GADgET - Online Gambling Addiction Detection

Description

Ever since the worldwide demand for gambling services started to spread, its expansion has continued steadily. The main factor contributing to this ongoing expansion is the explosion of telecommunication technologies that has facilitated the development of new gaming services, along with a wide variety of new delivery channels for gambling services [5]. According to the European Gaming and Betting Association, Europe currently represents the largest international market for online gambling, with a Gross Gaming Revenue (GGR) expected to reach €24.9 billion in 2020 [2]. To wit, online gambling is a major industry in every European country, generating billions of Euros in revenue for commercial actors and governments alike. Despite such evidently beneficial effects, online gambling is ultimately a vast social experiment with potentially disastrous social and personal consequences that could result in an overall deterioration of social and familial relationships [3]. In 2012, the European Commission released a statement [5] highlighting the need for regulatory policies to aid in the detection of pathological gambling behaviors, citing “a responsibility to protect those citizens and families who suffer from a gambling addiction.” Despite this earnest recommendation, research aimed at developing a coherent set of responsible national policies has yet to make it past the embryonic stage. In Portugal, the Gambling Inspection and Regulation Service is “responsible for the control and regulation of gambling activities in casinos and bingo halls, as well as online gambling and betting.” To comply with operational objectives, this authority receives, on a daily basis, all data related to online gambling activities pursued by every user on every online platform with services that are accessible to Portuguese citizens. In spite of this prodigious collection of data, the authority still lacks appropriate tools for identifying gambling addicts. This authority acknowledges a profound scarcity of actionable data regarding the actual scope of gambling addiction, and a consequent lack of expertise about how best to deal with this problem. The same authority observes that “the human dimension and economical and social relevance of this issue (i.e., gambling addiction) demands scientific studies.” To answer this call, this project proposes an AI (Artificial Intelligence)-based tool that could capitalize on the vast amount of data collected every day and analyze online user behavior to model and detect the behaviors associated with addicted gamblers. The problem represents a major challenge for a few reasons: first, the massive amount of data involved (that will require efficient data- analysis algorithms); second, the temporal dimension of the phenomenon we intend to model; and third, the fact that we are trying to observe and affect behaviors associated with a very small fraction of the population. The analysis confronts an additional complication by virtue of the potentially infinite behaviors associated with various kinds of gamblers. To tackle this problem, we propose a system based on a version of Recurrent Neural Networks (RNNs) the architecture of which will be optimized by a neuroevolution algorithm. RNNs are ideal modes for addressing problems that can only be resolved by using previous events to predict the future events of a system; RNNs have been successfully applied in a plethora of different domains [6]. To effectively resolve the problem under consideration, this system must be able to render efficient comparisons of time series associated with different gamblers’ behaviors, in a way that also takes the temporal dimension of the problem into account. The system, therefore, must be able to: (1) identify common behavioral patterns among gamblers within in an acceptable timeframe; (2) detect actions that are representative of a risky behavior in the context of gambling; and (3) run in real-time, to allow for continuous control of gambling activities. Successful implementation of the system and its integration with the system currently in use by the gambling control authority will enable efficient modeling and detection of online user behaviors associated with gambling addiction. Armed with this information, the authority could deploy all actions it regards as necessary. The social impact of the project is enormous, given its inherent capacity to reduce the social costs associated with gambling addiction. In sum, the project will address a socio-economic problem that at least one public authority regards as critical in the following ways: analyzing a large volume of administrative micro-data derived from the routine operation of the gambling control authority; implementing a new system that can model and detect online user behaviors associated with pathological gambling, by proposing a state-of-the-art AI system; and furnishing the gambling authority with a tool for the successful achievement of its regulatory duties.

Researchers

Funded by

FCT

Partners

Universidade Nova de Lisboa, Universidade de Lisboa, Universidade de Coimbra, Turismo de Portugal IP

Total budget

295 291,00 €

Local budget

55 228,00 €

Keywords

Human Behavior Modelling, Deep Learning, Big Data, Times Series

Start Date

2019-01-01

End Date

2021-12-31

Conference Articles