People with long histories of regular drug use tend to be inconsistent in how they use information about negative effects to guide their decisions. This contradiction suggests a new explanation for why some people continue to use substances despite experiencing severe personal and economic costs. These findings were recently published in the journal translational psychiatry.
The main characteristic of severe drug use is the tendency to continue using drugs or alcohol even when it has negative consequences. These negative effects can include strained family relationships, financial debt, and serious health problems. To understand this behavior, scientists study how people evaluate the costs and benefits of their actions.
Deciding whether to use a substance requires comparing various potential costs. For example, a person may weigh the social costs of continuing to use against the costs of discontinuing use, such as withdrawal symptoms or loss of social circle. This decision-making process often takes place in complex environments that can change without warning.
“People who struggle with drug use often continue despite negative consequences (strained relationships, health problems, career problems, etc.), but little is known about how they actually weigh those costs when making decisions,” says Sonia G. Lewis, a doctoral candidate in clinical psychology at Yale University.
“For example, how do we compare the interpersonal strain that comes with using to the loneliness of discontinuing use, even though social circles are still used to determine which costs are ‘worth it’?” Lewis explained. “Most research focuses on how people respond to rewards or learn how to avoid just one cost in situations that don’t change. We wanted to test how people compare multiple potential costs in situations that change.”
To fill this knowledge gap, the researchers designed an experiment to test how changes in the environment affect risky choices. They recruited 137 adults from New Haven County, Connecticut. This group was selected to represent a wide range of substance use histories. The scientists excluded people with mental disorders, bipolar disorder, or medical problems that might interfere with understanding the tasks.
Scientists interviewed each participant to determine the total number of years they regularly used drugs. Regular use was defined as using a substance three or more times per week. The researchers added up the number of years of regular use of a variety of substances, including alcohol, cannabis, opioids, and stimulants.
This cumulative number served as the primary measure of substance use severity in the experiment. Approximately 75 percent of the sample had regularly used at least one substance over their lifetime. Participants earned $15 an hour and could earn up to $10 in bonuses based on their performance in the game.
Participants completed a specialized decision-making task on a computer. For each of the 200 rounds of the task, participants had to choose between a black card and a white card. Each card displayed a random amount between $1 and $5 representing a potential loss.
Behind the scenes, there was always a high chance that a single card would cause a participant to lose their money. If a participant selects the wrong card, they will lose the amount shown on the screen. If you choose the correct card, the loss will be avoided and your loss will be zero.
Participants were told to build a sense of which card was safer based on past results. You can monitor overall performance via the green bar at the bottom of the screen. This bar tracks the amount of money participants have cumulatively lost during the game.
The task was split into two different environments to test different types of learning. In a stable environment, a single card had a 75% chance of causing a loss for 100 rounds in a row. Participants were not informed when they were in a stable or unstable environment.
In an unstable environment, one card predicted a loss 80% of the time, and the other card predicted a loss 20% of the time. These probabilities switch between the two cards every 25 rounds. This design tested how people react when trusted patterns are suddenly disrupted.
To understand the thought process behind every choice, the researchers applied a mathematical model to the participants’ data. These computational models estimate how people learn over time in situations of uncertainty. The model calculates how quickly people update their beliefs and how consistently they use learned information to choose the best option.
Researchers found that people who had used substances regularly for more years behaved differently than those with a shorter history of use. Specifically, these people were less likely to repeat the same choice after successfully avoiding a loss. They tended to change their answers regardless of whether their previous choice resulted in a loss or zero dollars.
“We expected to find that people with higher severity of drug use would be less susceptible to loss, as most previous studies have suggested,” Lewis told Cypost. “We found that while they can learn from cost information, they don’t consistently translate it into choices.”
“It’s not so much a lack of learning ability as it is something that is preventing you from using that learning in the moment (which probably varies from person to person),” Lewis explained. “This nuance is actually very important because it allows us to have interventions that focus on helping people use the information they already have more consistently.”
Mathematical models provided evidence that these individuals have a fundamental discrepancy in how they use value information. They were not at all sensitive to negative costs. Rather, they were unable to consistently apply what they learned about probabilities and amounts to guide future choices.
“People with high levels of drug use learn about what is expensive just like everyone else, but what makes them different is how they use that information to make choices,” Lewis said. “Even if you understand the consequences, your choices may seem random or inconsistent because something is getting in the way of you acting consistently on what you’ve learned.”
“This is consistent with a situation we often see in practice: people know what the costs are, but that knowledge is not always consistently reflected in their choices,” Lewis said. The most notable discrepancies occurred during the stable portion of the task. For this part, the rules remained exactly the same, but the choices seemed more random.
In interpreting these findings, researchers note that the effects are relatively modest. “The strength of the relationship we found between drug use and differences in decision-making is consistent with other studies, but it is modest and only helps explain part of the picture,” Lewis said. “I want to use these results to think conceptually about how human decision-making works and what gets in the way of using what we know to make choices.”
The authors also note some potential limitations and misconceptions that should be avoided. “First, we can identify that information about cost is not used consistently, but we still cannot say why it applies to a particular person (for example, it may depend on someone’s emotional state, the specific situation, or someone’s history),” Lewis explained.
“Second, this study is a first step. Our sample was relatively small, and the costs of our work were monetary rather than material, so future research should replicate these findings with larger samples and with stimuli more relevant to the decision-making situations people actually face,” Lewis added.
The study also relied on a single measure of substance use severity that combined past and current use into one total value. Active use can have different effects on the brain than past use, so grouping everyone together can mask important differences. The authors caution against using these specific findings for clinical prediction.
“Third, this study was not designed to predict individual behavior, so these results cannot be used to predict who is likely to develop a substance use disorder or relapse,” Lewis cautioned. “These results can be used to help understand the underlying decision-making process.”
The researchers suggest that future tests could separate current and past users to better understand how decision-making changes during recovery. Emotional distress and a lack of working memory, the brain’s ability to retain and process information in the short term, can also contribute to inconsistent decision-making.
“Real-world choices require comparing costs and benefits simultaneously, so the immediate next step is to incorporate rewards into decision-making tasks,” Lewis said. “There is a manuscript under consideration that addresses this.”
“In the long term, I think exploring how people respond to decision-making situations (e.g., is the situation fair? Do their choices affect the outcome?) will be important for understanding why cost information is sometimes used to make choices and when it is not,” Lewis concluded.
The study, “The Relationship Between Regular Substance Use and Cost Comparison in Stable and Unstable Learning Environments,” was authored by Sonia G. Lewis, Samuel Paskewitz, and Ariel Baskin Somers.
