Scientists tricked a rat's brain into thinking salt water tasted good
Researchers used optogenetics to amplify reward-signal neurons in rat brains so that when the rat drank salt water it hated, the brain still judged 'this was good'—sensory input and reward signals can be directly severed.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Rationality is economics' limiting case, not how humans actually decide
Camerer describes the utility-maximization model as a ‘limiting case’ in economic theory, like a simplifying assumption in physics—expert tasters with extensive experience who can rank all options might approach perfect rationality, but that is not the state of most people most of the time. The model resembles the endpoint of learning rather than its starting point. Textbook ‘rational actors’ were never what the field actually believed; they were just a convenient launching point.
— Colin CamererLife and death probability questions should add to 100%—they don't
When asked ‘what is the probability you'll live to 75,’ people answer with a number like 90%. But when asked separately ‘what is the probability you'll die before 75,’ the answer will not be 10%—the two should sum to 100%, but they do not. Camerer's collaborator Eric Johnson attributes this to ‘default endorsement bias’: the question frames attention toward survival or death accordingly. Frame effects are not trivial—salad dressing labeled ‘10% fat’ sells differently from ‘90% fat-free’. Advertisers and political consultants have long understood this.
— Colin CamererChess masters guess numbers like everyone else in this game
In the guessing game where players choose 0–100 and the closest guess to two-thirds of the average wins, most people reason only one or two steps deep—‘others will guess 50, so I'll guess 33; or others will guess 33, so I'll guess 22’—and almost no one reasons all the way to zero. The New York Times ran this with 50,000 players; the data shows 33 appears far more often than 22, and deeper reasoning is nearly nonexistent. Surprisingly, professional chess players perform about the same as amateurs in this game—deep strategic thinking appears to be highly domain-specific and does not transfer across domains.
— Colin CamererAudiences can't reverse-engineer ‘no reviews’ as a bad signal
From 600 major theatrical releases (2000–2009), roughly 10% are ‘cold opens’—studios don't send films to critics beforehand. These films later show the lowest aggregate critic scores, indicating studios knew they had lemons and held them back. But the cognitive hierarchy model predicts that some viewers will read ‘no reviews’ as neutral rather than bad, like Sherlock Holmes' dog that did not bark—most people won't reverse-engineer ‘why isn't anyone reviewing this?’. Because of this reasoning gap, cold-open films outperform their true quality by roughly 15% at the box office.
— Colin CamererNeuroeconomics began with a silent room of economists
In 1997 Camerer and George Loewenstein organized a conference in Pittsburgh bringing together economists and neuroscientists to explore whether economic questions could be studied with neuroscience methods. Most economists stayed silent throughout—in the field, admitting ignorance carries low status. Notably, Peter Diamond, Al Roth, and Paul Romer were all in that room and later won the Nobel Prize, yet except for Romer's one paper on emotion, virtually none pursued this direction. Neuroeconomics actually advanced only when Caltech acquired an unused brain scanner and a few professors were willing to bet their tenure on it.
— Colin CamererPeople who think one step deeper activate stronger mentalizing circuits
In the guessing game, players who reason forward—choosing 22 instead of 33—show stronger activity in brain regions responsible for ‘mentalizing,’ understanding others' thoughts: the right temporoparietal junction and dorsomedial prefrontal cortex. The test works by showing subjects the same photograph twice, once asking ‘what is this person doing’ and once asking ‘why is this person doing that’; the activation difference in the second version indexes mentalizing. Camerer acknowledges research here is sparse, but existing studies point the same direction: strategic depth and mentalizing circuitry use are linked.
— Colin CamererGym check-in intervals form a nearly perfect power law
Analyzing gym attendance data from roughly 20,000 people over 1,500 days, Camerer plotted visit intervals—many short gaps, many long gaps, no bell curve—and on a log-log graph it traces almost a perfect line (R²=0.97), a ‘burstiness’ that should be extremely rare in social science data. He later discovered mathematicians proved in 1954 that non-FIFO queuing, where ‘priority queue’ admits some people ahead of order, produces power-law distributions; such results appear mostly in physics journals. Their ‘Neural Autopilot’ model—habit-driven decision-making tracking only two values, the last choice and reward reliability—reproduces this effect with a slope of roughly -2.
— Colin CamererLight can make rat brains misread foul tastes as rewards
The brain contains neurons encoding ‘reward prediction error’, discovered in 1997 by Schultz, Dayan, and Montague in rodents. Using optogenetics, Camerer's team amplified these neurons when a rat pressed a lever and drank salt water it hated, creating artificial ‘this was good’ signals in the brain—the rat kept pressing despite its facial expressions and behavior showing it disliked the taste. Camerer calls this ‘neural gaslighting’: the sensory system knows the truth, but the hijacked reward signal drowns it out.
— Colin CamererIn their own words · checked verbatim
I think of it as a limiting case, like you might think about in physical science or natural, in the sense that there may be people with extraordinary experience, like a wine expert who has tasted a lot of wine and is immune to herd behavior and who can...
Colin Camerer3:45
Yeah, I was at the University of Chicago 100th anniversary of their PhD program, of which I'm a graduate, and somebody asked that. And I think I said 75% rational.
Colin Camerer12:14
If you call a salad dressing 10% fat, you sell a lot fewer than if it's 90% fat-free.
Colin Camerer13:41
You might think that chess players can do more strategic thinking, but they look very much like regular folks.
Colin Camerer17:11
the fact that there's no review is bad news. In this case, no news is bad news. But psychologically it's a little tricky. It's kind of like Sherlock Holmes, the dog that didn't bark.
Colin Camerer29:55
Now, did we have a causal effect? [laughter] I'm just saying it's a small sample. You know, we don't have much statistical power.
Colin Camerer40:19
more strategic thinking, as evidenced by these kind of extra levels, like 22s or 18 or 19 in the 11-20 game, those extra steps of thinking are recruiting this mentalizing.
Colin Camerer52:11
You upregulate, so when it presses the lever and gets salty water, the brain neurons say, "That was great," because you've hijacked, you've convinced the brain that salty water is really good because you know where these reward prediction error neurons are.
Colin Camerer1:22:35
Figures
| Cold-open films (not screened for critics) | ~10% | 29:55 |
| Box-office boost from cold opens vs. true quality | ~15% | 31:18 |
| Film study sample size | 600 major releases (2000–2009) | 29:55 |
| Gym check-in interval power-law fit | R²=0.97 | 1:01:29 |
| Caltech students' math SAT median | 800 | 23:07 |
| Level-k reasoning gap between Caltech and nearby community college students | ~0.5 levels | 23:07 |
| Food-value boost from whispering food names during sleep | ~10% | 1:22:35 |
Glossary
- mentalizing
- Inferring others' mental states; the brain process underlying theory of mind.
- level-k reasoning (cognitive hierarchy model)
- Stepwise strategic thinking where players reason only a finite number of moves deep.
- framing effect
- Identical information triggers different judgments when presented differently.
- burstiness (Poisson non-homogeneity)
- Events cluster unevenly with intervals following a power-law distribution, not random or uniform.
- Neural Autopilot model
- Habit-based decision-making storing only two values—the last choice and reward reliability.
- neural gaslighting
- Artificially amplifying reward neurons to override sensory judgment of actual experience.
How to listen
Entrepreneurs and product managers interested in behavioral economics and decision psychology, plus researchers wanting to understand neuroeconomic empirical methods.
If you're not interested in queueing theory mathematics, skip 1:04–1:16 on power-law mathematical derivation.