basic emotions (anger, disgust, fear, joy, sadness, surprise) plus neutral to automatically label the headlines. Results show an increase of sentiment negativity in headlines across written news media since the year 2000. Headlines from right-leaning news media have been, on average, consistently more negative than headlines from left-leaning outlets over the entire studied time period. The chronological analysis of headlines emotionality shows a growing proportion of headlines denoting anger, fear, disgust and sadness and a decrease in the prevalence of emotionally neutral headlines across the studied outlets over the 2000–2019 interval. The prevalence of headlines denoting anger appears to be higher, on average, in right-leaning news outlets than in left-leaning news media.
In another Substack piece Rozado uses ChatGPT to code headlines as pessimism, optimism, or neutrality. The results showed pessimism increasing over time, with big spikes in the 60s/70s and starting around the late 90s.
Economics Ain’t So Bad
Sociologists tend to throw shade at our sister fields, so as to justify our own distinctiveness. Regarding economics, the usual criticisms revolve around the limitations of the rational choice paradigm and oversimplified models based on homo economicus. But economics does tend to attract bright and serious people — the role of mathematics in theory sometimes strikes me as cargo cultish, but math requirements tend to keep out the dummies and activists. And if this post by Cremieux Recueil has any validity, their empirical work has substantially less p-hacking than statistical analyses in many other fields.
Note also that “psychology and sociology” — which are lumped together — come out ahead of “informatics, mathematics, and physics.” That actually strikes me as a bit odd, given all the N = my-intro-class studies in social psychology, versus the propensity for fifth decimal place accuracy in physics, so I wonder if the way fields are lumped and split here is muddying things up.
For another win for economists, see Alex Tabarrok’s we told you so regarding the effects of lost Russian gas supply on the German economy. Short story is that a model published by top economists was more accurate than the doomsaying of politicians. The reason doom didn’t come is that normies underestimate elasticity and substitutability:
When the Chancellor and the average person think about a 40% reduction in natural gas supplies, they implicitly assume that each natural gas-dependent industry must cut its usage by 40%....
When the economists replied that there were opportunities for substitution they were typically met with disbelief and misunderstanding. The disbelief stemmed from a lack of appreciation of the many opportunities for substitution that permeate an economy….
The misunderstanding came from thinking that we need every user of fuel to find substitutes. Not at all! In reality, as fuel prices rise, those with the lowest substitution costs will switch first, freeing up fuel for users who have more difficulty finding alternatives. Just one industry with favorable substitution possibilities, combined with a few moderately adaptable industries, can produce a significant overall effect.
I’ve made it a point lately to learn a little more about economics, even though it’s not close to my normal specialties. Along those lines, here’s Arnold Kling explaining how banks work.
Speaking of money matters….
WVU Budget Crisis
I’m not one to air the family’s dirty laundry in public, but it’s getting a little hairy around here. After announcing a $55 million budget shortfall, the administration is talking about downsizing, and is sending us notifications about updates to our severance packages. Which is why an anonymous group of WVU employees wrote a letter arguing that our administration is bloated, expensive, and is doing a questionable job if it indeed got blindsided by a budget crisis of this magnitude.
I suppose the implicit appeal is for the administrators to fall on their swords and at least take paycuts or trim their staffs before they jettison the ones who actually carry out the university’s manifest functions — teaching and research. But I wouldn’t hold my breath.
It Brings On Many Changes
This week I gave a talk on “The Social Causes and Consequences of Suicide” at Kyungpook National University in Daegu, South Korea. Special thanks to sociologist Sohoon Yi for arranging the event and for being a gracious host. Thanks also to her colleague Hyoung-jin Shin for joining us for lunch and coffee. The faculty and students present had good and astute questions.
I’ll be giving a similar talk next week at the Ulsan National Institute of Science and Technology. Thanks in advance to anthropologist Bradley Tatar for his help arranging that event.
Both talks summarize a lot of material from my book on Social Causes of Suicide.
One social cause is downward mobility, as when someone loses wealth or other kinds of social stature. We can see this in the case of a college student who killed himself in 2020 after logging into the trading app Robin Hood to find his balance now $730,000 in debt. The wrinkle in this case is that he killed himself over false information; the displayed balance was the result of a glitch.
One social consequence of suicide is to find someone to hold liable for the death of a loved one, even if it was at his or her own hands. In this case, the family of the student sued Robin Hood for wrongful death and unfair business practices. The company recently negotiated an out-of-court settlement.
Such responses to suicide are of greater interest to me lately, as I’ve joined sociologist James Tucker’s ongoing project of studying them. Some years ago, Tucker wrote a good paper on the legal punishment of suicide (full text paywalled). If you wonder how authorities can punish someone who has already killed herself, consider an example from the paper’s introduction:
….The ecclesiastical authorities of the Inquisition tribunal of Ciudad Real concluded that this was a suicide and that Isabel should be “punished” accordingly … [T]he inquisitors pronounced Isabel excommunicate and damned her memory (en memoria damnada). They went to the length of ordering the exhumation of her bones, which were broken into pieces and then [turned over] to the secular authorities. El brazo seglar – in the person of the public executioner – burned Isabel's shattered bones, thus eradicating all physical evidence of her existence. Subsequently, the authorities deprived her heirs in the male and female lines of their rights to her property and barred her descendants in perpetuity from a number of private occupations, including those of pharmacist and attorney. They were forbidden to receive honors of any kind whatsoever, to own or even ride horses, to carry guns, or to own any precious metals, gems or rich fabrics, all in perpetuity.
For a shorter item on the topic: When Suicide Was Illegal.
Crime and Policing
Speaking of crime, for all my libertarianish distrust of armed state agents, we do have quite a bit of evidence from criminology that policing reduces crime. A new paper in Criminology and Public Policy claims that even a classic study thought to have found a null effect of police patrols, when reanalyzed to correct flaws in the original, shows evidence of deterrence. From the abstract:
The Kansas City Preventive Patrol Experiment (KCPPE) was seen by its developers to have produced “consistent evidence of the lack of effects of any consequence on crime,” a conclusion that was to have a strong impact on assumptions about police patrol for almost half a century. We identified the original official crime data from the KCPPE, and reanalyzed outcomes focusing on a comparison of the “proactive” versus “control” beats (“reactive beats” were criticized because of violations of treatment integrity); examining broad categories of crime (to increase statistical power); and using count regression models. Our findings are not unequivocal, but point to modest impacts of police patrol on crime in police beats.
A randomized controlled experiment is the best sort of evidence for this sort of thing, though it seems like even casual observation might get you to the same conclusion:
Screencaps from sociologist Peter Moskos.
Not all crime is stick-em-ups: Some people get ripped off by confidence men. An older relative’s encounter with such a scammer has me thinking about the techniques con artists use and who is most susceptible to them. One technique is to create some urgency — buy now, the stock won’t sit at this price long! Rob Henderson’s observation on Twitter that sensitivity to social comparison is correlated with a fear of missing out makes me wonder if that sort of con plays better in places (like parts of East Asia) where people are relatively more attuned to social comparison. But at the moment I can’t get through the paywalls to the articles he cites, so maybe “fear of missing out” means something different than what I’m thinking. And to gauge the real effect of anything on falling for a scam, seems like you’d need to control for intelligence.
Mortality and Health
One of the most shocking things I’ve come to appreciate as an adult is the insane levels of child mortality in the premodern world. For all my gripes about modernity, it’s the immediate cure to any romanticizing of traditional society. Sure, I’d probably have more children if I were middling freeholder Hugh of Norwich, but I’d almost certainly have buried one or two by now.
I can barely fathom how different life is in a world where such loss is normal and expected, even though that was the norm for almost all human history and I’m the one living in the weird outlier society. Over at Scientific Discovery, Saloni Dattani discusses the scale of the changes in mortality. Some highlights:
Globally, 27% of children died before they were 15 years old in 1950.
But that figure became 4.3% in 2020, which is a 6-fold decline in seventy years. For some wealthy countries, the decline has been more than 100-fold over a couple hundred years.
In total, between 20–40% of women surveyed had lost a sibling and a child under five in most low- and middle-income countries, in the 1960s.
Women in France who died at the age of 90 in 1900 were in the top 1% of survivors in their cohort. But by 2018, they were only in the top 40%. Now, the top 1% are those who die over the age of 100.
80 year olds in 1900 France had only 10% of their birth cohort still alive with them. 80 year olds in 2018 had 75% still alive with them.
Also from Saloni: Seasonal flu isn’t seasonal everywhere.
Assorted Books
I don’t only read about sociology, suicide, and race riots. In fact, lately I’ve been consuming science fiction on Kindle.
Robert Kroese’s Codex Babylon is a time travel story about a modern man who has to gain some information from the Knights Templar back in the 12th century. Also, social media engagement algorithms are literal demons adapted to the modern world, which is an interesting twist on David Chapman’s “unaligned AI is already here” argument.
Fenton Wood’s Hacking Galileo is a spin on the 80’s teen sci-fi adventures. I was reminded of The Explorers, the movie where River Phoenix and his buddies make a spaceship out of spare parts. It’s hard sci-fi engineering procedural, which is the sort of thing I might find boring. But the book worked for me, maybe because it was well-paced and had enough weird revelations.
Since becoming a father I’ve developed strong opinions about children’s books. After saying so on Twitter, Jonathan Tweet was kind enough to send me a complimentary copy of his children’s book on evolution: Grandmother Fish. I like that it’s very much written for children, keeping it simple and without adult-oriented snark or “religion of science” woo.
Thanks for reading! To support Bullfish Hole, you can leave a tip in the Tip Jar, and/or become a subscriber at the link below.
Substacks referenced in this post: