Five papers on how games taught a generation to read chance as skill, and what happens when feedback stops being about you.
This is the first half of a two-part research arc. Five papers, written across 2026, asking one question from five directions: how does a generation raised on mastery-shaped systems end up in relief-shaped ones, and what changes in them on the way? This half is the diagnosis. The second half is about what to build instead.
A note on why it sits under craft. Craft is the principle about tight feedback loops: short reps, clear focus, learning through iteration. Everything below is about feedback loops that look exactly like craft and teach nothing.
Each section links the paper it draws on. All five are free to read.
1 · The disguise
Contemporary games convert gambling’s core mechanic, the variable-ratio schedule, into something that feels like a mastery journey. The method is fusion. Unpredictable, high-salience rewards get wrapped in visible signals of effort: XP bars, ranks, battle passes, mastery tracks. The language does heavy lifting. Forge. Craft. Upgrade. Mastery chest. Every one of those words describes work.
The sharpest device is the skill-coded gate. You have to win, or hold a streak, or reach a rank, before you may open the random thing. The requirement is real. The randomness behind it is untouched by it. But the win happened first, so the reward inherits its credibility.
A 2019 UK prevalence review found randomised reward mechanics in roughly 71% of top-grossing titles. Surveys and meta-analyses since have found consistent small-to-moderate associations between loot-box spending and problem-gambling symptoms, in adults and adolescents alike.
The neurological account is unglamorous and well established. Anticipation and reward-prediction error amplify dopaminergic response. Near-miss outcomes and illusions of control raise arousal and sustain “one more try” behaviour without altering a single probability. Self-Determination Theory supplies the motivational half: games satisfy a genuine need for competence through skill-contingent feedback, so when chance outcomes are framed as earned, perceived competence inflates on evidence that was never there.
That is the part worth sitting with. This design works because the underlying hunger is legitimate.
Regulation has been uneven, with Belgian enforcement, Chinese odds disclosure and platform rules pulling in different directions, and the industry’s own mitigations tend to be cosmetic. Pity timers and battle passes reduce visible frustration without restoring any real link between skill and outcome. They fix the feeling and leave the structure.
2 · The chase
The second paper asks why young men in particular re-enter and escalate after losing.
At the neurocognitive core, the act of betting partially mimics a reward signal on its own. Reward-prediction-error mechanisms in the ventral striatum and anterior insula respond to the wager, not only the result, which blunts the negative signal that would otherwise prompt someone to stop. Under high-frequency, variable-ratio conditions, engagement decouples from monetary outcome altogether.
Development compounds it: heightened reward sensitivity, steeper delay discounting, and less mature prefrontal control over striatal systems bias post-loss choices toward immediate get-even action. This is a schedule, not a character defect, and products are built for it.
Then the cultural layer. Masculine norms favouring emotional suppression raise physiological arousal and cognitive load after a loss. Narrowed reflective capacity channels distress into action. And where losses are visible, on leaderboards or public profit-and-loss feeds, an ordinary error becomes a status threat. Same money, different injury, and the second one escalates.
The interventions that survive scrutiny are dull: triggered cooldowns after loss streaks, cumulative net-loss made impossible to avoid seeing, muted near-miss feedback, opt-in rather than default rankings. One is worth naming for later. Recognition for walking away. That changes what gets scored.
3 · The concealment
The Sociological Rupture (PDF)
The third paper moves from the individual to the household, and produces the single most useful idea in the set.
The risk signal is not the channel. It is loss velocity plus concealment: how fast money moves, and how well the movement hides.
Cash used to be a governor. You could see the pile shrink. Cashless, fragmented, credit-enabled payments do more than accelerate losses, they obscure the total. Multi-accounting and instant credit finish the job.
What follows is the mechanism that makes this so much worse than a financial problem. When losses collide with provider and breadwinner ideals, they register as evidence of failed competence, and that produces shame rather than guilt. The distinction is load-bearing. Guilt says I did a bad thing and moves toward repair. Shame says I am the bad thing and moves toward secrecy.
Secrecy is the harm multiplier. It removes corrective feedback from other people and leaves the loop running with nothing outside it, which is how this arrives at relationship rupture, over-indebtedness, depression and elevated suicidality rather than merely at debt. Policy that constrains credit availability reduces acute risk. Detection built around velocity and concealment outperforms detection built around channel.
4 · The story
The fourth paper examines how these products are sold, and finds the persuasion is not in the claims.
Advertising analysis shows sports-betting marketing emphasising knowledge, expertise and controllability while minimising chance and harm. Interface studies find the same argument in the mechanics: dashboards, streak counters, build-a-bet tools, coaching-style copy about strategy and discipline. Gamified trading and betting interfaces produce hot decision states, with attention herding, rapid repeats and poor median returns, among largely young male cohorts.
Endorsement of traditional masculine norms, meaning risk-taking, emotional control and self-reliance, correlates with greater gambling frequency, higher problem severity, and reduced help-seeking. The same trait cluster that makes the product attractive makes the exit harder.
And the cruellest move in the whole set: qualitative work documents losses being reframed as failures of discipline rather than as structural mathematics. A person who loses to a house edge concludes he was insufficiently disciplined, and responds by trying harder at something that cannot be won. Stigma around asking for help follows directly.
Which is why “gamble responsibly” underperforms. It is a disclosure aimed at claims, in a system that runs on interface and social proof. It regulates the one layer that is not doing the work.
5 · The rewiring
The fifth paper is the synthesis, and it makes the strongest claim.
Mastery contexts, meaning ranked ladders and competitive play, supply tight action-outcome contingencies, dense feedback and visible skill curves. These strengthen what the literature calls controllability priors: a working belief that your actions shape outcomes. Relief-oriented products employ variable-ratio reinforcement, near-misses, high event speed and opaque odds. These decouple reward from control, raise arousal, deliver short-term mood relief, and blunt the informational value of prediction errors.
Sustained migration from the first to the second is argued to recalibrate the person, from an agentic stance (my actions shape outcomes and teach me things) toward a passive, dissociative one (outcomes happen to me; I play to feel different).
Harms concentrate in rapid, continuous digital formats and are overrepresented among young men. Problem gambling correlates with illusions of control, distorted beliefs about randomness, elevated dissociation, and a measurable shift from goal-directed to habitual responding. Exposure to gambling-like mechanics, loot boxes above all, predicts later gambling involvement.
What the five agree on
Read together, the arc is not about money and it is not about willpower. It is about attributability: whether the feedback you receive is about you.
Games are unusually kind environments. Act, receive a signal that reflects your action, adjust. That is craft. Relief products preserve every surface feature of that loop, the speed, the stakes, the ritual, the feeling of being absorbed, while severing the one property that made it teach. From the inside, the two are close to indistinguishable. That is the entire product.
The recommendations converge accordingly: screen for relief motives and dissociation rather than for spend; score risk at the level of features, meaning event speed, near-miss salience and availability, rather than of category; deploy default friction; and pursue mastery-restoring interventions rather than relying on the regulatory column alone.
What this half does not settle
Honest limits. Much of the evidence is cross-sectional, so causation and selection remain tangled. Whether relief-seeking causes the migration or the migration causes relief-seeking is not resolved here. Direct longitudinal measures of controllability priors barely exist. Cultural and gender moderators are thin. And the strongest recommendation the set produces, restore mastery, is asserted more often than it is tested.
The takeaway, plainly: the problem is not that these systems are too fast or too rewarding. It is that they removed the only property that made speed and reward worth anything, which is whether the result had anything to do with you.
Part two takes up the obvious next question, and finds the obvious answer is wrong.