[Tech Insights] Mapping Intimacy: Analyzing Relationships via Mobile Communication Logs

TL;DR

Researchers from Politehnica University of Timisoara have developed a method to quantify human relationships using nothing but your phone's call and text history. By applying a weighted mathematical model to communication frequency and duration, they can categorize contacts and visualize your "social solar system" in a intuitive star graph.

Problem & Motivation: The Gap Between Public Social Media and Private Reality

While social networks like Facebook or LinkedIn provide a "macro" view of our connections, they often fail to capture the nuances of our most intimate "micro" relationships. Someone you interact with daily via phone calls might have zero presence on your public social profile.

Prior work in this field often relied on surveys (which are subjective and labor-intensive) or complex multi-modal sensors (GPS, Bluetooth, App usage) which raise significant privacy concerns. The authors' insight was simple yet profound: the most accurate record of our social habits is already sitting in our Call Logs and SMS Databases.

Methodology: Quantifying the "Interaction Degree"

The core of the paper is the Interaction Degree (Gr) formula. Unlike simple frequency counters, this model recognizes that not all interactions are created equal.

The Weighted Formula

The authors define the relationship strength as: $Gr=0.2\cdot TNbC+0.4\cdot TNbSMS+0.4\cdot TCD$

• TNbC (Total Calls): Includes incoming, outgoing, and even missed calls (signals of intent).
• TNbSMS (Total Messages): Weighted higher (0.4) because text-based communication represents a deliberate exchange of information.
• TCD (Total Call Duration): Also weighted highly (0.4) as the time spent talking is a direct proxy for intimacy.

Table 1: Comparison of intimacy factors used in related research vs. the proposed model.

Experiments & Results: Categorizing the Social Circle

Over a three-month study period, the authors tracked interaction patterns and observed how seasonal events (like the Easter holidays) created "spikes" in communication with estranged contacts.

They identified four clear tiers of relationships:

1. Very Close: Constant, high-duration interactions (e.g., family/partners).
2. Close: Small but highly consistent interactions (at least once a month).
3. Friendly: Inconsistent, low-frequency contact.
4. Estranged: Interactions primarily triggered by holidays or specific events.

Figure 1: Visualizing call and SMS counts across the first month of the study.

The Star Graph Visualization

To make this data useful for a regular user, the team implemented a Weighted Star Graph.

• The Logic: Distance ($D$) from the center (User) is defined as $1/Gr$.
• The Visualization: A contact you talk to for hours appears right next to your central node; a contact you messaged once three months ago is pushed to the edge of the screen.

Figure 5: Fluctuations in Interaction Degree for different contacts over three months.

Critical Analysis & Conclusion

This work provides a robust framework for Social Interpersonal Analysis without needing invasive tracking. Its reliance on "Basic Services" (Call/SMS) makes it universally applicable to any Android device.

Limitations: The most obvious drawback is the exclusion of Instant Messaging (IM) apps like WhatsApp, Telegram, or WeChat. In the modern era, IM has largely supplanted SMS for many demographics. The authors acknowledge this and aim to integrate these APIs in future iterations.

Future Outlook: As we move toward a more privacy-conscious tech landscape, "local-first" analysis tools like this—which process sensitive metadata on-device without uploading it to a cloud—will likely become the standard for personal CRM and digital wellbeing applications.