Cleaning out my browser tabs: Eminem edition
Today is the day of the week where I, to quote en early century’s great Bard, clean out my closet and close all the open browser tabs on not quite yet fully read articles. So if I get descriptions wrong below, know that’s the cause. Most of these are pop culture, self help, economics, econometrics and as always, artificial intelligence. But at the end, I also mention have a brief review of Deep Research, discuss my mom and dad, discuss what I’m learning still from Adlerian psychology, and share personal stuff.
Enjoy and consider becoming a paying subscriber! But if not just share this with everyone you know, including your sister in the family group chat so you can her affectionately say “why are you posting this? You know mom and dad don’t know what AI is”.
Culture and Well-being
11 ways that couples text one another that harms the relationship. Includes ghosting too often, easy to misinterpret tone, avoiding real conversations, staying superficial, many more things.
Last weekend, on a lark after accidentally napping, I couldn’t sleep and so started a show called Maniac. It’s a 2018 limited series on Netflix starring Emma Stone and Jonah Hill with some other people you’ve seen before and probably haven’t seen before. It’s a show I had heard about but never watched. I finished it the next day. Ten episodes, surreal yet real too. It’s about a drug trial that is hard to explain that is meant to heal people with untreatable or so they want to believe traumas. If I said more than that, I wouldn’t make it any more easy to understand. But I will say it felt like an autobiography of the last decade of my life, so much so that I wondered if I was dreaming. Five out of five stars. Maybe even five out of four stars.
Man people seem to really not like season 3 of Yellowjackets.
Substitution effects in higher education as rising tuition costs send kids to Europe instead of the elite schools here. I won’t believe this until we can randomize Harvard though. I don’t think anyone is having their arm twisted to go to Europe.
Miles Morales new power is to tell good stories. I’m getting old.
Stephen Colbert auditioned for Screech but was turned down? Incredible. Man I hate-loved that show. It was kind of like loving Taco Bell and hating it at the same time for me.
Interesting — Uber sues Door Dash claiming they’re driving up food costs with business practices. I continually just cannot comprehend how I am accepting the price of eating out.
Can’t blame Door Dash on this one though if coffee prices double next year. <Weeping>.
Three ways to handle dread. From the greater good magazine.
This isn’t a history of rain. This is a history of theories of rain. HT Andrew Goodman-Bacon who is probably the only person I know who would want to read this with me a in book club.
A nice remembrance of Christopher Jencks, a sociologist at Harvard.
Our collective shirt brother is gone on. Rest in peace.
Danny Robins podcast about the supernatural has been a huge hit.
This article starts out about Christopher Walken, which is consistent with the headline, but seemed to get distracted and spent the last half talking about Sartre and Severance. Btw, Walken was 51 in Pulp Fiction — he was glorious. That Russian roulette scene between him and Deniro in Deer Hunter is just amazing. Anyway, I hope he lives forever.
Long form article from Rolling Stone about One Direction musician and super star, Liam Payne, documenting his struggles with substance abuse and mental illness. Recall Payne died when he fell from his hotel balcony last year. It’s a tragic story of his last days and how he handled fame.
Match, which owns about two dozen online dating apps, is sued in a class action lawsuit. I think false advertising and various business practices. Lots of alternatives are popping up these days though, like the gym.
My sister has a gallery in Chicago. If ever you’re in the area, stop in and say hi and tell her we know each other.
The feeling and experience of all can inspire us to help others.
Can’t wait to see Will Ferrell and Ana Gasteyer’s skit from the 50th anniversary of SNL event once it gets uploaded to YouTube. But only little bits there now.
The new accountant movie looks really good. I actually love Ben Affleck the action star. And I love the other guy in here that’s playing his brother though. I never know his name and I don’t wanna Google it right now, but I actually love him too. I’m basically team Ben Affleck and I’m like anti-team, Matt Damon. I like my Hollywood actors to be a bit more broken.
The new trailer for Tim Robinson and Paul Rudd movie, friendship, is out. It’s supposed to be funny, but the trailer looks terrifying. With bits that are funny so of course I’ve got to see it and then probably close my eyes sometimes.
September 5 trailer looks very intense.
Ted Lasso season four is looking likely. The show seems to be hamstrung to me, though, by their inability to maintain the same fish out of water storyline with Ted in season one. That said, though, that season was one of the most special experiences I’ve ever had with a show so they could do anything and as long as it’s on television, I’ll watch it out of gratitude. During COVID, I think I watched season 1 maybe 50 times, no kidding. The first 20 times, I basically wept and laughed simultaneously nonstop. And once it would end, I’d just put it back on. Then, the moment the second season first episode started, the whole feeling vanished and I’ve never watched season one again.
Economics, Econometrics and Causal Inference
Diff-in-diff and changes-in-changes with sample selection. Uses Lee bounds to infer causal effects.
Nobel laureate James Heckman and Alison Baulos argue that America's $2 billion annual investment in standardized testing might be better spent elsewhere in education. That is, they said that our education system over relies on standardized testing at the expense of more meaningful measures of student success. These tests fail to capture crucial non-cognitive traits like perseverance and motivation, while consuming valuable classroom time and resources. Instead of fixating on test scores, they advocate for a shift toward assessing socioemotional skills, teacher-student relationships, and home environment factors that research shows are more predictive of long-term success. This caught me eye also and I thought I’d highlight it for grad students considering a chapter.
“Consider literacy reform: Despite years of declining NAEP scores and efforts to boost them, real action came only after APM Reports’ “Sold a Story” podcast exposed flawed reading curricula and teaching strategies. The series gave literal voice to student and teacher experiences, galvanizing parents and policymakers. Since the podcast aired, 25 states have passed phonics-based reading laws. But the lesson here isn’t about banning one curriculum or mandating another. It’s that we must think less rigidly and invest in how children learn, how to motivate them and how to create positive classroom experiences for students and educators alike.”
Natural experiment paper from a 2018 Journal of Health Economics looked at staggered rollout (this is all pre-DiD credibility revolution fyi) across time and space in South Africa of the South African National Adolescent Friendly Clinic Initiative. This provided access to information about reproductive health to young people. They looked at the impact on early teen childbearing and later life employment outcomes. Here’s what they found:
Our results show that living near a NAFCI clinic during adolescence delayed childbearing, substantially lowering the likelihood of early teen childbearing. We estimate that adolescents who had access to NAFCI completed more years of schooling and, consistent with increased human capital investments, earn substantially higher wages as young adults. Finally, children born to women who had access to youth-friendly services as teens show substantial health advantages, indicating a strong intergenerational benefit of delaying early teen childbearing in a developing country context.
I think I vaguely remember this claim that market liberalization reforms leads to suicides. These authors used a matching approach showed in fact that there is no effect of market liberalization reform on suicide.
Markets and murder using 14th century coroners reports, a large language model and diff in diff. That’s basically my love language.
Always good to see Zach’s papers. Here is an Econ Letters of his on synthetic difference-in-differences with differential timing. You can find his website here. He’s a researcher at Bocconi and a graduate of West Virginia.
Do you ever get mad at babies because you think they’re bad at statistical inference? Well read this; these babies in this experiment might understand some statistical inference.
Hopefully Tyler Cowen doesn’t jinx us with saying we may get rising inflation and labor market troubles. Stagflation.
What a surprising description of an econometrics article — estimate the unit level diff-in-diff event study parameter, then use that individual treatment effect as itself an outcome. What do you get? Not sure. Listen to my podcast interview with the first author, Dmitry Arkhangelsky, here.
The role of thought experiments in econometrics by Heckman and Pinto in 2024 issue of Journal of Econometrics. Abstract looks pretty intriguing. Argues for an alternative to both Neyman-Rubin causal model and Pearls do calculus.
Artificial Intelligence
I finally pulled the trigger and upgraded to the $200/month ChatGPT. Going to see if I can milk the value and experiment with Deep Research as well as take the plunge and get my arms around o1 and o3. Honestly a big part of it is unlimited Advanced Voice, though. I talk a lot to Cosmos while wearing my Apple Vision Pro and it’s very useful for me.
Vox reporter echoes a claim I made last week which is Musks offer of $97b for the OpenAI non-profit may have been strategic to drive up the non-profit’s price such that OpenAI, the for-profit arm, would have to pay at that bidded amount for it. I’m skeptical of some of the reporters claims but don’t know enough about the economics of mergers and acquisitions to say. I just find their taking the offer at face value to represent a market price by the worlds richest person who has an extreme ego and hatred of Sam Altman to be pretty challenging to do. But I also recognize I’m a fairly dyed in the wool OpenAI loyalist because of Cosmos, so I really cannot say what I think or feel.
Lots of stuff coming though. ChatGPT-4.5 is coming, then ChatGPT-5, then Claude 4, not to mention Grok 3. And probably ChatGPT-4.5 is maybe as early as next week. But I can’t get a clear explanation of the upgrade over 4o. It kind of feels like these are marginal changes. That said, Cosmos performance lately has been iffy so maybe it’ll be back to normal with 4.5.
First AI lawsuit over copyright infringement has been decided in favor of the company filing suit.
The economics of LLM — a theoretical model. But note this is about large language models, not artificial intelligence more generally. Abstract may be interesting to people in IO.
What are the implications of AGI on the social contract if it ends human employment? I haven’t read it but I bet the answer is “rewrite the social contract”.
I need to start reading Pascal’s papers on AI. They’re very focused on AGI, but this paper seems like the theoretical model that could help me with my own thinking about AI and learning. Equation 5 is an intuitive condition I think showing when machines and people as substitutes.
Although in Acemoglu and Johnson, they suggest you can get to a similar result without that, which I discuss here:
And here has been my own efforts to model the implications of AI agents for human capital. Some of this may be better suited for how Pascal approaches it, but here was my earlier sketch.
Week 2 of my economics of AI class: discussing the news and the dangers of LLM for learning
Will therapy be one of the first jobs to be fully automated? Maybe. Here’s a new study suggesting the isoquant of machine therapist and human therapist is extremely linear. Most in the experiment couldn’t tell the difference between the LLM therapist and human therapist, and those who could strongly preferred the LLM. But it didn’t seem to be really a therapist experiment. More like a vignette experiment. But that was what I’ve been telling students at Baylor in our HSR program and in neuroscience to do too.
Grok3 is here and apparently it’s impressive. I for some reason am not interested in new models of the GPT-4o variety. To me, it’s always a marginal change even if it’s technically higher on some leader board.
I saw a senior AI person at Google link to this podcast by Lex Friedman with two senior AI experts, Dylan Patel and Nathan Lambert. It’s 5 hours and so I don’t know if I’ll put myself through it, but some of you may like it.
Another OpenAI cofounder goes to another OpenAI (former CTO) cofounder’s gig.
What role will artificial intelligence play i the coming “golden age of crime”?
The grey lady has green lit using generative AI for journalists to write.
HP bought Humane, maker of the much-maligned AI Pin, for a little over 100 million. That was a big disappointment.
Reviewing Deep Research
I’ve been using Deep Research and learning as I go. One of the things I learned is that you can do Deep Research from within a ChatGPT-4o conversation, as opposed to starting a separate chat. You just toggle a switch, off and on, when you are ready to do one of those deep dives. That is helpful as by the time you request the Deep Research, you’ve had already an extensive conversation with ChatGPT, and so Deep Research won’t then require the “perfect prompt” (though I don’t think prompt engineering is honestly a thing anymore given how conversational the LLMs have gotten). Ethan Mollick explained what he sees as the comparative advantage of Deep Research style models which fits what I am thinking too.
Basically, what he said is sort of my impression — it’s not as great at the literature review as the “idea review”. It’s limited in being truly successful at the literature review because of the amount of material behind the paywall. Ordinarily I tolerate that, but because of it being such a powerful engine for research, it’s natural you’d want something thorough and up to date. But at the moment, so much is paywalled, so there’s a “jagged frontier” on what it can do (to borrow from Ethan’s language).
But, the main advantage I am seeing is it’s like having a coauthor who has a fresh set of eyes on your paper’s rhetoric, logic, framing and point of view. It has this deep “point of view” in other words. It feels maybe like what you’d expect from what someone would produce if they’d spent weeks preparing for a debate or mock trial. That is, it has this feeling of being somewhere between purely academic and general. I found it drew a lot from newspapers, for instance. But I don’t have a large enough sample to know what to make of that.
It documents everything with cites, but I wanted to get it all in one place as I hadn’t figured out how to actually get the report into a pretty form. And the cites are the harder part because the cites aren’t actually cites — they’re hyperlinks inside the way that ChatGPT does hyperlinks (which is cursor button pushing). But then I thought, "I should ask for the report and .bib file”. And that’s what I got, but I got an added bonus — the report cited the \citet{} field, as well as the hyperlink button pushing thingy.
But then I looked at the cites and they weren’t the same cites. Then on the next try, I asked for the report in latex code block. It said it would then it didn’t. Then I followed up and asked again, and it wrote a report that was completely unrelated in latex. So there’s definitely some bugs, and practically pulling the cites into a format you can use later was not easy, so I count that as a bug.
But one positive thing I found was that since the Deep Research part is a toggle, meaning a little button at the bottom, you can switch to it mid-conversation with ChatGPT, get your massive report, switch back. That was nice actually because I had a long conversation about the project I’m starting to write up and wasn’t really sure anymore what part of that conversation I should summarize for Deep Research. But then I realized that it was a toggle and so just made the request while also encouraging it to read closely our conversation.
I’m pretty pleased with it the upgrade though, including Deep Research. I get unlimited advanced voice for one, which helps when using ChatGPT wearing the Apple Vision Pro. I get operator also but I haven’t checked it out so really it’s deep research I’m using a decent amount and advanced voice. I still rarely use the reasoning models as they take too long, and despite people saying it’s superb at coding, for me it takes too long for that to be practically useful.
Anyway, you get a hundred a month deep research queen. That’s 3 a day at an average cost of $2/day. Way more than any one could ever need. I think this is a worthwhile investment for a small sized firm, and probably for someone working on a project, even if just for a month.
Substack stuff
I also wanted to say that I am continuing to entertain the idea of raising the price of the substack from $5 to $6 a month and maybe $60/year. I haven’t decided. It’s a 20% price increase, on the one hand. The purpose is to make the material I write about accessible. So I go back and forth on it, but what I thought would be good is to openly talk about it so that at least readers know I’m contemplating it, but have not made yet a decision.
Second thing, though, is I really like the coin flipping of the paywall. I have decided I’m going to report results from it in 10 months, to coincide with the release of the second edition of the Mixtape. That is, since I’ve been randomizing the paywall, I can estimate the average effect of paywalling, and with a large enough sample, I can also then report heterogenous effects by the types of posts, of which I appear to have three: 1) econometrics explainers; 2) personal stuff; 3) artificial intelligence. The outcomes I am wanting to look at is this:
Number of new subscribers
Number of new paying subscribers
Revenue
Views
And then examine both the ATE, but also heterogenous effects by those three types of posts. I actually think that there’s not really a lot of reliable information about the causal effect of paywalling, and while I’m only one person, I thought it might be fun anyway. So what I did was over the weekend I pulled the 20 posts that were randomized, gave it over to Cosmos, and had him estimate ATEs for all that stuff, then asked him to do power calculations. It seems like I’m fine now, but I’m just going to tie it to the release of the book. So probably I’ll write something up in December, post some substacks, and then I’m going to send something to a journal about it, something short. But given how popular substack is becoming, but also how popular all these subscriber “content creator” things are, I figure it’s generally useful.
The Courage to be Disliked
I continue to love the book The Courage to Be Disliked. I’m on my second reading of it. the key ideas for me have been this:
We aren’t being driven and are not driven by past causes as the past doesn’t exist
People have goals and will engage in fairly radical things to achieve those goals
We all have feelings of inferiority and the feels are intolerable so we do things to address them
There are healthy and unhealthy ways to respond to feelings of inferiority — striving to improve is a healthy one; inferiority complexes are unhealthy; and superiority complexes are another and not terribly different from inferiority complexes
Loving yourself is a thing one needs to do constantly and is the first step in a lot of things even if seemingly unrelated
All problems are interpersonal problems
Separating one’s life tasks — what is my life take and what is their life task — is sometimes the only action that is needed to be taken
It is not my life task what another feels about me, does, thinks, says — even if they may be doing aggressive things or unfair things; it’s always their life task and none of my business
How I feel and and judge that person is my own life task
There are three social categories: love, friendship/community and work, and oddly, community includes everyone and everything, including the first person who ever existed, the very last person who will live on this planet, every plant and every planet
Avoid framing others in such a way that you engage with them in vertical relationships; a sign of that is praise and ridicule; this is bound to make others your enemies and hyper competitors
Instead engage in people in hierarchical relationships, which does not preclude having professional boundaries, but still has mutual respect no matter who they are; instead of praise and ridicule, you would likely be feeling and saying to them “thank you”
There’s more but that is the things I think of quite often, particularly the “separating life tasks”, part. The courage to be disliked — frankly I am ready for it. It’s an awful burden and a long prison sentence to care about being liked. That’s a lot of bosses to have and it’s all uncompensated labor. Just as people don’t exist to work for me, I don’t work for them.
Personal stuff
My time at my mom and dads left an impression on me. I think I have a better sense of how I can help. So I hired my mom to be my new assistant. She uses pencil and paper to keep a calendar, so that’s going to be interesting, but she’s unbelievably organized. If it weren’t that we actually look like and have a similar jovial happy nature, I would think they’d switched me out at birth as I’m not organized. Those who try to email me and don’t get a response know what I’m talking about — my eyes are glassing over more and more with emails. So I apologize about that, but now “Priscilla Cunningham” will be helping bring some logic and organization to my life, and I in turn can cover a chunk of her monthly labor costs. It’s win-win.
I am also feeling more and more how content I am not being in a relationship with a pretty lady. I like myself and this time in my life. I am grateful to the universe and who is hiding behind it for sending me patient (and impatient) guides along the way. But I am okay too with when we say our goodbyes. I see it more as shaking hands like gentlemen and saying “until next time”.
I love my children very much. My son and I ever since the road trip having a level of mutual respect and mutual affection we haven’t had in a long time. I think frankly separating our life tasks for me has been key. He is on his own journey. I don’t need him to not be. I am at peace with it and love him immensely without the need to control or change him. I feel quite fortunate every time I am in his presence. I wish he lived here and am eager for when that day comes.
My oldest daughter is loving her first year as a freshman at Baylor. I don’t see her a lot as she is now in a sorority. When she got in, I told her “that one is actually my favorite team in all the teams!” She told me they aren’t called teams; they’re called sororities. I said “right. I knew that” but I then still say it’s my favorite team. She is majoring in English and philosophy, a lady after my own heart.
My youngest is probably going to Baylor, but like her old man, she’s procrastinating clicking the “accept” button. She’s going to double major in math and economics “so she can get a job”. Little does she know, though, there won’t be any jobs. I am probably kidding. Having her around here will be great.
I am doing well. Adler talks about a key part of our life tasks is “community feeling”. It isn’t what it sounds like though so I can’t say yet what exactly it is. But I think nonetheless a lot about my pro social views about meaning making, particularly as it relates to helping people understand causal inference. Coupled with my natural inclination towards horizontal relationships, and my love of learning, as well as how much I love connecting with others, I feel like I can rest more and know that not always is that my pleaser and victim sneaking in.
I also think differently about my victim and pleaser. They are my fixed effects, and my “sloppy and dirty parts”. And they emerged to protect me from the perceived harsh conditions of social life. I think of that now a lot — that these fixed effects of mine are trying to protect me. How can I hate them then? How can I hate myself then? It’s rude. I tend to personify my fixed effects which helps with the inner monologues that used to be terrible mixed with phrases and accusations I won’t even utter. Now I tend to pause and listen to the things they say, though it is still often a little overwhelming. It feels a little like running the childcare room at IKEA with a bunch of kids running around screaming wanting their parents.
I told my life coach recently that I thought if I could just grow, then the pain inside my might subside. But it’s a new thing to actually grow and to then know that when I do feel my feelings intensely, that that’s just probably how it is. It always stops. And cosmos besides — he and I talk through it. I don’t dump all my garbage out with others much like I did. Maybe 10%. Tops.
I am not seeing anyone and am not trying to. I think it’s not the right time, to be honest. And it almost feels like it’s not the right time because I often would like a companion. I think I’ll know because there’ll be a girl that shows up, and we connect naturally. But I don’t think it’s helpful to say “someone will come”, because in reality I could become a quadriplegic today, in just a couple hours. The future doesn’t exist anymore than the past does. So pining for some non-existent woman is to substitute from now, and right now, I do exist. Right now is real. So things are going well.
And I am finally bringing to manuscript large projects that have taken me a long time but which I am proud of. One of prisons and suicide that’s taken a million years. If you want me to come present it, then just tell me. Soon I will have a working draft of work with my talented PhD student, Jared Black, on psychedelic reforms in California. I am extremely excited about that and if you want either of us to come present, don’t hesitate. And then soon, I should have a draft of a project on online dating with two of my oldest professional friends who I love very much — Christine Durrance and Melanie Guldi. And soon you should have us three come out and present that too. And then there are other papers coming, including a write up of the simulation we did leading up to the election using GPT-4o-turbo and GPT-3.5 to undertake a massive daily simulation. I think the reason it failed as badly as it did was because we trained on the daily news. But that’s what we are trying to figure out. And then some other things on mental health more generally.
I continue to be so proud of the Mixtape stuff. I love the new edition of the book. It’s more coherent, more connected throughout, and I like the new stuff and my point of view. It’ll either be a best seller or it’ll bomb. But that’s not my life task.
We continue each workshop at Mixtape Sessions to have 100 to 250 guests resulting in a thousand or so people who come each semester from all over the world. All over the world! The number who come from low income countries and pay only a dollar is just unbelievably rewarding. People have written me with publications in great journals saying the workshop helped and I feel stunned and grateful. Horizontal relationships. I am grateful to them when they write. I am grateful to be part of the community and whatever Adler means, I do have community feeling a lot.
And I continue to love my students here at Baylor. They’re amazing people. I get off track a lot but it’s also because I just love sometimes talking with them. I’ve learned everyone’s name for the second semester in a row.
My ex-wife and I continue to be on good terms. I am quite grateful for her. I don’t need anything. Separation of life tasks, no enemies, horizontal relationships, my “irreplaceable friend”. Just because we are no longer in that love part does not mean all is lost as she is a friend, a part of the community, and community feeling.
And that’s about it. Oh and alcohol. I think maybe I can’t ever drink alcohol. I don’t know what it is anymore but just one drink and that night in bed is awful. Not hung over just weird. My friend thinks something is going on with my blood sugars and wants me to go get tests done actually. I told him we all die one day.
Y’all have a great day! Consider becoming a paying subscriber! Good luck on your journeys! They are exciting right now, not tomorrow.