Published on May 20, 2025 11:43 PM GMT

[Epistemic status: I'm not a neuroscientist nor any kind of medical nor biological professional. I have an amateur understanding of neurochemistry and cognitive science. This post is primarily based on what has worked for me and my intuitions around why. I err on the side of epistemic transparency and caution where possible.]

Prerequisite Concept: Hedonic Set Point

Disclaimers/Warnings

I think this can be a genuinely very useful technique for people. However,

This can Probably be Dangerous

This is a post about doing some light amount of hacking your brain's chemicals, specifically in a way that involves temporarily taking away enjoyable things. Do not do this if you think that would be long-term bad for your mental health. This is not a technique for dealing with e.g. severe depression (nor probably most smaller-magnitude depressions). This is not a technique for fixing complex mental health issues, it is a technique for dealing with a very specific thing in some subset of people whose brains are somewhat like mine along some relevant axes.

I think there are some mental states and neurotypes where this is an actively harmful technique if you try to push through it for >0.5 days when that isn't the right move. I suspect it's relatively safe if you have the mindfulness to realize if things are making you long-term worse off and immediately Nope Out of the detox (described near the end). This is worth being cautious about.

And I'm Noticeably Biased

Also, this post is about a mental technique that

made me feel good after doing itrecently temporarily fixed what happened to be the most [big × tractable] bottleneck to my productivity.

For these reasons, the picture I paint of the benefits may be rosy and portray an effect size that does not generalize reliably to other people. Given those forces acting upon me, I try to paint a true picture and convey a general technique to whatever extent I reasonably can.

Introduction

(You can probably skip this section if you already know what a dopamine detox is and why you might want to do one.)

A "dopamine detox" is, for the purposes of this post, a period (usually a day) in which you try really hard to get rid of superstimuli in your environment. I've written in some probably-helpful additional detail about how superstimuli can hack your cognition, but this post does most of the necessary conceptual explaining.

I'd appreciate if anyone who tries this technique because of this post reviews it, since my anecdata is limited to my own experience and that of a small number of others.

Superstimuli and Detoxing

Stimulating activities give people a little bit of dopamine and make them more likely to do that activity in the future. This is mostly fine within the approximate ancestral distribution of stimuli: you eat a tasty, healthy salad, you like the salad^[1], you continue eating salads when you're hungry for them, and this gets you utility. Importantly, the amount of stimulus you get from eating a salad causes you to eat more salads approximately the optimal amount, from a utility-maximizing point of view. You don't have to do conscious optimization over how often you eat salads, the reward-seeking part of your brain does that automatically. A superstimulus, for the purposes of this post, is something where this property is not true.

Pizza is a superstimulus^[1]. Pizza is entirely outside even the far reaches of the ancestral tasty food distribution available to humans. It is not, however, correspondingly healthy. By default, many human brains exposed to pizza get hacked by the density of the stimulus, and pursue pizza more than is utility-maximizing (even after accounting for the fact that experiencing tasty food is also utility).

Additionally, in the context of abundant pizza, healthy salads seem much less appealing. Pizza moves the hedonic set point upwards, and suddenly salads don't seem so appetizing in comparison. Beyond just receiving undue priority among foods, pizza competes against other types of stimuli as well: a higher hedonic set point makes "write a long LessWrong post about meditation" seem less fun than it would be otherwise (i.e. usually pretty fun).

A dopamine detox is about resetting the hedonic set point to "normal" so that subjective reward corresponds more closely to utility.^[2]

The next section follows an archetypical progression of a dopamine detox. I have had several experiences separated by months which closely match this progression, but this description is not a literally true story of any one particular dopamine detox.

The Detox Experience

The Day Before

The day before the detox is especially filled with the stimulus-noise that fights for my attention every day. My friends send me memes. I scroll Substack. I eat pizza. I get a couple hours of work done, but I have a couple times where I sit down at my computer, find a video of someone doing something entertaining, and then the next thing I know it's mysteriously two hours later.

I say oops. I notice that something happened that I wish hadn't. I open up my journal and start dumping out my thoughts, but my mind feels sluggish. It's not immediately clear what's going on, just that something weird is going on in my brain. My mind wanders back to that video I was watching earlier. I almost open up Youtube again, but I stop myself. Oops again.

I've been through this enough times to notice what's happening: I've been afflicted by the superstimuli. I delete Youtube from my phone and adjust my browser's site blocker. It's getting late though, and I should sleep instead of trying to optimize any harder. Tomorrow, though, I'll do a dopamine detox.

The Day Of

I wake up to my phone's alarm. I consciously choose to not check my notifications, and instead only look at my phone long enough to turn on Do Not Disturb. I have filters designed to let through emergency communications, but I precommit to check my messages at midday for any urgent-priority interrrups anyways. And so I go about my day.

It hurts, not like injury or sadness, but like exercise and difficult work. It hurts like Actually Trying hurts, sometimes. There is a part of my brain that is nagging in the background for superstimuli. I sit down to eat, and it nags for me to read something on my phone. I sit down to work, and it nags me to check Discord. The theme of the day, however, is that I expend willpower to do Not That.

It's taxing, both on my willpower to avoid superstimuli and on my mindfulness to notice when that's happening. But left with nothing more stimulating to do, I write: I journal, I do work for my job (which also happens to be writing), and I write a LessWrong post. There are several moments when I don't have the energy to do anything detox-allowed but sit and exist, but I choose those instead of the superstimuli.

I go to bed. My brain is still noisily nagging me about things as I try to sleep, but I am Tired and the Tiredness wins.

The Days After

I feel noticeably less bad than yesterday! Sleeping is good for becoming less tired, as well as for adjusting the hedonic set point. I go outside. Sunshine and birds and trees are unusually pleasant. I talk with a friend. Talking with people is really pleasant!

I very carefully reintroduce some limited superstimuli, since I no longer want to spend willpower on them. I read a fun article, I check Discord, I eat some candy when it's offered. They are unusually pleasant, but that pleasure is not the main point of a dopamine detox.

The real benefits are in the normal-level stimuli. Working on something more difficult than eating candy is suddenly very doable. My workflow becomes closer to being the default, instead of something that takes non-negligible willpower. I'm more readily able to form habits and enter deep work.

I find that I get less lost when having complex conversations, and I organize my thoughts better when explaining myself. I pick up on subtle things in my environment that I didn't notice before in all of the stimulating noise.

Importantly, dopamine detoxing puts me in a headspace where preserving these benefits into the future is less taxing. Sometimes I fail and need to detox again, but the default action in most cases is to keep being not-hacked by stimuli and keep being unusually productive. The "wow birds and trees and sunshine are way cooler than I remember" effect is mostly a temporary artifact of the rapid change, but some of that is also preserved.

Further Notes

Duration

As I mention above, sleep is a very good way to reset. Willpower is replenished, hedonic set points move, memories and new cognitive patterns consolidate, etc.

I keep a biphasic sleep schedule, but my mid-day nap isn't long enough to get these benefits, so the most convenient duration for a dopamine detox is 24 hours. The mid-day nap helps with all of those things a little bit, but I find most of a day necessary to get the ~full benefits of such a dopamine detox.

People definitely do things like this for longer. They go on meditation retreats, they go camping in a remote part of the world, they simply decide to expend willpower for longer. I haven't tried this, since I predict the willpower expenditure would not be worth the marginal productivity gains, beyond the single-day experience.

Intensity

I pick a slightly arbitrary point for the intensity of my dopamine detoxes. Intensity falls on an axis from "restrict yourself to only 15 hours of Youtube and 30 cookies maximum per day" to "run off into the woods for a week and forget about anything invented since agriculture." I choose a point somewhere around:

no endless scrolling on Substack or Youtube (nor Reddit/Instagram/Tiktok/Facebook/etc. if I ever used them). Productive news-reading is okay.very limited or emergency-only messagingmy normal moderately-healthy food choices, as well as not going out of my way to get candyno miscellaneous stimulus-seeking behaviors, e.g. repeatedly checking my LessWrong karmaInternet access is totally fine, being active on Slack is totally fine.

This is the point where most marginal reward-removal would remove stimuli that are actually really useful in making me pursue my goals. I don't get rid of reading the news since my literal job is to be up to date on certain news, and even though I get a lot of (sometimes super-)stimulus from reading, I don't cut it out for this reason.

It also happens to be a goal that is entirely tractable, given the amount of willpower I have available to expend, although it's non-negligible. You may be different along both of these axes and thus have different forces acting on you, and your threshold should be adjusted accordingly.

While the extremes of the given "detox intensity" axis are intentionally too extreme and should not be done, there are very practical reasons to go closer to those extremes for shorter durations:

Noping Out, wherein you temporarily go towards superstimuliStaring at a Wall, wherein you metaphorically run off into the woods for a few minutes

Noping Out

If you're instructing someone on how to do something potentially dangerous, you both want to give disclaimers and give a way to safely exit whatever they're doing. I gave the disclaimers, here is the other half.

This is for when, instead of feeling like exercise or difficult work, the detox feels like mental health issues. If you notice yourself having an anxiety attack because of a detox, Nope Out. If you notice yourself being unusually depressed because of the detox, Nope Out. Hopefully these examples illustrate a very clear and general algorithm:

if bad:    NOPE_OUT()

This means go eat some candy or message a friend or whatever else the nagging-for-superstimulus part of your brain is asking for. Temporarily taking away nice things had a net bad effect, so please give your brain back its nice things.

There is no secret graceful maneuver required, like when trying to safely Nope Out of a backflip. You just stop expending willpower on things that are bad. This does not instantly fix things, but it responds to the signal of "this technique is making things worse" by no longer making things worse. The disclaimers are an attempt to prevent some of this, but they cannot prevent all harms, especially those which are due to unexpected events.

This general algorithm is not literally perfect, but I expect most of the exceptions to exist in thought experiments rather than real life.

Staring at a Wall

If you find that you have an unusual excess of willpower available for this technique (after accounting for slack), you can speed up the process of moving your hedonic set point downwards. This is done by further depriving yourself of positive stimulus for a much shorter period of time. Sometimes this makes the technique work better, but mostly it just makes things faster. Here is the technique I use, for when this Difficult Expenditure of Willpower is not difficult enough and does not expend enough willpower:

not

The timer is critically important because it means you no longer have to consciously track whether or not you're done with the exercise.

just the wall

This is usually unpleasant, and the unpleasantness is intentionally much more dense than my baseline dopamine detoxing, since intensity is what it takes to move the hedonic set point faster. I don't know what happens if you do this for more than 10-15 minutes because I haven't tried it for longer, but I could see it worst-case causing some of the weird meditation injuries if done in excess.

Most of the technique in this post aims to calibrate the reward signal to your actual utility by only doing rewarding things in rough proportion to their utility. This sub-technique, however, is not doing that. Many things are significantly more useful to me than staring at a wall. This sub-technique is just about moving the hedonic set point as fast as possible downwards, in order to reach the point of calibration faster. This sub-technique is overkill, but useful overkill in the sense that having robust interstellar transport probably means it's easier to get to the moon.

Other Reasons Not to Detox

If your reward environment is already well-calibrated and gets you utility and protects you from superstimuli, you probably don't need a dopamine detox.

If any of the previous disclaimers, warnings, or literally anything else in this post made you think this would be a bad idea, don't do it.

If you would misinterpret this post and miss out on utility that happens to require you to e.g. eat candy and gain rapport with someone important or watch a clickbaity Youtube video that happens to contain a full solution to the alignment problem, then don't do it.

^{^}
If you have atypical preferences about salads or pizza, feel free to mentally swap this out with something more appropriate. I am writing this example for what I believe to be the typical mind reading it, but I hope this post generalizes further.
^{^}
If you're paying close attention, you'll note I use a slightly different definition of superstimulus here than Eliezer in the original post. Eliezer was specifically focusing on stimuli that are stronger than the ancestral distribution, whereas I'm focusing on reward signals that don't automatically make you maximize utility. These are different! If I give you pizza for every 10 utilons you get, this is an Eliezer!superstimulus but not a this_post!superstimulus. For many cases, these definitions do overlap, which is why I am using the same word.

Discuss