User Acceptance Model for Smart Incentives in Sustainable Video Streaming towards 6G

TL;DR

The paper addresses energy and carbon efficiency in next-generation video streaming by proposing a user-acceptance framework that links QoE to sustainability through a greenness factor $\gamma_n$ and a personalized incentive threshold $r_{\min,n}$, with stochastic acceptance via a sigmoid $p_n(r_n)$. It extends this core model with altruistic behavior, educator-driven threshold reductions, and data-driven learning of user traits, including two extension schemes (heterogeneous and two-group populations) and a learning-from-data pipeline using logistic regression. Through synthetic-data experiments, it demonstrates trade-offs between provider cost and network flexibility, showing that personalized, gradually adaptive incentives can reduce costs while meeting sustainability goals, and that education and learning amplify these gains. The work provides a scalable, data-informed blueprint for deploying energy-conscious streaming services in future 6G networks, with practical guidance on incentive design and policy levers for user segments.

Abstract

The rapid growth of 5G video streaming is intensifying energy consumption across access, core, and data-center networks, underscoring the critical need for energy and carbon-efficient solutions. While reducing streaming bitrates improves energy efficiency, its success hinges on user acceptance--particularly when lower bitrates may be perceived as reduced quality of experience (QoE). Therefore, there is a need to develop transparent, user-centric incentive models that balance sustainability with perceived value. We propose a user-acceptance model that combines diverse environmental awareness, personalized responsiveness to incentives, and varying levels of altruism into a unified probabilistic framework. The model incorporates dynamic, individualized incentives that adapt over time. We further enhance the framework by incorporating (i) social well-being as a motivator for altruistic choices, (ii) provider-driven education strategies that gradually adjust user acceptance thresholds, and (iii) data-driven learning of user traits from historical offer--response interactions. Extensive synthetic-data experiments reveal the trade-offs between provider cost and network flexibility, showing that personalized incentives and gradual behavioral adaptation can advance sustainability targets without compromising stakeholder requirements.

Figures (13)

• Figure 1: Flexibility-Cost ratio vs the discreptives of the offered incentives $\sim \mathcal{U}[a, b]$. $a = 18$, $b = 22$ (left). $a = 10$, $b = 40$ (right).
• Figure 2: Flexibility-Cost ratio vs the descriptives of the offered-incentives $\sim \mathcal{N}(\mu, \sigma^2)$. $\mu \in [0, 12]$ and $\sigma = 5$ (left). $\mu = 10$ and $\sigma \in [0.5, 5]$ (right).
• Figure 3: Flexibility-Cost ratio vs the numbers of incentivized users. The offered-incentives sampling from the uniform distribution, with $\mathcal{U}[18, 22]$ (left), and $\mathcal{U}[10, 40]$ (right).
• Figure 4: Flexibility-Cost ratio vs number of incentivized users. Offered-incentives are sampled from the normal distribution, with $\mathcal{N}(0.1, 0.05)$ (left), $\mathcal{N}(1, 1)$ (left), and $\mathcal{N}(10, 1)$ (right).
• Figure 5: Provider offers $\mathcal{U}[a,b]$: i) group with small $r_{\min, n}$ (blue), ii) group with large $r_{\min, n}$ (red), and iii) mean $r_{\min, n}$ between groups (green). (top-left) $k = 100$, (top-right) $k = 300$, (bottom-left) $k = 500$, and (bottom-right) $k = 800$.