The Next Frontier of Pervasive Computing: Social Context Modelling and Recognition

TL;DR

In this keynote summary, Professor Jiannong Cao explores the evolution of context-aware systems from tracking individuals to understanding Social Context. By leveraging the digital traces left on smartphones and social networks, researchers can now infer social ties and group behaviors, though significant challenges remain in handling the heterogeneity of "Cyber-Physical-Social" data.

Moving from "Me" to "Us"

Historically, context-aware computing was a solitary affair—your phone adjusting brightness based on light or suggesting a route based on your location. However, humans are inherently social animals. The real value for business intelligence and public safety lies in the Social Context:

• Who are the influencers in this group?
• Are these individuals strangers or a cohesive social unit?
• How does an epidemic or information spread through these specific social ties?

The core challenge is that social context is not a single data point; it is an emergent property of multiple users interacting across complex, heterogeneous platforms.

The Methodology: Harvesting Digital Traces

Professor Cao’s approach centers on the convergence of three major data sources:

1. Sensor-rich Smartphones: Capturing Proximity (Bluetooth), Mobility (GPS), and Physical Activity.
2. Ambient Sensor Networks: Smart city infrastructure providing the environment's "state."
3. Social Network Services (SNS): Providing the "Cyber" layer of explicit relationships and interests.

Figure 1: Conceptual visualization of Social Context Recognition (Generic Representation).

The methodology involves inferring high-level "Social Ties" from low-level "Digital Traces." This requires mapping physical proximity data to social relationship strengths—a task that is non-trivial due to the "noise" of modern urban environments.

Complexity and Heterogeneity

The paper identifies two primary hurdles:

• Complexity: Unlike individual behavior, group behavior is non-linear and dynamic.
• Heterogeneity: Data from a heart rate monitor (physical) and a Twitter post (cyber) have different temporalities, formats, and semantic weights.

Insights and Future Directions

Professor Cao suggests that the field is moving toward a more holistic integration of hardware and social science.

Key Takeaways:

• Collective Intelligence: We must move beyond recognizing what a user is doing to who they are with and why that group exists.
• Cross-Platform Fusion: The future lies in algorithms that can bridge the gap between physical sensor data and cyber-social data seamlessly.

Conclusion: Why This Matters

Social Context Modelling is more than an academic exercise; it is the backbone of the next generation of "Smart" applications. Whether it's a company refining its advertisement strategy by targeting social hubs or a government managing public health during an epidemic, understanding the social fabric through digital traces is no longer optional—it’s the new standard for pervasive intelligence.

About the Author: Dr. Jiannong Cao is a Fellow of the IEEE and a Chair Professor at Hong Kong Polytechnic University, with extensive contributions to distributed computing and wireless networks.