Reinforcer-specific value-based decision-making in persistence of and recovery from alcohol use disorder

Lead Research Organisation: University of Sheffield
Department Name: Psychology


Alcohol places a considerable burden on society; it was responsible for more than 25,000 deaths and over one million hospital admissions in England in 2019. Much of the burden of alcohol-related harm can be attributed to the heaviest drinkers, of whom there are half a million (people with Alcohol Use Disorder; AUD) in the United Kingdom. The majority of people with AUD in the UK do not receive treatment and of those that do, most do not benefit from it.

Scientific progress in our understanding of AUD and other addictions, and the search for new evidence-based treatments, has been slow. This is partly because many addiction scientists believe that addiction is compulsive and habitual, meaning that drinking and drug use are behaviors that become insensitive to their consequences. However, there is evidence that all motivated behaviour, including the apparently irrational behaviour of people with addiction, is determined by the goals that people have, and the values that underlie their goals. Behavioural economics has applied this idea to the study of addiction, and has amassed a large body of evidence and coherent theories that can explain how addictions develop and persist, how treatments work, and how people recover from addiction even if they do not receive treatment.

Despite these important advances, the mechanisms through which addiction disrupts the psychological and neural processes that underlie value-based decision-making (VBDM), are largely unknown. Therefore, there is a need for neuroscience-informed methods that can characterise the internal processes that determine choice for alcohol and other rewards (particularly activities that are incompatible with drinking alcohol), and demonstrate how those processes are implicated in the persistence of AUD, recovery, and treatment response. To this end, we developed a computerized VBDM task and used computational methods to interpret participants' responses on the task. Our preliminary findings suggest that the task can discriminate people who currently have addictions from people who have recovered, and it is sensitive to intense alcohol cravings.

In this project we will apply these methods to systematically characterise the internal mechanics of value-based choice in AUD. The overarching aim is to demonstrate how distortions in VBDM contribute to important features of AUD, specifically characterisation of people with AUD versus people who have recovered, the influence of powerful cravings on drinking behaviour, and the response to treatment. To achieve these aims we will complete the following package of work. First, we will attempt to distinguish people who currently have AUD, people who have recovered from AUD, and a control group of light drinkers, on the basis of their VBDM. Second, we will conduct a laboratory study in which we will experimentally increase alcohol craving in people with AUD before measuring their VBDM and recording how much alcohol they voluntarily consume. This will tell us how VBDM changes during craving episodes, and which specific aspects of VBDM are predictive of alcohol consumption. Third, we will track changes in VBDM over the course of a three-week community-based behavioural economic treatment for AUD in which participants receive financial incentives proportional to the extent that they are able to reduce their drinking (contingency management). This will identify which specific aspects of VBDM are predictive of reductions in drinking at follow-up.

Important outcomes from the project include advancing the conceptual understanding of addiction by corroborating a novel framework of the drivers of value-based choice of alcohol versus alternative reinforcers. We will also validate a new measurement tool of individual differences in the cognitive-motivational constructs that distinguish people who are on track to recover from their addictions, and those who require additional support; this tool can be exploited in future work.

Technical Summary

This project applies computational methods to characterise the internal mechanics of value-based decision-making (VBDM) in alcohol use disorder (AUD). We developed a decision-making task in which participants choose between alcoholic drinks (in one block) and alcohol-free reinforcement (in a different block). A drift diffusion model is fitted to choice and response time data, to yield VBDM parameters (evidence accumulation rate and response threshold) for each reinforcer. Specific objectives are as follows:

1. We will identify the VBDM profiles that characterise AUD and recovery. Using a case-control design, the following groups will complete the VBDM task: (1) people with AUD, (2) people who have recovered from AUD, and (3) a control group of light drinkers.

2. We will identify how VBDM changes during craving episodes, and which VBDM parameters are predictive of drinking behaviour. Using a between-subjects design, participants with AUD will be randomized to complete either an alcohol craving induction procedure or a control procedure, before completing the VBDM task and offered alcohol to drink in the laboratory. Dependent measures are VBDM parameters and the volume of alcohol consumed.

3. We will track changes in VBDM following a contingency management intervention, and identify which VBDM parameters are predictive of reductions in drinking at follow-up. Using a between-subjects design, participants with AUD will be randomized to either a contingency management intervention in which they receive financial incentives for abstaining from alcohol over three-weeks, or a yoked control intervention. Dependent measures are VBDM parameters at the end of the intervention period and drinking behaviour at one-month follow-up.

Outcomes from the project include validation of a conceptual framework and measurement tool of the cognitive-motivational constructs that impede recovery. In future work, those constructs could be targeted by personalized interventions


10 25 50