Two Thermostats: Why Your Addiction Doesn't Work the Way You Think It Does

If you've ever struggled with a compulsive behavior like food, alcohol, nicotine, work, gambling, your phone, you've probably been offered one of two explanations:

1. You have a disease. You're wired wrong. Accept it.
2. You lack discipline. Try harder.

Neither of these is completely wrong, but neither gives you anything to work with. They're descriptions of outcomes, not mechanisms. If you're the kind of person who needs to understand how something works before you can engage, especially if "just trust the process" makes you want to leave the room, you've probably noticed that nobody has ever explained what's actually happening inside the system that's running you.

Here's what's happening. There are two mechanisms that can dominate compulsive behavior, and they feel different, respond to different interventions, and (here's the kicker) explain why the thing you've been trying to do about your problem may be perfectly rational and completely wrong at the same time.

Thermostat One: The Receptor Problem

Some substances work by occupying specific receptors in your brain. Nicotine binds to acetylcholine receptors. Benzodiazepines bind to GABA receptors. Opioids bind to mu-opioid receptors. Caffeine blocks adenosine receptors.

When you use these substances regularly, your brain adapts. It adjusts the number and sensitivity of those receptors to account for the external supply. This is tolerance. This is your brain remodeling itself around the assumption that the substance will keep showing up.

When you stop, the receptors are exposed. They're either under-stimulated or over-stimulated depending on the substance, and the result is withdrawal: anxiety, irritability, physical discomfort, craving. The system is asking one simple question: are the receptors occupied or not?

Here's the important part: this thermostat is roughly proportional, at least in the cleaner cases. If you skip your morning dose, the craving builds. But one normal dose at noon more or less resets the system. The room got cold, you turned the heat on, the room warmed back up. No drama. No overshoot. The thermostat is mechanical, and the correction is mechanical.

This doesn't mean quitting is easy. Receptor adaptation can be brutal, and the timeline for your brain to remodel back to baseline can be weeks or months. But the logic of the system is simple. You're missing a molecule. The system wants the molecule. Replacing the molecule resolves the signal in rough proportion to the deficit.

Tapering works here. Substitution works here (nicotine patches, methadone, etc.). The intervention matches the mechanism: you manage the receptor transition.

Imagine your ancestor stumbling onto a plant that dulls pain after a hunt. He chews it for a few days, runs out, feels lousy for a while, then his body recalibrates and he moves on. The receptor system evolved for a world where the supply was inconsistent and finite. The plant runs out. The withdrawal resolves. The thermostat resets because the environment enforces the reset. The modern problem isn't the mechanism, it's that the plant never runs out anymore.

Thermostat Two: The Prediction Problem

Now consider a different kind of compulsive behavior. You decide to eat clean. You skip breakfast. You're disciplined at lunch. Maybe you skip that too, or eat something small and virtuous. By 7 PM, you eat more than you would have eaten in three normal meals. Now, your logical brain knows your body doesn't need that much food. But something else is happening in that very moment.

This pattern shows up with food, alcohol (for many people), gambling, compulsive work, shopping, sex, doom-scrolling. Basically, any behavior where the reward is driven less by a molecule binding to a receptor and more by a prediction your brain is making about how good the reward will be.

Here's the mechanism: Your brain has a sophisticated system for anticipating rewards. It's dopaminergic (driven by dopamine) and its job is not to make you feel pleasure. Its job is to make you want. To motivate pursuit. To predict how rewarding something will be and then generate the motivational energy to go get it.

This system, what neuroscientists call the incentive salience network, can do something the receptor system does not. Under the right conditions, it compounds.

Those conditions are specific: deprivation, restriction, repeated cue exposure, stress, uncertainty. Defer a reward inside that state and the system doesn't just log a deficit. It revises the prediction upward. It reads the delay as evidence that the eventual payoff must be even more important, because why else would you be tolerating the deprivation? The anticipatory signal stops rising linearly. It escalates. It recruits stress hormones, hunger signals, emotional urgency. By the time you finally engage with the reward, the system has pre-written a check that no actual experience can cash.

Defer outside that state and the opposite can happen. The craving fades through extinction, habituation, a shift of context. The urge surfed until it passes. Whether a delay multiplies the craving or dissolves it depends on the surrounding state, and an all-day restriction stacked with mounting food cues is exactly the state that turns deferral into inflation.

This is why the binge overshoots. It's not a failure of willpower. It's the foreseeable result of a prediction engine that has been revising its estimates all day. You're not eating because you're hungry. You're eating because the wanting signal has inflated so far beyond what any amount of liking can satisfy that the system just keeps driving consumption, searching for a satiation point that has already moved past where the food can reach it.

This thermostat doesn't just measure the temperature in the room. It has a feedback loop on itself. The longer the room stays cold, the more the thermostat recalibrates what "warm" means, so when the heat finally kicks on, it blasts past the original set point.

That's one day. Run the cycle enough times and something more durable sets in. Over many rounds of restriction, overshoot, and the stress of both, the system stops returning to its old baseline and begins regulating around the new pattern. This is what addiction researchers call allostasis: the baseline itself moves. The single binge is the prediction engine overshooting inside a day. Allostatic load is what accumulates when the days stack up and the set point stops coming home.

Now imagine a different ancestor. He finds a grove of ripe fruit. Rare, calorie-dense, critical for survival. His brain doesn't just say "eat." It says "eat everything, right now, because this may not be here tomorrow." The prediction engine evolved to treat scarcity as the default assumption. Every hour without food was evidence that the next meal mattered more, not less. Bingeing wasn't a failure of regulation. It was the regulation: a system perfectly tuned to an environment where overshooting was the safe bet because undershooting meant death. You're running famine software in a world with a 24-hour DoorDash. The thermostat isn't malfunctioning. It's performing exactly as designed for a world that no longer exists.

Why This Distinction Matters to You

If you're dealing with a Thermostat One problem and you apply a Thermostat Two solution (say, trying to use cognitive reframing to manage nicotine withdrawal), you're going to have a bad time. The receptors don't care about your insights. They want their molecule.

If you're dealing with a Thermostat Two problem and you apply a Thermostat One solution (say, trying to taper your gambling by doing a little less each week), it can make it worse for some people, because you're feeding the prediction engine a series of small rewards that keep it primed without ever satisfying it.

If your mind is already organized around anticipation and optimization, if you live three moves ahead for a living, Thermostat Two may feel especially familiar from the inside. The prediction engine that makes you exceptional at your work is the same machinery that, when it locks onto a compulsive behavior, generates anticipatory signals no consummatory experience can match.

Your greatest cognitive asset and your most destructive compulsive pattern are running on the same hardware.

You're not broken. This is not a moral failure. It's not a disease in the way most people mean when they say that word. It's a mismatch between a system that evolved to motivate pursuit under scarcity and an environment where the thing you're pursuing is always available and the scarcity is self-imposed.

I'll say that again, because it's the single most important idea in this essay: The system that drives your worst compulsive behavior is not a broken version of you. It is the same system that drives your best thinking. Prediction, pattern detection, anticipatory modeling. These are why you're good at your job, why you see around corners, why people pay you to think.

But that system doesn't have a "work" mode and an "addiction" mode. It has one mode. And when it locks onto the wrong target with the wrong gain in the wrong environment, it does exactly what it does at work: it escalates commitment, narrows focus, and outpaces the available evidence. You are not fighting a disease. You are not failing a character test.

You are experiencing your own cognitive engine doing precisely what it was built to do, aimed at something it was never meant to track.

The Crossover Problem

Nothing is perfectly clean. Substances can recruit both thermostats.

Nicotine leans Thermostat One at first: receptor occupancy dominates early. But the cue learning starts almost immediately, and over years the rituals and associations (coffee and a cigarette, stress and a cigarette, finishing a meal and a cigarette) wire the behavior deep into the Thermostat Two prediction network. Now quitting isn't just a receptor transition. It's a dismantling of a prediction architecture that has woven nicotine into a hundred different internal states. This is why quitting after twenty years is qualitatively different from quitting after six months. The mechanism has shifted.

Alcohol is the most common hybrid, and the messiest. Even early use is never just one receptor: alcohol acts on GABA, glutamate, dopamine, the endogenous opioids, and the stress and sleep systems at once, which is part of why alcohol withdrawal can be medically dangerous rather than merely uncomfortable. On top of that pharmacology, chronic use layers in heavy incentive salience distortion. The person who can't stop after one drink is often well past a receptor problem. They're deep in Thermostat Two: one drink doesn't satisfy because the prediction engine has already priced in ten.

The practical implication: the longer and more ritualized your relationship with any substance or behavior, the more likely it is that Thermostat Two has been recruited on top of whatever Thermostat One dynamics were there originally. Duration and ritual convert simple dependencies into prediction-engine problems.

What You Can Actually Do With This

Knowing which thermostat is running your problem doesn't make the problem disappear. But it does something that might be more important: it makes your experience legible to you.

If you're in a Thermostat Two pattern, you now understand why "just one" is a lie, and it has nothing to do with that shameful feeling of weakness you load upon yourself every day. "Just one" is a lie because the prediction engine interprets small rewards as confirmation that the larger reward is coming, which increases the anticipatory signal rather than satisfying it.

You understand when deferral strategies backfire. Discipline works, and I'm not arguing otherwise. But under deprivation and mounting cues, discipline can feed the wrong thermostat. Every hour you white-knuckle through that state is another hour the prediction engine spends revising its estimate upward.

You understand why the binge feels so disproportionate: the system promised itself something that the actual experience structurally cannot deliver. The wanting outran the liking. That gap is the source of the overshoot, and it was baked in before you took the first bite.

And if you're someone whose prediction engine is already running at high gain, who can see twelve moves ahead in your work but can't stop checking your phone, you now have a framework for understanding why. The same system that makes you good at what you do is the system that's vulnerable to this specific failure mode. Not all systems. This one. And this one has a logic you can see.

Seeing it doesn't fix it. But for people who need to understand the mechanism before they can engage with the repair (and I'm looking right at you) it's the necessary first step.

The thermostat isn’t broken. It’s working exactly as designed, with the set point drifting too high, in the wrong environment, with no one at the controls who understands what it’s actually doing.

Now you do.

References

To read the science behind the difference between wanting and liking (anticipatory vs. consummatory reward signals) and the allostatic model of addiction see:

Berridge, Kent C., and Terry E. Robinson. 2016. “Liking, Wanting and the Incentive-Sensitization Theory of Addiction.” The American Psychologist 71 (8): 670–79.

Koob, G. F., and M. Le Moal. 2001. “Drug Addiction, Dysregulation of Reward, and Allostasis.” Neuropsychopharmacology: Official Publication of the American College of Neuropsychopharmacology 24 (2): 97–129.

Linnet, Jakob. 2014. “Neurobiological Underpinnings of Reward Anticipation and Outcome Evaluation in Gambling Disorder.” Frontiers in Behavioral Neuroscience 8 (March).

Morales, Ileana, and Kent C. Berridge. 2020. “‘Liking’ and ‘Wanting’ in Eating and Food Reward: Brain Mechanisms and Clinical Implications.” Physiology & Behavior 227 (December): 113152.

Robinson, Terry E., and Kent C. Berridge. 1993. “The Neural Basis of Drug Craving: An Incentive-Sensitization Theory of Addiction.” Brain Research Reviews 18 (3): 247–91.

Robinson, Terry E., and Kent C. Berridge. 2025. “The Incentive-Sensitization Theory of Addiction 30 Years On.” Annual Review of Psychology 76 (1): 29–58.

Wittenberg, Ruthie E., Shannon L. Wolfman, Mariella De Biasi, and John A. Dani. 2020. “Nicotinic Acetylcholine Receptors and Nicotine Addiction: A Brief Introduction.” Neuropharmacology 177 (October): 108256.

Essays on treatment resistance, altered states, and the conditions under which change becomes possible.

Subscribe