Here is an experience almost everyone has had and almost nobody has a decent explanation for.
You reach for the thing. You want it — really want it, the pull is physical. You get it. And it's... fine. Mildly pleasant. Nothing close to the size of the wanting.
Then, twenty minutes later, you want it again.
The intuitive reading is that you're deceiving yourself, or weak, or that the wanting is somehow fake. The actual explanation is that wanting and liking are produced by different systems, and one of them can grow while the other doesn't.
The finding
Terry Robinson and Kent Berridge published the incentive-sensitization theory in 1993. They revisited it in 2025, thirty years on, and the central claim has held up better than most things in the field.
The claim: mesolimbic dopamine mediates incentive motivation — "wanting," the pull toward a reward — but not the hedonic impact of that reward when you actually get it. "Liking" runs on different circuitry.
The evidence was awkward for everyone at the time, including its authors. Amplify dopamine signalling in animals and they will work harder, pursue more intensely, want more. Measure whether they enjoy the reward more when they get it — using hedonic reactions rather than behaviour — and the answer is no. The wanting scales. The liking doesn't.
Berridge has said this surprised him. He'd assumed, like nearly everyone, that dopamine was the pleasure molecule. It isn't. It's closer to the pursuit molecule.
Sensitization: the part that stings
The second half of the theory is where it starts explaining your life.
Repeated exposure can sensitize the wanting system — make it more reactive, not less — in a way that persists long after the exposure stops. And sensitization is selective. It amplifies wanting without amplifying liking.
So the gap doesn't stay constant. It widens. The pull gets stronger while the payoff gets weaker, and both changes have the same cause.
This is why Robinson and Berridge's theory explains something the withdrawal-based accounts struggle with: people relapse long after withdrawal has fully resolved, and long after the thing has stopped being enjoyable. If pursuit were driven by anticipated pleasure, that wouldn't happen. If pursuit runs on a separately sensitized system, it's exactly what you'd expect.
What this changes about self-blame
Most people run an implicit model where wanting something is evidence that it's good, and continuing to want something that isn't good is evidence of a defect in them.
Under the dissociation, that inference just doesn't hold. Wanting is not a report on quality. It's the output of a system that learned to flag this cue as significant, and that system doesn't consult the one that evaluates whether the experience was any good.
You are not being irrational when you want something you no longer enjoy. You're reading one system's output and assuming it came from the other.
That reframe matters practically, not just emotionally. "Stop wanting it" is an instruction to a system that doesn't take instructions and isn't currently tracking your assessment. "Change what the cues are doing" is an instruction that has somewhere to land.
What we do with it
Myo has no "liking" score. That's a design decision, not an omission.
Craving risk in the model rises with a depleted baseline, accumulated chronic load, and the opposing process — never with pleasure. There's no term where enjoying something more reduces the pull, because in the literature it doesn't. Sensitization appears as an accumulator that strengthens with each prior peak, so a system that's been trained harder generates more pull from the same input.
The practical output is a number that goes up when things are getting worse in the specific way that predicts tomorrow — not a mood score, and not a measure of how good your day was.
The structure follows the published work. The weights are ours, and we publish the mechanism and the literature behind it with an honest label on which is which.
The line to keep
Wanting and liking come apart, and only one of them sensitizes.
Which means the useful question isn't "why do I still want this when it isn't even good." That question has an answer and the answer is architectural. The useful question is which cues your wanting system has been trained on, and what would happen if you changed them.