A large-scale analysis of nearly 2,600 users of a cannabis-reduction app identified three distinct psychological profiles based on why people use cannabis and how ready they are to quit. Coping-driven users showed the highest risk of problematic use yet the lowest confidence in their ability to change, while enhancement-seeking users had moderate risk but little awareness of it. These findings are descriptive and cannot yet guide personalized treatment.

Most digital cannabis-reduction tools treat all users the same, despite substantial variation in why people use cannabis and how motivated they are to stop. This matters because a person using cannabis primarily to manage anxiety occupies a fundamentally different psychological position than someone using it at social gatherings for enjoyment. Understanding these differences at scale is the first step toward designing digital interventions that meet people where they actually are, rather than where a generic program assumes they should be.

Cannabis use disorder is one of the most prevalent substance use conditions globally, yet digital tools designed to help users reduce or quit tend to treat their audiences as homogeneous. The Stop-Cannabis app, a free French-language mobile application developed in Switzerland, has been available since 2010 and integrates validated screening instruments directly into its user experience. This study performed a secondary analysis of data from users who completed the Marijuana Motives Measure (MMM), the Stages of Change Readiness and Treatment Eagerness Scale (SOCRATES), and the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) between 2010 and 2020. The investigators used five distinct clustering algorithms with a composite validity index to determine the optimal grouping of users by their motivational and readiness profiles.

The analysis identified three clusters: individually coping users (ICU, 44% of the sample), social and coping users (SCU, 26%), and enhancement-seeking users (ESU, 30%). ICU and SCU clusters scored higher on ASSIST problematic use risk, demonstrated greater recognition of their cannabis use as problematic on SOCRATES subscales, and placed higher importance on making a change. Paradoxically, these same groups reported lower confidence in their ability to change. ESU users, by contrast, scored at moderate ASSIST risk while showing minimal self-recognition of a problem. The authors suggest these profiles could inform subgroup-tailored app interventions, but acknowledge this recommendation is extrapolated; no outcome or engagement data were collected, and the self-selected sample limits generalizability to broader cannabis-using populations.

The Coping Paradox: Why High-Risk Cannabis App Users Are Least Confident They Can Change

Nearly 30 percent of people who download a cannabis-reduction app show up ready to enhance their experience, not quit it. And among those most likely to have a serious problem, the ones using cannabis to cope with emotional pain, confidence in their ability to change is at its lowest. These are not just interesting data points. They are a design challenge for every digital tool we build to help people change their cannabis use. This study from Wegener and colleagues does something deceptively simple: it takes 2,578 real-world app users, measures why they use cannabis and how ready they are to change, and sorts them into groups using a rigorous multi-algorithm clustering approach. What it finds is clinically intuitive but empirically important. The three resulting profiles, coping-driven users, socially and coping-motivated users, and enhancement-seekers, are not just statistically distinct. They describe three different relationships with cannabis, each calling for a different clinical conversation. The paper’s genuine contribution is in making this heterogeneity visible at scale, within a naturalistic digital health setting where most interventions still operate as if every user arrives with the same psychology.

The central methodological limitation, however, is one the authors themselves partially acknowledge but whose consequences they may underestimate. This is a cross-sectional analysis of a self-selected population. Sorting library books by genre tells you what kinds of books exist on the shelf, but it does not tell you which ones readers will actually finish or benefit from. Similarly, identifying three user profiles tells us who downloads a cannabis app and why, but it cannot tell us whether matching app features to those profiles would improve anyone’s outcomes. The absence of outcome data is not a minor gap; it is the gap. Without it, every recommendation the paper offers about tailoring content, about confidence-building for coping users, about normative feedback for enhancement seekers, remains a plausible hypothesis rather than an evidence-based strategy. Compounding this, the 27 percent of eligible users who did not complete all required questionnaires are entirely uncharacterized, and their exclusion could systematically distort the cluster profiles in ways we cannot assess. Surveying people who already walked into a gym about their fitness motivation tells you about gym-goers, not about everyone who could benefit from exercise. The same applies here: these clusters describe app users specifically, not cannabis users broadly.

What I would tell a patient is straightforward: this research suggests that people use cannabis for very different reasons, and those reasons matter for how ready someone is to change. If you are using cannabis mainly to cope with stress or difficult emotions, that is worth knowing, because addressing those underlying feelings may be as important as the cannabis itself. What I would tell a colleague is that the cluster typology is a useful conceptual heuristic, grounded in established motive literature, but it should not be treated as a validated treatment-assignment tool until someone runs the prospective adaptive trial this paper implicitly calls for. This study will not change clinical practice tomorrow, and it is not designed to. What it does is hand us a map, imperfect and drawn from a self-selected population, but coherent, of the psychological terrain that cannabis app users navigate. Three profiles, each with its own risk signature and motivational posture, each deserving something different from the tools we build for them. The real work begins when someone runs the trial to find out if matching the tool to the terrain actually helps people get where they want to go. Descriptive precision about who is in a population is not the same as prescriptive evidence about what that population needs, and that distinction separates hypothesis generation from evidence-based practice.

This study sits early in the research arc for personalized digital cannabis interventions. It provides the descriptive foundation that adaptive intervention trials need: empirically identified subgroups with distinct motivational profiles. For clinicians who use or recommend cannabis-reduction apps, the three-cluster framework offers a practical mental model. Coping-driven users may need concurrent attention to the anxiety, depression, or trauma that drives their use. Enhancement-seeking users may not recognize their moderate risk at all, making standard motivational approaches less effective. These distinctions align well with established motivational interviewing principles regarding the importance-confidence matrix.

From a pharmacological standpoint, the study’s emphasis on coping motives underscores the need to assess for comorbid psychiatric conditions and concurrent medication use before attributing all behavioral patterns to cannabis alone. Patients using cannabis to manage anxiety who also take SSRIs, benzodiazepines, or other psychoactive medications present complex pharmacological profiles that a motive-based screening cannot capture. The one concrete recommendation clinicians can implement now is this: when a patient discloses cannabis use, asking not just how much and how often, but explicitly why, can meaningfully shift the clinical conversation and may identify coping-driven users who need additional mental health support regardless of their cannabis trajectory.

This is an original cross-sectional study performing secondary analysis of observational data collected from a mobile health application over approximately a decade. It uses cluster analysis to identify user subgroups and does not test an intervention or follow participants over time. In the evidence hierarchy, it sits below prospective cohort studies and randomized trials, and its most important inferential constraint is that no causal, temporal, or outcome-related conclusions can be drawn from its design.

The finding that coping motives are associated with greater problematic cannabis use is well established in prior research using the Marijuana Motives Measure, including earlier work by Simons and colleagues who developed the instrument. This study extends that literature by demonstrating the association at scale within a naturalistic digital health setting and by pairing motive profiles with readiness-to-change indicators. A previous study using the same Stop-Cannabis dataset examined cannabis use motives but did not perform the cluster analysis linking motives to readiness and problematic use risk that this study contributes. The three-cluster solution is broadly consistent with typologies seen in other substance use research, though no directly comparable cluster analysis of a cannabis mHealth population exists for comparison.

The most consequential analytic choice was the preference for the 3-cluster solution over the 6-cluster solution, which was statistically comparable by the composite validity index. The authors justified this on the grounds of clinical meaningfulness, noting that the 6-cluster solution produced small clusters without significant inter-cluster differences. This is a reasonable judgment, but it introduces subjectivity. Had the 6-cluster solution been selected, a more granular typology might have emerged, potentially separating anxiety-driven coping users from those coping with chronic pain or social isolation. Additionally, including demographic variables such as age, gender, and cannabis use frequency in the cluster model could have produced meaningfully different groupings, given the well-documented influence of these factors on use motives.

The most likely overinterpretation of this study is treating the three clusters as validated cannabis use disorder subtypes that should guide clinical treatment decisions. They are not. The clusters describe a self-selected group of people who downloaded a specific French-language app and completed all of its questionnaires, a highly particular population that may not represent cannabis users in general or even cannabis users who seek help through other channels. Relatedly, the association between coping motives and greater problematic use should not be read as evidence that coping motives cause problematic use. Third factors, particularly comorbid anxiety and depression, almost certainly underlie both, and the cross-sectional design cannot disentangle these relationships. Finally, the paper’s recommendations for tailoring app content to clusters, while clinically sensible, are explicitly the authors’ extrapolation and have not been tested in any intervention study.

This study contributes a statistically rigorous, clinically coherent three-cluster typology of cannabis app users that reveals meaningful heterogeneity in use motives and readiness to change. It does not establish that these profiles are stable over time, generalizable beyond app users, or predictive of treatment response. For clinical practice now, its primary value is as a reminder that asking patients why they use cannabis, not just how much, is the beginning of a more productive conversation.

Does this study prove that cannabis apps should be personalized for different users?

No. The study identifies three distinct user profiles, which suggests personalization could be beneficial, but no intervention was tested. Whether matching app features to user type actually improves outcomes remains an open question that requires a prospective trial to answer.

I use cannabis mainly for stress relief. Does this mean I have a more serious problem?

Not necessarily, but the study does find that people who use cannabis primarily to cope with difficult emotions tend to score higher on measures of problematic use. This association does not mean coping use automatically equals a disorder. It does suggest that exploring what you are coping with, and whether other strategies might help, is a worthwhile conversation to have with a healthcare provider.

Can I use these results to figure out which “type” of cannabis user I am?

The three profiles described in this study, coping-driven, socially motivated, and enhancement-seeking, offer a useful framework for self-reflection. However, these clusters were derived statistically from a specific app-using population and are not designed as a diagnostic tool. A clinician can help you explore your personal motivations and their implications far more reliably than self-categorization based on a research typology.

Are cannabis-reduction apps effective?

This study did not measure whether the Stop-Cannabis app helped users reduce or quit. Some prior research suggests that digital health interventions for substance use can produce modest benefits, but the evidence base for cannabis-specific apps remains limited. This study’s contribution is in understanding who uses such apps, not whether the apps work.

Have thoughts on this? Share it: