For much of behavioural science and psychiatry, compulsive digital use has been explained through a relatively simple framework. Dopamine is treated as the central driver of reward. Social media, gaming, and short-form content are described as systems that "hack" this pathway. Addiction becomes a matter of overstimulation. Regulation becomes a matter of withdrawal (Volkow et al., 2016).
That framing has helped popularise interest in digital mental health. It has also flattened a more complex set of behavioural and cognitive processes into a single neurochemical narrative. Many users do not exhibit clinically defined addiction, yet still describe patterns of use that feel automatic, difficult to interrupt, or disproportionate to intent. The gap between perceived loss of control and clinical classification remains significant. This is where current research into compulsive online behaviour begins to shift, not away from reward systems, but away from overly simplified interpretations of them (Kardefelt-Winther, 2014).
The central hypothesis emerging across recent literature is more restrained. It is not that dopamine is "overactive" or "depleted", but that digital environments are structured around highly optimised reinforcement schedules. Variable rewards, intermittent feedback, and constant novelty shape attention and behaviour through well-described learning mechanisms. In this sense, the issue is less about a single neurotransmitter and more about how behaviour is trained over time through repeated exposure to specific patterns of reinforcement (Berridge & Robinson, 2016).
Recent behavioural and clinical studies on problematic social media use support this distinction. Much of the observed engagement appears better explained by habit formation and cue-driven behaviour than by classical addiction models. Individuals often report strong urges to check platforms, but these urges are frequently linked to contextual triggers such as boredom, notification cues, or routine rather than persistent compulsive drive in the clinical sense (Brand et al., 2019).
The dopamine narrative remains popular largely because it offers a simple and intuitive explanation for complex behaviour. It is also easier to communicate than more nuanced models of reinforcement learning and behavioural conditioning. The downside is that it can draw attention away from other important factors, such as attentional processes, platform design, and the learned habits that keep people engaged over time.
More recent work in computational psychiatry and behavioural science has begun to frame digital engagement in terms of reinforcement learning models. Within this framework, platforms are not understood as sources of direct neurochemical manipulation, but as environments that shape prediction, reward expectation, and habit formation. Behaviour emerges from the interaction between individual sensitivity to reward and the structure of the environment in which that behaviour is repeatedly reinforced (Huys et al., 2016; Montague et al., 2012).
This distinction is particularly relevant when considering the boundary between high engagement and pathological use. Current evidence suggests that many individuals who self-identify as "addicted" to digital platforms do not meet clinical thresholds when assessed against behavioural addiction criteria. At the same time, this does not negate the presence of functional impairment or subjective distress in specific cases. The challenge lies in distinguishing between intensive but adaptive use and patterns that reflect diminished behavioural control (Kardefelt-Winther et al., 2017).
Across recent studies, a consistent limitation remains the lack of stable diagnostic criteria for what constitutes "problematic" digital behaviour. As a result, prevalence estimates vary widely depending on measurement approach. This variability complicates both clinical interpretation and policy development, particularly in contexts where behavioural labels carry normative weight (Billieux et al., 2015).
The emerging direction in research is therefore less focused on identifying a single causal mechanism and more focused on modelling digital behaviour as a dynamic system. Attention, reward sensitivity, impulse control, and environmental structure are increasingly treated as interacting variables rather than isolated explanatory factors. This aligns more closely with contemporary models of decision-making than with older dopamine-centric accounts (Huys et al., 2016).
For the people building in this space, the shift matters. Products designed around simple reward dysregulation tend to reach for blunt instruments: usage caps, detox features, screen-time alarms. If compulsive engagement is instead shaped by reinforcement structures embedded in the design itself, the more useful interventions are quieter ones, concerned with where friction sits, how cues are timed, and what a default does when no one is paying attention.
This does not discard the neurobiology. It complicates it. Compulsive digital behaviour looks less like a chemical being hijacked and more like a behaviour trained over time through repeated contact with environments built to reinforce it. For the people designing those environments, that is the difference between treating a symptom and shaping the conditions that produce it.
🪺