Craving Doesn't Break Your Willpower — It Rewrites the Math Your Brain Runs on Every Choice

You know the moment before you know it’s happening. A smell, a taste that isn’t there, a physical hum under your skin like the substance is already in the room even though it isn’t. A 36-year-old man in Ethiopia, three relapses behind him, described it to researchers this way: “I feel the substance’s presence through smell, taste, and physical sensation despite its absence.” He wasn’t talking about a memory. He was describing a body that had already made a decision before his mind caught up.

Here’s the thing new brain science finally has a number for.

Craving isn’t a mood or a weakness — it’s a measurable change in the actual math your brain runs when it weighs risk and reward, which means “just use willpower” was never coherent advice.

That’s not a metaphor anymore. It’s the finding of a 2026 study in Nature Mental Health led by Yale researcher Kaustubh Kulkarni, and it’s worth sitting with because it explains something most craving advice gets backward: it treats craving as a feeling to manage, when it’s actually an event that changes how you compute.

The study put 132 people who reported moderate-to-heavy alcohol or cannabis use through a repeated-choice task — two “slot machines,” one paying out 80% of the time, the other only 20%, with which one was winning switching several times over the session while participants tried to maximize their take. Some played for money. Others played for images of the thing they craved — a beer, a bong. Researchers used computational modeling — essentially, math that traces exactly how a person updates their next choice after a win or a loss, move by move — to track what craving did underneath the surface, moment to moment, as people played.

What they found: rising craving didn’t just make people want the substance-linked outcome more. It changed the learning rate itself — the actual rule the brain used to update from what just happened. Alcohol craving sped that updating up. Cannabis craving slowed it down. As Kulkarni told Yale Medicine, “We’ve now demonstrated there’s some computational dysfunction that has occurred” — and it showed up specifically around the substance-related choices, not the money ones, meaning it wasn’t general impairment. It was targeted. Senior author Xiaosi Gu put the stakes plainly to Yale: “Addiction is very hard to treat, and one of the reasons is that we don’t fully understand the craving linked with it.” The alcohol and cannabis mechanisms also ran in opposite directions, which the researchers and outside writeups have flagged as evidence that a one-size-fits-all approach to treating craving is already scientifically outdated — what recalibrates an alcohol-craving brain may not touch a cannabis-craving one.

Now put that finding next to what people who’ve actually lived it have been saying, in their own words, all along.

A separate 2026 qualitative study in Scientific Reports interviewed nine men in treatment at a hospital in Addis Ababa, Ethiopia about their relapse experiences — a different country and a different method than the Yale lab, but the same underlying psychological event: sensory craving, shame, and cue exposure pulling a person back into use. It’s worth naming honestly that this was not a U.S. sample — but the mechanism these men describe is one clinicians and researchers across cultures recognize, which is exactly what makes it useful here: it’s independent confirmation of the same event the computational study measured, from the inside. A 53-year-old man with five prior relapses connected his craving directly to collapse in the rest of his life: “I feel deep failure, stemming from both romantic and occupational instability, which has contributed to my renewed substance use.” A 40-year-old man held two full months of recovery in a residential setting and relapsed almost immediately after returning to his old address — proximity did what argument couldn’t. A 37-year-old man, five relapses in, described the same geography problem from the treatment side: recovery programs placed close to where substances were easy to get produced a repeating loop of recovery and relapse, not because he didn’t want to stay clean, but because his environment kept re-triggering the exact computation the Yale study found — reward-weighting tilted toward the cue, over and over, in the same location where it had tilted that way before.

None of that is a character flaw showing up on a schedule. It’s a cognitive event with an environmental trigger, recurring in a brain that’s been trained, repeatedly, to run this specific equation in this specific place.

It’s a cognitive event with an environmental trigger, recurring in a brain that’s been trained, repeatedly, to run this specific equation in this specific place.

That’s also the treatment implication that should actually change practice, explored more in Rize’s science and medicine coverage: craving deserves substance-specific intervention, not a generic “urge surfing” script applied identically to a person struggling with alcohol and a person struggling with cannabis, and treatment placement that ignores exposure to old cues is treatment working against its own math. If the environment recreates the equation, the equation recreates the relapse.

You didn’t fail a test of character in that moment by the smell, the taste, the hum under your skin. Your brain ran the equation it’s been trained to run. The science just caught up to what you already knew from the inside — and now it has a name.