[Paper Review] Intelligent Notification Systems: A Survey of the State of the Art and Research Challenges

Abhinav Mehrotra, Mirco Musolesi|arXiv (Cornell University)|Nov 28, 2017

Personal Information Management and User Behavior92 references18 citations

TL;DR

This survey presents a comprehensive analysis of intelligent notification systems that leverage user context and behavioral modeling to optimize notification timing and reduce disruption. By integrating sensor-based context awareness and machine learning to predict user interruptibility, the system improves receptivity and minimizes adverse effects on task performance and user affect.

ABSTRACT

Notifications provide a unique mechanism for increasing the effectiveness of real-time information delivery systems. However, notifications that demand users' attention at inopportune moments are more likely to have adverse effects and might become a cause of potential disruption rather than proving beneficial to users. In order to address these challenges a variety of intelligent notification mechanisms based on monitoring and learning users' behavior have been proposed. The goal of such mechanisms is maximizing users' receptivity to the delivered information by automatically inferring the right time and the right context for sending a certain type of information. This article provides an overview of the current state of the art in the area of intelligent notification mechanisms that relies on the awareness of users' context and preferences. More specifically, we first present a survey of studies focusing on understanding and modeling users' interruptibility and receptivity to notifications from desktops and mobile devices. Then, we discuss the existing challenges and opportunities in developing mechanisms for intelligent notification systems in a variety of application scenarios.

Motivation & Objective

• Address the growing problem of disruptive mobile notifications that impair task performance and user well-being.
• Identify the limitations of current notification systems in adapting to users' real-time context and interruptibility.
• Survey state-of-the-art approaches that model user receptivity using sensor data and behavioral patterns.
• Highlight the need for large-scale, in-the-wild evaluations to ensure ecological validity and generalizability.
• Outline open challenges in designing adaptive, user-centered notification systems across diverse devices and contexts.

Proposed method

• Systematically review studies on user interruptibility and receptivity using data from mobile and desktop devices.
• Analyze context-aware notification mechanisms that use physical and cognitive context (e.g., activity, location, device usage) to infer optimal delivery times.
• Categorize notification types based on source (human or machine-generated) and relevance to user context.
• Evaluate interruptibility prediction models using offline datasets collected from real users in natural settings.
• Integrate findings from large-scale studies (e.g., Sahami et al., 2014; Mehrotra et al., 2015) to assess real-world user behavior.
• Propose the need for adaptive systems that learn from user interactions and adjust delivery strategies dynamically.

Experimental results

Research questions

• RQ1 What contextual factors significantly influence users' receptivity to notifications in mobile and desktop environments?
• RQ2 How do current intelligent notification systems model and predict user interruptibility using sensor-based context?
• RQ3 What are the key limitations of existing interruptibility prediction models, particularly in terms of ecological validity and scalability?
• RQ4 How do notification types (e.g., social, promotional, system alerts) affect user acceptance and dismissal behavior?
• RQ5 What are the open challenges in deploying intelligent notification systems at scale across diverse user demographics and social contexts?

Key findings

• Notifications delivered at inopportune times significantly disrupt ongoing tasks and negatively affect users' affective states.
• Users frequently dismiss irrelevant or uninteresting notifications (e.g., app updates, game invites), indicating low receptivity to non-urgent alerts.
• Current interruptibility models are predominantly evaluated offline, limiting their real-world validity and robustness.
• There is a strong need for large-scale, in-the-wild deployments to validate prediction models across diverse populations and environments.
• Existing systems often fail to account for cognitive and physical context, leading to suboptimal notification timing and user annoyance.
• The integration of contextual inference with adaptive learning mechanisms can significantly improve notification effectiveness and user experience.