Your Messages Aren’t Just Messages — They’re Behavioral Data
Series: Seeing the Social Program (2/4)
The Message Was Never Just the Text
You send a text.
It feels direct. Personal. Contained.
A message to one person, in one moment.
But on most platforms, the message itself is only part of what is generated.
Alongside it is a layer of metadata—information about the interaction rather than its content. This can include:
- Who you contacted
- When the interaction occurred
- How frequently communication happens
- Response timing patterns
Individually, these signals may seem insignificant. Aggregated over time, they can reveal structured patterns of behavior.
The Pattern Behind the Conversation
In 2018, reporting by multiple outlets, including Ars Technica and The Guardian, documented that Facebook had collected call logs and SMS metadata from some Android users through permissions granted within its applications.¹
The data collected in these cases included:
- Call frequency
- Duration
- Historical contact records
Importantly, this reporting indicated that the content of communications was not collected in the same manner. However, researchers have shown that metadata alone can be highly informative.
A widely cited study by Stanford University demonstrated that telephone metadata—such as call timing and frequency—can be used to infer sensitive personal attributes and relationship structures with a high degree of accuracy.²
These findings support a broader conclusion:
Even without message content, communication patterns can be used to approximate social networks and relationship strength.
When Platforms Model Interaction Patterns
Once interaction data is aggregated, it can be used to model behavioral patterns.
In 2021, internal documents reported by The Wall Street Journal—commonly referred to as the “Facebook Files”—indicated that platform systems were designed to optimize user engagement, including through the ranking and amplification of content that generated stronger reactions.³
These systems rely on behavioral signals such as:
- Click-through rates
- Time spent on content
- Interaction frequency
While these signals do not directly measure emotional state, they are often used as proxies for user response and engagement intensity.
Influence Through System Design
The implications of these systems are not always immediately visible.
Users typically experience platforms as neutral environments. However, the structure of those environments is shaped by:
- Ranking algorithms
- Notification systems
- Content prioritization mechanisms
These systems can influence:
- Which interactions are surfaced
- Which contacts remain visible
- Which conversations continue or diminish
This influence is generally indirect.
It operates through selection and timing, rather than explicit instruction.
The Core Tension
This creates a measurable tension:
Connection vs. behavioral influence
Users make choices about who to engage with. At the same time, the system:
- Filters available information
- Prioritizes certain interactions
- Reinforces specific behavioral patterns
Over time, this can shape:
- Communication frequency
- Relationship maintenance
- Attention allocation
Why This Matters
Relationships are not only social—they are functional.
They inform:
- Trust
- Decision-making
- Emotional support structures
If interaction patterns can be modeled, they can also be:
- Analyzed
- Predicted
- Influenced through system design
Not through direct control, but through environmental shaping.
Companion Intelligence Perspective
Companion Intelligence is designed around a different premise:
That relationship data should remain under user control.
In this model:
- Communication history is stored locally
- Relationship timelines are private
- Context is not used for engagement optimization
Instead of external modeling, the system supports user-directed recall and awareness.
For example:
- “When did I last speak with this person?”
- “What commitments were made?”
- “Which relationships have become inactive?”
These queries operate on user-owned data, rather than platform-aggregated behavioral models.
Closing
If interaction patterns can be mapped, they can also be interpreted.
And if they can be interpreted at scale, they can influence what remains visible in your world.
The structure of communication is not neutral.
Citations
- Ars Technica, “Facebook collected call, SMS data for years from Android phones” (2018); also reported by The Guardian (2018).
- Mayer, J., & Mutchler, P., “MetaPhone: The Sensitivity of Telephone Metadata”, Stanford University (2014).
- The Wall Street Journal, The Facebook Files series (2021), including reporting on engagement-based ranking systems.
