Trust Propagation in Vee Structures

What is Trust?

Before modeling its propagation, it's essential to have a clear definition. Trust is a cornerstone of any relationship, but its meaning can be multifaceted. At its core, it involves a calculated risk.

As the video explains, trust is "putting oneself in a vulnerable position based on the belief that another person will act with our best interest in mind." This definition is powerful because it highlights the inherent risk involved. In a vee relationship, the hinge (H) trusts both partners (P1 and P2) with their time, emotions, and resources. Similarly, P1 and P2 trust H.

However, trust is rarely uniform or perfectly reciprocated. The video goes on to discuss that trust is often asymmetric and context-dependent. For instance, H may trust P1 more with emotional support and P2 more with practical matters. This asymmetry is fundamental to the vee, where the relationships H-P1 and H-P2 are distinct, and the P1-P2 link is often non-existent, at least initially. The question we want to explore is: how do the trust dynamics in the H-P1 and H-P2 dyads influence the potential relationship between P1 and P2?

The Logic of Balance: Structural Balance Theory

One of the oldest and most intuitive frameworks for understanding relationships in a signed network (where links can be positive or negative) is Structural Balance Theory (SBT). Developed by social psychologists, it posits that some network structures are more psychologically stable than others.

The theory is based on a simple premise that can be summarized by familiar adages:

• "A friend of my friend is my friend." (+, +, +)
• "An enemy of my friend is my enemy." (+, −, −)
• "An enemy of my enemy is my friend." (−, −, +)
• "A friend of my friend is my enemy." (+, +, −) --> This is considered unstable or "imbalanced."

This last case creates social tension. If you are friends with two people who dislike each other, you are in an awkward position. SBT predicts that this tension will drive the network to change, either by one relationship flipping sign (e.g., you drop one friend) or a new relationship forming to resolve the imbalance.

How does this apply to a vee? A simple vee is a path, P1-H-P2, not a closed triad, so there's no initial imbalance. The theory becomes predictive when we consider the potential for a relationship to form between P1 and P2.

• Scenario 1: H trusts P1 (+) and H trusts P2 (+). SBT suggests there is a "force" or "tension" for the P1-P2 link to become positive. A positive P1-P2 relationship would form a balanced (+,+,+) triad. This is the structural basis for compersion.
• Scenario 2: H trusts P1 (+) but distrusts P2 (-). Here, the theory predicts the P1-P2 link should be negative to form a balanced (+,−,−) triad. This provides a structural explanation for jealousy or conflict between the two partners.

A network is globally balanced if all its cycles have an even number of negative edges.

Beyond Binary: Extended Structural Balance Theory

Classical SBT is powerful but limited, as it treats trust as a simple positive/negative binary. Your background in statistics will appreciate that this is an oversimplification. Relationships have varying strengths—from strong trust to mild acquaintance, and from mild dislike to strong distrust.

Extended Structural Balance Theory (ESBT) refines the model by incorporating relationship strength. It does this by mapping relationships to a 'relation distance', where:

• Strongly positive relationships have a small distance.
• Neutral relationships have a medium distance.
• Strongly negative relationships have a large distance.

With this mapping, the rule for balance in a triad (A, B, C) becomes elegantly simple: the three relation distances must satisfy the triangle inequality.

And its permutations. This means that for a triad to be balanced, the distance between any two nodes must be less than or equal to the sum of the other two distances. A triad is imbalanced if this geometric constraint is violated.

This leads to a powerful predictive tool. For the vee (P1-H-P2), we can state the balanced range for the P1-P2 relationship:

This inequality defines the "tolerance" for the P1-P2 relationship. If the actual d(P1,P2) falls outside this range, the triad is imbalanced and under pressure to change.

Let's reconsider our vee scenarios with this quantitative lens:

• Scenario 1 (Compersion): H has strong trust in P1 (e.g., d(H,P1)=0.2) and strong trust in P2 (e.g., d(H,P2)=0.3). The tolerance for the P1-P2 distance is [|0.2-0.3|, 0.2+0.3] = [0.1, 0.5]. This narrow range on the low end of the distance spectrum predicts a positive relationship between P1 and P2.
• Scenario 2 (Jealousy): H has strong trust in P1 (d(H,P1)=0.2) but strong distrust in P2 (d(H,P2)=1.2). The tolerance is [|0.2-1.2|, 0.2+1.2] = [1.0, 1.4]. This range, high on the distance spectrum, predicts a negative relationship between P1 and P2.

ESBT provides a much more nuanced prediction than classical SBT, accounting for the strength of the existing relationships.

An Algorithmic View: Atomic Propagation

While SBT describes states of equilibrium, a different class of models focuses on the process of inferring trust. This approach, which will resonate with your background in operations research and algorithms, views trust propagation as a series of computational steps on the network graph.

The paper by Guha et al. proposes several "atomic" mechanisms by which trust can propagate.

These mechanisms are:

1. Direct Propagation: If A trusts B, and B trusts C, we can infer that A has some trust in C. This is the most intuitive form of transitivity.
2. Transpose Trust: If A trusts B, we might infer that B has some trust back in A. This assumes a degree of reciprocity.
3. Cocitation: If A trusts C, and B trusts C, we can infer that A and B might trust each other, as they share a common positive judgment.

For the vee, cocitation is the most relevant mechanism for the P1-P2 dynamic. Since P1 and P2 both have a relationship with the hinge H, we can use that to infer a relationship between them. If P1 trusts H and P2 trusts H, cocitation suggests P1 and P2 might trust each other.

This algorithmic perspective also forces us to confront the complexity of distrust. If A distrusts B, and B distrusts C, what is A's opinion of C?

• Multiplicative View ("The enemy of my enemy is my friend"): Distrust can be seen as a negative value. A chain of two distrusts (-1 * -1) results in trust (+1).
• Additive View ("My enemy's judgment is flawed"): You might believe that someone you distrust has poor judgment. Therefore, you would also distrust anyone they distrust.

The choice between these models for distrust is not a mathematical one, but a psychological one that depends on the context of the relationship. In a vee, if P1 distrusts H, and H distrusts P2, P1's feeling towards P2 could go either way. P1 might see P2 as a potential ally against a problematic hinge (multiplicative), or P1 might see P2 as "tarred by the same brush" as H (additive). These models give us the language to describe these different dynamics formally.

Conclusion

In this lesson, we explored how trust and distrust, foundational elements of any relationship, propagate through a network structure. We saw that this process, while complex, can be analyzed using formal models that provide powerful insights into relationship dynamics.

Key Takeaways:

• Structural Balance Theory (SBT) provides a qualitative model for network stability. It predicts that triads with an odd number of negative links are unstable, creating pressure for change. In a vee, this explains the structural basis for compersion and jealousy.
• Extended Structural Balance Theory (ESBT) offers a more nuanced, quantitative model by representing relationship strengths as 'relation distances'. Balance is achieved when the triangle inequality holds, providing a predictive range for the nature of a missing link (e.g., between P1 and P2).
• Atomic Propagation Models provide an algorithmic perspective, defining mechanisms like direct propagation and cocitation to infer trust. For a vee, cocitation is a key mechanism for modeling how partners P1 and P2 can form opinions of each other through their shared connection to the hinge H.
• The propagation of distrust is ambiguous, with competing logical models ("the enemy of my enemy is my friend" vs. "my enemy has poor judgment").

In our previous lesson, we saw how emotional states diffuse. Today, we've seen how more stable judgments like trust and distrust propagate. In our next and final lesson of this module, we will examine how the failure of one part of the network—such as the dissolution of a single relationship—can trigger a cascading failure, potentially leading to the collapse of the wider structure.