Asian Journal of Probability and Statistics

Online dating fraud against older adults in the United States causes financial losses in the billions of dollars and harms that extend well beyond money, including psychological distress, damaged relationships, and long recovery times. Existing accounts of the problem are largely descriptive: stage models explain how a scam unfolds and official statistics count reported cases, but neither connects the risk faced by one user in one relationship to the burden carried by the population, so the effect of a platform or policy intervention cannot be estimated before it is tried. A model that makes that connection would let interventions be compared on the harm they remove rather than on the number of accounts they block.

Aim and Objectives: The aim of this study is to propose a model of online dating fraud against older adults in the United States that connects individual level risk with population level burden. Its objectives are to specify the dyadic processes that carry an engagement toward exploitation, to derive the population dynamics those processes produce, to measure harm as a burden with financial, psychological, social, and temporal components, and to set out how the model can be estimated and tested. The purpose is practical: to give platforms and policymakers a way to compare interventions before deployment. The justification is that decisions about warnings, filtering, and account termination are already being made, currently without a framework that states what each option is expected to achieve.

The Digital Relationship Exploitation Risk Model (DRERM) is specified at two linked scales. At the level of a single user and correspondent it is a discrete time hazard model in which progression to exploitation depends on contact exposure, contextual vulnerability, deception indicators, financial escalation, elapsed duration, reporting, and platform intervention, with intervention also weakening the effect of deception. At population level the dyadic hazards aggregate into a flow model driven by offender supplied fraud pressure, with steady states derived in closed form. Harm is tracked as four separate components: financial loss, psychological distress, social disruption, and recovery duration. The model is examined through a worked scenario analysis using explicit synthetic parameters.

In the baseline scenario, 19.4% of engagements progress to exploitation, giving a steady state of 318 exploited persons per 100,000 users and 15,873 exploited person weeks over 52 weeks. Faster platform termination of deceptive engagements reduces exploited person weeks by 42.8%; warnings that slow progression, by 21.5%; pre contact filtering, by 23.5%; and the first two combined, by 56.0%. All parameters are synthetic and illustrative, not empirical estimates. DRERM therefore turns general calls for platform action into testable quantities: how much harm an intervention removes depends on when it acts and on which component of burden it reaches. The article sets out the estimation and validation protocol and the falsifiable hypotheses needed to test the model on platform data.