Theories of Valenced Interactions

Karma, Status, Solidarity, and Dominance in Sentiment Networks

Part 1: Valenced Interactions & Karma Theory

Introduction to Valenced Interactions

Beyond Simple Co-membership

  • In social network analysis, we move beyond simple presence or absence of links to examine sentiment networks.
  • Relationships have an emotional and evaluative character, labeled as valence:
    • Positive valence (e.g., liking, helping, admiring).
    • Negative valence (e.g., disliking, bullying, fighting).
  • Valenced Interactions are governed by deep-seated psychological micro-motives.
  • These principles drive how networks grow, change, and cohere into stable macrostructures.

Karma Theory: Core Concept

What You Send Out Predicts What You Get Back

  • Core Principle: “You get what you spread out!”
  • The valence of the ties you send out predicts the valence of the ties you receive back — from the network at large, not necessarily from the same alter.
  • Predicts received valenced indegree (what you receive) based on expressed valenced outdegree (what you send out).
  • Illustrates common expressions:
    • “Likers get liked” (positive).
    • “Bullies get bullied” (negative).

Karma Theory: Network-Wide Flow

  • Karma Flows can be generalized across broader spheres in the network.
  • Positive Sphere: Node A (positive spreader) actively initiates friendly ties and receives deference/support in return from the network.
  • Negative Sphere: Node B (negative spreader) actively initiates hostile ties and attracts reciprocal conflict/bullying from multiple alters.
  • No alter returns a tie directly; the match is between what each sends and what it receives from others.

Karma Dynamics: Virtuous vs. Vicious Cycles

  • Karma principles drive valenced networks to grow and evolve via feedback loops:
  • Virtuous Cycles: An initial positive act (\(+\)) breeds matching positive sentiment from others (\(+\)) — not necessarily from the same alter. These positive feedback loops lock actors into mutually reinforcing, cooperative cliques.
  • Vicious Cycles: A single unprovoked negative act (\(-\)) breeds hostile counter-attacks (\(-\)). This triggers cascading negative spirals, spreading localized conflict throughout the network.
  • Over time, outdegree and received indegree valences converge on a node-by-node basis.

Part 2: Status Theory

Status Theory: Introduction

  • Status Theory proposes that sentiments towards an alter are heavily driven by how others feel about them.
  • It extends the triadic balance logic from the previous lecture.
  • This represents a social prestige alignment mechanism:
    • If someone is liked by many, you align your sentiment and like them too.
    • If someone is disliked by many, you align your sentiment and dislike them.
  • A node’s status is determined by the valence signs of incoming ties:
    • High status = receiving many positive links.
    • Low status = receiving many negative links.

Prestige and Deference Alignment

  • Inside status-seeking systems, tie directions carry implicit status expectations:
    • Directed positive tie from Ego to Alter \(\implies\) Ego considers Alter to have higher status (Deference).
    • Directed negative tie from Ego to Alter \(\implies\) Ego considers Alter to have lower status (Dominance).
  • Status-seeking actors strive to align their sentiments with prestige to avoid losing their own social standing.

Status and Outgoing Sentiments

  • Your outgoing ties affect your own social position:
    • Status Gain:
      • Sending positive ties to high-status actors (basking in reflected glory).
      • Sending negative ties to low-status actors (distancing from pariahs).
    • Status Loss:
      • Sending positive ties to low-status actors (guilt by association).
      • Sending negative ties to high-status actors (hostility to elites backfires).
  • Actors actively prune or withdraw ties that cause status loss.
Outgoing tie Target status Sender's status
Positive (+) High Gain (+)
Negative (−) Low Gain (+)
Positive (+) Low Loss (−)
Negative (−) High Loss (−)

Status Dynamics: Centralization and Hierarchy Growth

  • How status networks grow and change over time:
    • Preferential Deference: High-status nodes attract a disproportionate share of incoming positive ties over time (a “rich-get-richer” effect).
    • Dynamic Pruning: High-status nodes lose their negative ties, while low-status nodes lose their positive ties.
    • “Gang-up” Effect: Subordinates consolidate negative ties onto low-status pariahs to boost their own standings.
    • Hierarchy Emergence: Sentiments organize into highly centralized, strictly ordered prestige rankings.

Part 3: Solidarity Theory

Solidarity Theory: Core Concept

  • Central idea: “People in the same position in the valenced interaction network are likely to link via positive ties.”
  • Time 1: Ego and Other already occupy the same position — each sends (+) and receives (+) ties from distinct alters (the popular “in-crowd”).
  • Time 2: The shared position generates a new positive tie between them — status solidarity: “Cool kids stick together!”

Solidarity on the Negative Side: The Downtrodden

  • The same logic holds at the bottom of the status order: actors who send and receive negative ties are similarly situated (karma solidarity).
  • Shared unpopularity motivates positive bonding between them — “misery loves company.”
  • Note: the solidarity tie is itself positive, even though the shared position is built entirely out of negative exchange.

Karma Solidarity: Bullies Stick Together

  • Solidarity can also form around the negative-sender position: actors who spread hostility occupy the same spot in the network.
  • Bullies direct (−) ties at their targets — and form a positive bond with each other.
  • Same-position actors close ranks; their targets (grey) receive the hostility but are outside the solidarity tie.

Measuring Shared Position

  • Structural similarity acts as a strong motivator for positive bonding:
    • Status Solidarity: Elite, popular actors share a similar high status, motivating them to like and support one another (“popular kids stick together”).
    • Unpopularity Solidarity: Disliked, downtrodden actors share a similar lower-tier position, motivating them to bond (“downtrodden stick together”).
    • Vertical Cleavages: Since different positions motivate negative sentiments, positive ties spanning across tiers are actively pruned.
Actor A Actor B Position match Predicted sentiment
High status High status Similar Solidarity (+)
Low status Low status Similar Solidarity (+)
High status Low status Different Hostility (−)

Solidarity Dynamics: Stratification and Clique Growth

  • How Solidarity networks grow and stratify over time:
    • Horizontal Integration: Sharing the same status position drives the creation of reciprocal positive ties within the same tier (clique formation).
    • Class Cohesion: Solidarity processes stabilize clusters into horizontal social classes (e.g., unions, associations, exclusive clubs).
    • Vertical Hostility: Cross-tier bonding is severed. What remains across tiers is asymmetric — deference flowing up to the elite, dominance flowing down.
    • Clique Stratification: The network evolves into a horizontally integrated, vertically polarized system.

Part 4: Dominance Hierarchies

Dominance Orders in the Animal Kingdom

  • Dominance hierarchies are observed across many species — most notably among chickens. No big brain needed to build a complex social network!
  • A “pure” dominance order has three algebraic properties:
    • Directed: a tie means “i dominates j.”
    • Antisymmetric: if A dominates B, B cannot dominate A.
    • Complete: for every pair, someone dominates — either A dominates B or B dominates A.

One tie per pair (antisymmetric, complete) — but arranged as a cycle, which a pure dominance order cannot contain.

Transitivity in Dominance Orders

  • Transitivity: dominance passes through nodes.
  • If A dominates B, and B dominates C, then A also dominates C — just like \(A > B\) and \(B > C\) imply \(A > C\).
  • In a directed network: if A is above B, A is above everyone B is above.
  • A linear dominance order must be transitive — the A → C arc has to exist.
  • An intransitive triad has the full chain A → B → C but the A → C tie reversed (C → A) — the closure is contradicted, not merely absent.
  • A merely absent A → C tie is a different case, the pass-along triad, which we return to below.

Vacuous Transitivity

  • What if A → B, but B does not direct a tie to C?
  • The precondition for transitivity — A → B and B → C — is not met.
  • The triad cannot violate transitivity, because the necessary chain does not exist.
  • This configuration is vacuously transitive:
    • It does not actively demonstrate transitivity…
    • …but it does not violate it either.
  • Only triads with a full chain A → B → C can be transitive or intransitive.

Perfectly Linear Hierarchies: The Pecking Order

  • Chickens placed in a new environment scuffle until a clear pecking order emerges.
  • The winner asserts dominance; this continues until every chicken knows its place.
  • Alpha dominates Beta and everyone below; Beta dominates Gamma and everyone below, and so on.
  • The hierarchy is ritually reinforced: the higher-status chicken pecks, the lower-status one assents.
  • A classic perfectly transitive dominance order.

Human Hierarchies: Deviations from Perfect Transitivity

  • Unlike chickens, human hierarchies are typically not perfectly transitive — and they vary by context.
  • A teacher has final authority in the classroom, but those relations may not apply outside it.
  • Deference to a President can invert during an election — the position depends on the voters’ will.
  • Two people can each lead in different clubs; their relative standing shifts with the decision at hand.
  • Humans tacitly recognize orderings, but disagreements over what determines status arise, and hierarchies are context-dependent.
  • Human hierarchies still tend toward the linear ideal:
    • Deviations are often temporary or local.
    • Ambiguity about the order itself breeds conflict and competition.
  • Still, human networks can be analyzed by comparing them to these idealized structures.

Seven Triad Types in Dominance Networks

Triad Structure Meaning
Null (003) No ties No competition yet
Single edge (012) A → B The starting point for iterative competition
Double-dominant (021D) A → B, A → C Signature of winner effects
Double-loser (021U) A → C, B → C Signature of loser effects
Pass-along (021C) A → B → C, no A–C tie Transitivity not yet resolved
Transitive (030T) A → B, B → C, A → C The expected triad under a linear order
Cycle (030C) A → B, B → C, C → A The violation — no clear hierarchy

The Seven Triads at a Glance

  • A dominance network can contain seven possible triadic configurations built entirely of asymmetric or null ties.
  • For a linear order to exist, it must be transitive and contain no cyclical triads.
  • Everything except the cycle is compatible with a linear order; the pass-along triad simply has not yet resolved its A–C tie.

Models of Hierarchy Formation: The Attribute-Based Model

  • Attribute-based model: position in the hierarchy reflects inherent qualities and abilities — talent, athletic ability, humor, being a good friend all count as “quality” — with the “biggest, fastest, strongest” rising to the top.
  • Deference does the compensating (Gould 2002):
    • A less talented person “exchanges” deference for association with a more talented one.
    • Yielding signals the recipient’s quality and status, increasing their influence.
  • If required deference exceeds the acceptable limits of reciprocity, the tie may not form at all.
  • Implication: a relatively stable hierarchy mirroring underlying capabilities.
  • But is rank really just a readout of attributes?
  • Enter the interaction-based model…

The Interaction-Based Model: Agonistic Competition

  • Interaction-based model (Chase 1982): hierarchies emerge from the sequence and outcomes of past agonistic encounters — not only from attributes.
  • Winner effects: winning brings confidence, experience, testosterone — winners dominate again (double-dominant triads).
  • Loser effects: losing brings lowered confidence and injury — losers lose again (double-loser triads).
  • Bystander effects: observers update from watching:
    • A bystander defers to the winner and rarely dominates them.
    • But a bystander becomes stronger relative to the loser, and rarely gets dominated by them.

Bystander Effects: A Worked Example

  • Three chickens: A is strong, B is moderately strong, C is weak.
  • A fights and defeats B — which suffers injury and lost confidence.
  • C only watched. Yet it can now defeat B, something it could not do before.
  • C’s new tie is a bystander update. It never fought A, and it never beat B before.
  • The final hierarchy is not a readout of abilities alone; it is shaped by the order in which contests occur.

Why Iterative Competition Guarantees Linearity

  • Combine the three effects over repeated contests:
    • Winners keep winning; losers keep losing (double-dominant and double-loser triads).
    • Bystanders almost never dominate the initial dominant — and almost never get dominated by the initial loser.
  • Bystander updating therefore locks in the outcome of every observed fight rather than reversing it.
  • Step by step, every triad that could be cyclical or intransitive gets resolved toward transitivity.
  • The resulting hierarchy is not a pure readout of attributes: it is shaped by the history and sequence of interactions.
  • Iterative, social competition pushes the dominance order strongly toward transitivity and linearity over time — the model’s prediction, even if real human hierarchies only approximate it.

Part 5: Comparative Summary & Synthesis

Comparison of Karma, Status, Solidarity, and Dominance

  • Each theory posits a different micro-motive, and each implies its own local pattern and global structure.
Theory Micro-mechanism Local pattern Global structure
Karma Match what you send with what you receive Virtuous / vicious loops Positive / negative regions
Status Align with others' sentiments (prestige) Deference / dominance Centralized hierarchies
Solidarity Align with actors in the same position Horizontal mutual ties Cohesive horizontal classes
Dominance Win, and let others watch Transitive triads Linear dominance order

Co-Evolution of Valenced Networks: A Combined Model

  • In real-world sentiment networks, these three principles do not act in isolation; they interact and co-evolve:
    1. Phase 1: Raw Interaction: Initial encounters are governed by Karma (direct reciprocity loops).
    2. Phase 2: Hierarchy Emergence: As actors seek status and prestige, Status seeking creates centralized deference paths.
    3. Phase 3: Factional Stratification: Sharing the same status position triggers Solidarity, freezing actors into cohesive horizontal classes.
  • Dominance (Part 4) adds a fourth dynamic, agonistic competition, on top of these — an illustrative synthesis, not a chapter claim.