Steven Quartz PhD

Dopamine isn't just about individual reward learning. In 2005, my colleagues and I found that a dopaminergic reward prediction error signal extends to social exchange. People use this signal to build a model of others as they develop trust and cooperation. Once formed, this becomes a predictive, anticipatory 'trust signal'. PMID: 15802598 (illustration: cover of science issue featuring the study, which required developing an fMRI 'hyperscanning' platform to probe multiple brains interacting in real-time).

@hubermanlab pubmed.ncbi.nlm.nih.gov/3463980/

Also Ralph Linsker’s series of papers in PNAS in 1986 was the first to show how such a Hebbian mechanism could produce neural organization with random spontaneous activity - foundational work in computational neuroscience. the first in the series: ‘ From basic network principles to neural architecture: emergence of spatial-opponent cells.’

A common quote misattribution that needs to be corrected. “Fire together, wire together.” Are NOT the words of Donald Hebb (as people mistakenly claim). Dr. Carla Shatz is the one who said that first. (in reference to activity dependent wiring of the developing brain.)

1. The Cochrane review highlights the lack of evidence for ketogenic interventions among adults with epilepsy, 2. for childhood epilepsy, the Cochrane review concludes the certainty of the evidence to be low to very low. 3. The autism study had an attrition rate of 75%. In epilepsy, about 2/3 of people abandon a ketogenic diet within a few years.

@ChrisPalmerMD Yeah, it's truly amazing how anything works in small, single-center trials w/ high risk of bias.

É impressionante como dieta cetogênica funciona tão bem em pequenos estudos não-controlados! 🤯

@jennymatt73 @BenBikmanPhD Glucose is an obligate substrate for critical brain processes. Severe hypoglycemia can lead to seizures, coma, and death. Ketones, by contrast, are a contingent, compensatory fuel.

@BenBikmanPhD So given the brain can use fuels other than glucose, are there parts of the brain that must have glucose, like red blood cells? And if there are plenty of ketones available for brain, how much glucose is needed in addition (exogenous or endogenous)?

Glucose is such a fickle fuel. No other nutrient experiences the swings glucose does, rising and falling dramatically throughout the day. And the brain appears to suffer the most, undergoing periods of feeding frenzies and starvation within hours of each other. pubmed.ncbi.nlm.nih.gov/39044609/ Give your brain a break. Controlling carbs not only stabilizes glucose availability, but also introduces ketones. Ketones, incidentally, are the brain's preferred fuel. If given equal access to glucose and ketones, the brain selects ketones at about 2/3 the rate of glucose.

These results are based on only 10 completers after 75% attrition, making them not only limited in generalizability but also highly susceptible to survivorship bias. The high attrition rate may be this study's most important finding: dropouts aren't noise; they are part of the treatment-effect profile, and attrition at this level indicates serious challenges for the intervention’s feasibility and scalability, a pattern that appears across ketogenic intervention studies.

How does high-intensity interval training (HIIT) compare with moderate-to-vigorous continuous training (MVICT)? Our new meta-analysis synthesises 115 RCTs (n = 3196) across populations from sedentary adults to elite athletes. DOI: bit.ly/4rJXGgt Key findings below 🧵

Nice to see someone summarise it well. They talk about the Norwegian training method. Well this is the South African training method. 😜

This is wrong at 3 levels. 1, it relies on the ‘great men of history’ framework, which is more outdated literary trope than serious historiography. 2, it's psychologically shallow because it conflates introspection and self-doubt: many actors are formidable because they are strategically self-aware. 3, it assumes a lawlike relation between personality and historical outcomes, but that's undermined by equifinality: similar outcomes can be produced by very different personalities even under the same structural conditions. So, what's needed is less sloppy psychology, less literary hero-worship, and a more rigorous account of historical causation.

@REV_Insulin_Res @ScottAppliedSci is making a documentary 'that could upend everything we think we know about cholesterol' based on an unvetted study that's now retracted acting in good faith?

@StevenQuartz @ScottAppliedSci Is this a good faith question?

The misinformation superhighway.

@ScottAppliedSci @REV_Insulin_Res is this the documentary based on a retracted paper?

@REV_Insulin_Res They’re talking about something the content of their speech shows they just don’t understand.

The lead researcher on this study has a note on her project page asking people to stop using words like "brain damage," "brain rot," "harm," "destroy," and "collapse" to describe her findings. This tweet is doing exactly what she asked people not to do. Here's what they actually tested. 54 people wrote SAT essays over 4 months. One group used ChatGPT, one used Google, and one used nothing. Researchers tracked brain activity continuously with EEG caps. The ChatGPT group showed the weakest brain connectivity. By session three, most were just pasting ChatGPT's output. 83% couldn't quote a single sentence from the essay they'd just "written." Two teachers who graded the essays called them "soulless." That part is real. But session 4 changes the picture. They swapped the groups. ChatGPT users had to write solo. Brain-only users got ChatGPT for the first time. The brain-only group's neural activity actually went up when they used ChatGPT. Their brains treated it as a collaborator, not a crutch. Three sessions of thinking independently meant they had something to bring to the table. The paper is a 206-page preprint. Still not peer-reviewed. A formal academic response in December 2025 flagged issues with the sample size, EEG methods, and the reporting of results. This pattern isn't even new to AI. A 2011 Columbia study in Science found people who knew they could Google something were less likely to remember it. GPS users show steeper declines in spatial memory over time. Calculators, cameras, even writing itself. Every tool humans have built changes what our brains bother to store. The actual finding: passive, copy-paste use of ChatGPT during essay writing reduced brain engagement over 3 sessions in 54 college students. People who already knew how to think independently used ChatGPT without losing anything. The researcher's own term is "cognitive cost," not "cognitive destruction."

@MattLutzPhi it reflects the larger problem that most of the scenarios used in moral psychology/neuroscience lack ecological validity.

I'm actually pretty sympathetic to Haidt here, but it's genuinely hilarious that this is an actual famous study, and I'm not exaggerating at all.

Jon Haidt: Suppose someone were to have sex with a chicken, then kill it, cook it and eat it. Would that be wrong? Normal person: Uh, yeah! Haidt: But why? Normal person: That shit is gross. Haidt (writing in his notepad): The savage is unable to articulate genuine moral reasons.

How many more meta-analyses of this stuff do we need? This one (k = 153, N = 18,933 kids) finds the usual tiny/non-significant *correlations* between social media use and bad things (r ≈ 0.10), explaining ~1% of the variance #smma jamanetwork.com/journals/jamap…

It’s not that crazy… 15k steps per day of low-level activity, in the absence of any other exercise training, isn’t really that active. I wouldn’t expect EE to be much higher than the average American. (Kudos @zbitter for pointing this out)

@jonatanpallesen @MattWalshBlog identity theorists are hardly religious or superstitious - it's a reductive materialist view. Those who thik AI will be conscious are committed to multiple instantiability, which is far more metaphysically 'spooky.'

@MattWalshBlog The only reason to think that AI will never have consciousness or any other trait is if you think there is something magical about the human brain. Either due to religious reasons or other superstition.

This is dumb. AI can’t ever be actually conscious because it doesn’t have subjective experience. It isn’t like anything to be AI. There is no experience there. Consciousness is the awareness and experience of self. AI has neither, and never will. The real risk (which I’m extremely worried about) is that AI becomes kind of a version of what has been called a “philosophical zombie,” which is something that acts and speaks entirely as though it has consciousness even though it has no genuine inner experience. When this happens with AI, millions of very lonely people will isolate themselves from the world even more, believing that their relationship with AI is a sufficient substitute for human interaction. So the nightmare scenario is a world where the average human has friends, coworkers, and even a spouse, who are all AI, all really nothing inside, not real. I think this probably will happen, and is already in the process of happening. And to me it’s an even greater horror than AI actually becoming conscious.

It's a reassuring myth that greatness is about showing up and being consistent. But at the elite level, it's just the entry fee. Genetics, including durability and trainability, largely determines the ceiling and who separates from the pack. Latest example: 19 year-old cycling phenom Paul Seixas dropping world tour riders and matching Pogacar's record time on a climb (arguably the GOAT).

"Not statistically significant" does not necessarily mean "no differences." The two terms are commonly conflated in exercise science research, which confuses messaging to the general public. We need to do better...

@LoreofRunning1 @PhilipPrins11 @AKoutnik @ProfTimNoakes where do they admit that?

But the authors don't provide any evidence that this is actually ergogenic. They truthfully admit as much. @PhilipPrins11 @AKoutnik @ProfTimNoakes

In my 40 years of bike racing, I've seen a lot of innovations: helmets, clipless pedals, electronic shifting... I'd rate changes in carbohydrate formulations allowing for higher and higher fueling as among the most important. Excellent new review on the case for raising recommendations up to 120 grams/hour.

@MarkBski @BStulberg that's great - have fun!

At 64, I got my 30 year old touring bike (Cannondale T-600) back on the road. I couldn't power my way up the steep desert hills like I used to. Got an e-bike. I've been tooling around town like I did when I was a kid and loving every minute of it. The e-assist helps me up the steep hills.

As @BStulberg says, "one of the greatest competitive advantages is having fun." I've raced bikes for 40 years and take it seriously. But it's also one of the best sources of fun in my life. And yesterday's ride after the LA storms reminds me at 62 you're not too old to play in the mud! Train hard - but have fun!