From The Vault: Craving predictability

The thin blue curl of smoke dances into the night sky from Bogart’s cigarette as their eyes meet. The film’s black and white denouement is upon us but all they can see is that cigarette. They can almost taste it. They are the smokers and the recently ex-smokers whose neuronal circuitry is lighting up in anticipation of that next cigarette.

Some of them will acquiesce and slink off silently as the credits roll whilst others will shift uncomfortably in their seats and wait for the cravings to subside. Whatever action they take it will be of no suprise to the team of researchers at UCLA’s Laboratory of Integrative Neuroimaging Technology, who have been using functional MRI techniques to observe the changes which arise in the brain during craving.

In the study, participants underwent fMRI scans whilst watching films designed to induce cravings, be neutral towards smoking, or simply watched no film at all. In all cases though they were instructed to fight their cravings.

The data produced by the scans was analysed and traditional machine learning methods were augmented by Markov processes in an effort to produce predictive algorithms. In essence, the researchers found that by observing the changes in the underlying neurocognitive structures of the participants, they were able to predict what film the participant’s were watching, whether cravings were being induced and perhaps most importantly whether the cravings were being resisted.

“We detected whether people were watching and resisting cravings, indulging in them, or watching videos that were unrelated to smoking or cravings… Essentially, we were predicting and detecting what kind of videos people were watching and whether they were resisting their cravings.”said Dr. Ariana Anderson, the study’s lead author.

In layman’s terms the algorithm produced worked much in the same way as Google does in predicting an entire search based on only a few words. So the algorithm would note that the activation of a particular neural network was indicative of the inducement of a strong craving for cigarettes.

Learning which neural networks are responsible for our cravings and urges gives researchers a clearer understanding of the obstacles that those of us with addictions face when our cravings surface. From coffee to cigarettes, chocolate to the illicit side of the spectrum, these machine learning methods may eventually provide a real-time biofeedback system capable of telling us when a craving is surfacing and how intense it may well be. These neural seismographs will give us a mechanism by which to train and suppress our cravings. Or perhaps they’ll just tell us that the single shot latte just isn’t going to cut it this morning. Better make it a double.

Leave a Reply