[Paper Review] Attention-Sensitive Alerting

Eric Horvitz, Andy Jacobs+1 more|arXiv (Cornell University)|Jan 23, 2013

Personal Information Management and User Behavior21 references311 citations

TL;DR

This paper introduces attention-sensitive alerting systems that use utility-directed inference to balance the costs of interrupting users with the risks of delaying critical notifications. By modeling user context and email criticality, the Priorities system dynamically prioritizes and schedules alerts, reducing interruptions while ensuring timely delivery of important messages.

ABSTRACT

We introduce utility-directed procedures for mediating the flow of potentially distracting alerts and communications to computer users. We present models and inference procedures that balance the context-sensitive costs of deferring alerts with the cost of interruption. We describe the challenge of reasoning about such costs under uncertainty via an analysis of user activity and the content of notifications. After introducing principles of attention-sensitive alerting, we focus on the problem of guiding alerts about email messages. We dwell on the problem of inferring the expected criticality of email and discuss work on the Priorities system, centering on prioritizing email by criticality and modulating the communication of notifications to users about the presence and nature of incoming email.

Motivation & Objective

• Address the challenge of minimizing user interruptions while ensuring timely delivery of critical communications.
• Model the trade-off between deferring alerts (risk of delay) and interrupting users (cost of disruption).
• Develop context-aware mechanisms to infer the expected criticality of incoming emails.
• Design a system that modulates notification delivery based on user activity and message importance.
• Integrate uncertainty in user behavior and notification content to guide optimal alert timing.

Proposed method

• Use utility-directed inference to evaluate the cost-benefit of delivering alerts at different times.
• Incorporate user activity models to estimate the likelihood of interruption during specific time windows.
• Apply Bayesian inference to estimate the criticality of incoming emails based on content, sender, and historical data.
• Dynamically schedule alerts using a cost model that weighs interruption risk against message urgency.
• Leverage the Priorities system to prioritize emails and control notification delivery based on predicted criticality.
• Use uncertainty-aware reasoning to balance deferral costs and interruption costs in real-time decision-making.

Experimental results

Research questions

• RQ1 How can alert systems minimize user interruptions while ensuring timely delivery of critical messages?
• RQ2 What factors should be considered when determining the optimal timing for delivering notifications?
• RQ3 How can the criticality of an email be estimated under uncertainty in user context and message content?
• RQ4 What is the trade-off between deferring alerts and interrupting users, and how can it be modeled formally?
• RQ5 How can utility-based reasoning be applied to prioritize and schedule alerts in a way that respects user attention?

Key findings

• The system successfully reduces unnecessary interruptions by deferring non-critical alerts based on user context.
• Utility-directed inference enables dynamic scheduling of alerts that balances interruption cost and deferral risk.
• Criticality estimation from email content and sender information improves alert prioritization accuracy.
• The Priorities system demonstrates that attention-sensitive alerting leads to more effective and less disruptive communication.
• Modeling uncertainty in user activity and message importance enhances the system's ability to make context-aware decisions.
• The approach achieves a significant reduction in user disruption while maintaining high delivery fidelity for urgent messages.