Now that scientists can manipulate dream content, advertisers want in

The start of a harmful trend

“The best minds of my generation are thinking about how to make people click ads.” - Jeff Hammerbacher, early Facebook employee.

The number of quiet glens in life is decreasing. I was recently at the beach where I watched people in swimsuits search frantically for their phones in the sand. My conclusion was that there are certain modes f being that should remain untouched by the attention economy. Quiet glens should exist in your life where neither price, nor function, is primary. Market creep is a real thing, and you don’t transform into a flag-waving marxist when pointing out that perhaps some things shouldn’t be commoditized. Advertising is now reaching into the heavens, with the launch of the first “space billboard” slated via SpaceX in 2022. And it appears companies would like to reach further, and deeper as well, into that most private of glens: your dreams.

This is because several techniques within the scientific study of dreams have proven it is possible to reliably influence dream content from the outside. Of course, like many things “discovered” scientifically, this has always been anecdotally obvious. Who among us has not woken to a siren within a dream, guised under some occurrence like a fire truck, only to find our real bedside alarm going off?

Normally, the reason not everything makes it into your dreams is that during sleep there is “sensory gating”—which is just a fancy term for saying that incoming sensory signals are much weaker at propagating up and through your cortex. But propagate they can, which in turn influences dream content. Specifically, it’s possible to influence your dream content by prompting the brain precisely at the time it’s entering hypnogogic sleep. While sleep so often feels, in retrospect, binary, it’s actually a succession of stages of mind-wandering as your consciousness collapses, and often the same holds during waking. Dreams are, in a sense, just mind-wandering, but instead of the mostly abstract mind-wandering we do during the day, they are more hallucinatory and imaginative due to having no competing bottom-up input (much like how you start seeing things in a sensory deprivation tank). Personally I think there’s a lot of beauty in this—it’s why I started every chapter of my novel The Revelations with a waking scene.

If you can prompt someone during this mind-wandering hypnogogic period, in that liminal stage between wake and sleep, you can trigger specific dream content. This is called Targeted Dream Incubation (TDI). It’s surprisingly easy to do: during the early stages of sleep just have an audio message repeat to “think of [x]” wherein x is some theme (like “Santa Claus sunbathing”). This leads to early dream content about this theme, although each person will of course dream it in their own. The fact that TDI works can be confirmed by asking subjects to report on what they’re dreaming during a brief waking not too long after the prompt. This opens a possibility that one could play an audio message telling the dreamer to, say, think of having sex with a celebrity (an exercise I will leave to the reader).

TDI, by the way, is incredibly old, and was practiced across many ancient cultures. So the recent interest is really a rediscovery of well-known ancient dream technology that was used for all sorts of purposes. For example, many cultures such as the Ancient Egyptians used TDI in order to have personal meetings with their gods, like Bes (also called “Besa”), a dwarf deity with a lion head who helped women during childbirth, who protected against snakes, and was the god of things like music, art, dance, and general revelry. Here’s some instructions from ~1350 BC describing Egyptian rituals to for TDI:

". . . Make a drawing of Besa on your left hand and enveloping your hand in a strip of black cloth that has been consecrated to Isis (and) lie down to sleep without speaking a word, even in answer to a question. Wind the remainder of the cloth around your neck. . . come in this very night." (Budge, 1901, British Museum Papyrus, No. 122, lines 64 ff and 359 ff, Catalogue of Greek Papyri, vol. I, p.118).

In other words, think of them and they will come. And here’s Bes the dwarf/lion god who you were to meet.

I mean, why wouldn’t you want to dream of Bes? Looks like a pretty fun guy / hermaphroditic deity possessing (according to descriptions) “flabby man-boobs.”

Now these ancient techniques for meeting your gods have given way to the scientific and the capitalistic. Given the easy means to track sleep stages with smart devices, and how Targeted Dream Incubation has had some big headline recent papers, advertisers and corporations have picked up on TDI. As an example, consider the collaboration between Coors beer and Harvard sleep scientist Deirdre Barrett, who has been on a media whirlwind after her project to track pandemic dreams (I am skeptical there is anything very interesting to track, but it sure is clickable). Here’s the description of the collaboration:

“Visit, watch the dream inducing film three times, play the soundscape, go to sleep, and leave those “quarandreams” behind. Yes, it’s that simple.”

In response to these new initiatives, just last week a number of prominent sleep scientists signed a petition against advertisers using TDI by companies having people play soundscapes during their sleep that supposedly lead to Coors dreams. The letter reads:

“[It] is easy to envision a world in which smart speakers […] become instruments of passive, unconscious overnight advertising, with or without our permission,” the letter authors said. “These tailored soundtracks would become background scenery for our sleep, as the unending billboards that litter American highways have become for our waking life.”

One of the letter’s authors was motivated to write the letter specifically because he’d already been contact by three companies (Microsoft included) trying to rope him into dream incubation advertising.

TDI is not the only recent advancement in sleep technology, either. There’s been a lot of progress in both triggering lucid dreams, with the pitch that it is some sort of immersive super-realistic video game, as well as communicating with those who are dreaming. And this communication (and sleep tech in generally, really) has all sorts of potential applications, like how:

…dreams could be curated in accordance with an individual’s objectives, such as to practice a musical or athletic skill…. interactive dreaming could also be used to solve problems and promote creativity—the next moonshot ideas could be produced with an interactive method that can combine the creative advantages of dreaming with the logical advantages of wake. Artists and writers might also gain inspiration from sleep communication.

These advances in dream manipulation combined with the growing panoply of smart devices make for some interesting hypotheticals. For instance, there’s a, ah, new method to trigger lucid dreams on cue, although it works only in men because they experience Nocturnal Penile Tumescence (NPT), which is the dry medical way to say men get erections during their dreams. So of course someone is kickstarting a device that monitors for NPT (in the obvious way) and plays a distinctive triggering sound when you’re dreaming to bring on lucidity.

All to say that dream tech is expanding (or sensitive to expansion gaaaaaaah sorry), and with the growing smartness of our homes and devices maybe it’s not so crazy to think that companies might be paying attention. On first hearing about the various experimental projects for placing advertising in dreams, I had a reaction similar to the letter signers: I was appalled. It’s worth noting that Deirdre Barrett has given a defense of her work with the Coors ad, pointing out that:

In the US, Section 5 of the Federal Trade Commission Act prohibits “unfair or deceptive acts or practices in or affecting commerce” including advertising that a reasonable consumer wouldn’t recognize to be an advertisement… This FTC prohibition of deceptive advertising applies across all media and automatically includes new ones as they emerge. At the very least, any attempt to play ads designed for sleepers would have to feature anti-deception statements for every advertisement that was going to play that night.

In other words, Barrett argues people would have to be alerted to the ads, so the dystopian scenario outlined in the letter of smart devices detecting your dreaming and piping in ads without warning isn’t really a concern. At least unless something changes. She also speculates that dream ads might not be very effective compared to awake ads (but her points have a lot of assumptions that brands are after an immediate purchase, not, say, subliminal positive attitudes or implicit brand recognition.)

So her dismissal that there is any problem at all seems to misunderstand the signatories’s concerns. They aren’t solely worried that ad companies could do this without our consent, they are more generally worried about the growth in dream tech, scientist’s ability to manipulate dream content, and the possibility of advertisers using it in some scenario in the future, given how easy it is (just audio messages repeating at the right time and tracking of sleep stages, which plenty of smart devices can do).

And its naive to think advertising companies don’t want in on people’s dreams, that’s totally natural for them—they already have in to all the content you imbibe (outside of subscription services like substack itself). Sure, as Barrett argues, dream advertising might not be particularly effective, but there’s no strong evidence for traditional advertising being effective anyways, and yet all big companies do it. The intrusion of commercials into dreams is truly a certain kind of modern horror, a final penetration of our most private moments. And instead of Bes, we get Coors. Every culture meets their gods in their dreams, I guess.

Beyond the moral question, there’s also the health question to consider before any of this takes off. Peter Watts, the Hugo-award winning sci-fi writer, pointed out on his blog a few months ago the health risks of new dream technologies. He used my work on the Overfitted Brain Hypothesis (OBH) to make his point. The OBH claims that the evolved purpose of dreaming is help prevent you from overfitting: since animals have such a limited daily repertoire of experiences, they are constantly in danger of statistical overfitting. This means they substitute a too simple model of the world for a complex one—essentially animals are learning too well, and thus the need to dream. The OBH is supported by the surprising similarity between the phenomenology of dreams and techniques in deep learning, like domain randomization and dropout, that are commonly used to prevent artificial neural networks from overfitting during training. The idea is that dreams are essentially a combination of this class of techniques used nightly for the same purpose.1

What Watts argued was that this means that the content of dreams should be, well, dream-like. If you spend too much time lucid dreaming, you might actually prevent the ameliorative effects of dreams. He says:

So: two papers, one theoretical [the OBH], one applied [communicating with lucid dreamers]. One on what dreams are for, the other on how they can be repurposed. If the first one’s right, the second points us down a dangerous road—one which might ultimately short-circuit an essential cognitive adaptation older than our species. You could strap on your DreamweaverTM headset with every intention of learning Mandarin or breaking some cognitive logjam—and after a few nights, find yourself not much good at anything but playing Tetris.

Perhaps science gives us an ought from an is here (or at least, helps us along the way) in banning dream commercials for health reasons. Although it’s worth noting how easy for sleep scientists to overstate their case; as Alexey Guzey has pointed out, sleep’s effects on health shouldn’t be exaggerated. I think dreaming is an important evolved function, but I’m skeptical that effects of sleep and dreaming on learning or cognitive function (in the sense of improving them) are immediate. Rather, I think it’s more likely the strong diurnal sleep drive is homeostatic, restoring you to some baseline in the morning. Getting enough sleep such that you don’t get dementia, or avoid overfitting of your performance on tasks, are likely life-long cumulative effects, with the occasional immediate expression of the positive effect being much rarer.2

As dream technology increases in its potential perhaps there will come a time when actually things like TDI could help, rather than hurt. Instead of being used for advertising or trying to learn some new skill, I want to open up the positive idea that, even if a hypothesis like the OBH is true, perhaps TSI could actually improve its effects. After all, why did the Ancient Egyptians use TDI to meet with their gods? Because the conversation gives you something back you couldn’t have gotten otherwise. Maybe we can use TSI to have more cathartic dreams, talk to a dead loved one, or just boost the degree to which dreams combat overfitting by prompting especially Dali-esque dreams. Maybe even we’ll get to meet Bes.


I’ve also pointed out that identifying evolved purpose of dreams seems similar to solving the puzzling problem of why humans are so concerned with fictions (TV watching, novel reading, story-following). Perhaps the OBH explains both, as what is a fiction but an artificial dream, and a dream but a biological fiction?


This makes it extremely difficult to test these sorts of hypotheses. Let’s say the one goes to test the OBH and finds that generalization abilities aren’t statistically significantly correlated with the overnight amount of dream activity (or something like that). Does this contradict the hypothesis? Well, if the effect is very subtle but cumulative every day, then no. This sort of thing is why I am very wary of conclusions from “mom and pop” neuroscience labs of small sample sizes and short-term experiments.