ARGO - An AI Based Responsible Gamification Framework for Online Skill Gaming Platform

ARGO - An AI Based Responsible Gamification Framework for Online Skill Gaming Platform

ARGO - An AI Based Responsible Gamification Framework for Online Skill Gaming Platform

Pulkit Agrawal, Aditya Pareek, Rukma Talwadker, Tridib Mukherjee

Abstract

"Games of skill provide an excellent source of entertainment to real-ize self-esteem, relaxation and social gratification. Engagement in online skill gaming platforms is however heavily dependent on the outcomes and experience (e.g., wins/losses). There is an inherent tension between being: (i) demotivated (and hence dis-engaged) and ultra-motivated (to chase wins) in response to losses; or (ii) pos-itively engaged and over-indulged under a winning streak. These extreme transient intents make any gamification effort non-trivial in outcome-oriented platforms because of the inherent trade-off between engagement and prudence (avoiding over-indulgence). Pro-viding long challenges and high milestones can generate immediate engagement bump while adversely impacting its sustenance (re-sulting in possible churn eventually) depending on the players’ intent. Unlike in most consumer internet platforms, the intent and relevancy (challenge completion) become intrinsically orthogonal. This paper presents ARGO, a deep learning based responsible gamification framework, which: (i) learns players’ long term “game behaviours” (as psychological imprints emanating from game play evolution), via SOTA collaborative neural network, CognitionNet; (ii) succinctly combines these game behaviours to estimate players’ inherent “intent” towards a challenge, via an engagement guided regressor; and (iii) addresses the issue in building an imbalanced classification model for challenge completion via quantile regres-sion with a novel data augmentation trick. Using online Rummy as a case-study, we show that the player engagement can be maximized based on both intent and challenge completion likelihood. Personal-izing the challenge milestones accordingly results in—15% lift in milestone completion, 14% decline in unsuccessful target chasing with 4% improvements in target completion rates for the target chasers and 7% reduction in loss chasing and 5% improvement in win rates for the over-indulging players. We contend that ARGO is first-of-a-kind framework for outcome-oriented platforms that— performs personalized gamification contingent to both: (a) players’ transient intent and (b) challenge relevancy—thus balancing the trade-off between engagement and over-indulgence."

Reference

Agrawal, P., Pareek, A., Talwadker, R., & Mukherjee, T. (2023). ARGO - An AI based responsible gamification framework for online skill gaming platform. In Proceedings of the 2023 ACM SIGKDD Workshop on Online and Adaptive Recommender Systems (OARS '23) (pp. 1–8). Association for Computing Machinery. https://doi.org/10.1145/3632410.3632455

Keywords

Time Series, Clustering, Gamification