Collusion Relations and their Applications to Balance Theory

Jean-Baptiste Joinet, Carlos Olarte · 2026

Read Collusion Relations and their Applications to Balance Theory by Jean-Baptiste Joinet, Carlos Olarte (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-01843.txt.

Introduction

Jean-Baptiste Joinet and Carlos Olarte study binary relations through an agonistic lens, where one agent attacking another is the basic link. They look at properties defined on groups of four elements and ask when a derived notion of protection stays consistent. Their work connects that question to balance in signed networks, where allies agree inside a group and oppose those outside it.

The authors show that balance can be captured by one such four-element property, called collusivity. That gives a new way to test whether a network splits into opposed camps. They extend the classic balance result to relations that need not be symmetric, and they add a modal logic and proof rules for collusive frames.

Catalog record

CREATORJean-Baptiste Joinet, Carlos Olarte
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE72 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.01843

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Collusion Relations and their Applications to Balance Theory Jean-Baptiste Joinet, Carlos Olarte Abstract We study quadrangular properties of binary relations on a set X X –i.e., properties defined on configurations of four elements– within an agonistic interpretation, where x ​ R ​ y xRy is interpreted as x x “attacks” y y . Such relations induce a suitable notion of “protection,” and we provide necessary and sufficient conditions for this notion to be consistent. We characterize the balance property in signed frames in terms of a specific quadrangular property, namely collusivity. In this way, we generalize a classical result in balance theory by offering an alternative method for determining whether a network is polarized. That is, one can identify well-formed groups of agents that agree with one another within the same group (a set of allies) while disagreeing with, or attacking, agents outside the group. Furthermore, we extend the balance theorem to non-symmetric relations, thereby relaxing a condition required in standard balance theory. We conclude by giving a modal characterization of collusive frames, together with corresponding rules in a labeled sequent calculus, and we show that previous modal characterizations of balance are derivable within this system. 1 Introduction Methods for reasoning about group polarization have recently attracted the attention within the proof theory community, since modal logics have proven to be useful for analyzing such phenomena. Consider for instance the Facebook epistemic logic [ 16 ] endowed with a symmetric “friendship” relation; Tweeting logic [ 19 ] (formalizing announcements in a network); the logic for reasoning about social beliefs [ 10 ] ; the positive-negative modal logic in [ 14 ] to reason about friendship and enmity relations, etc. We may also mention the game-semantic approaches that we have explored in this context in [ 4 , 15 ] . This paper focuses on the problem of balance in networks of agents, where relations may be either positive (e.g., friendship or agreement) or negative (e.g., enmity or disagreement). The theory of Balance originates in social psychology [ 12 , 7 ] and political philosophy [ 17 , 18 ] . It studies how such relations evolve toward configurations in which triads of agents forms “stable” patterns. For instance, if x is a friend of y and y is a friend of z , then the network tends to evolve toward a state in which x and z are also friends (capturing the intuition that the friend of my friend should be my friend). We contribute to this line of research in several ways, including a novel characterization of balance, relaxing a condition required in “standard” Balance theory, and the development of new proof methods. The methods proposed in this paper are grounded in the study of quadrangular relations ( Section 2 ), i.e., properties of a relation R\subseteq X\times X (for a set X ) expressible by quantifying over four elements of X . We focus in particular on collusive relations, introduced in [ 8 , 9 ] , defined by the quadrangular condition \forall x,y,z,w\in X,\big((xRy\wedge xRz\wedge wRy)\Rightarrow wRz\big) . Intuitively, whenever x and w share a target y , every target of x is also a target of w . Collusions extend equivalence relations, which are precisely the reflexive collusions. Moreover, the standard equi-quotient construction (partitions induced by equivalence relations) appears as a special case of a more general quotientation process induced by collusions. In fact, this process is fully general: collusions characterize quotients [ 8 , 9 ] . Moreover, the (collusional) quotients X/R and X/R^{-1} (if R is collusive then its inverse R^{-1} is collusive too) can be recovered as standard equi-quotients via two dual relational indiscernibility predicates (see Definition 9 ). In Section 3 , we endow a binary relation R with an agonal interpretation, where xRy is understood as x attacks y , x disagrees with y , or x is an enemy of y . We then provide necessary and sufficient conditions under which R induces a meaningful notion of protection against attacks. Intuitively, x protects z if x defends z against all its enemies. Here again a quadrangular property, namely co-confluence, plays a central role to ensure the consistency of the protection relation, in the sense that no agent can simultaneously protect and attack another. In Section 4 , we show a tight connection between collusive relations and Balance theory. The classical balance theorem provides three equivalent characterizations of stable networks ( Section 4.1 ). First, a network is stable if it can be “completed” by assigning signs to missing edges so that all triads are stable. Second, this holds iff the network can be partitioned into groups such that agents within the same group are not negatively related, while agents in different groups are not positively related. Third, this is equivalent to the condition that all cycles in the network satisfy certain constraints. In this paper we propose a fourth characterization ( Section 4.2 ): the previous conditions are equivalent to check whether the positive and negative relations are collusive. Our approach generalizes standard Balance theory in two complementary directions. First , classical Balance theory assumes two given relations (positive and negative) satisfying a fixed set of axioms. In contrast, we begin with a single relation, the agonal (attack/disagreement) relation, without imposing any a priori axiom. The positive relation is then derived from this attack relation, as shown in Section 4 . Subsequently, we identify the properties that the attack relation must satisfy in order to have a “natural” agonistic interpretation and how this interpretation leads to balanced configurations. Second , in Balance theory, both relations are assumed to be symmetric; additionally, the negative relation is anti-reflexive, while the positive relation is reflexive. In our case, anti-reflexivity of the negative relation is assumed since it has a natural agonistic interpretation. However, symmetry is not required ( Section 4.3 ). In fact, we show that irreflexive collusive relations (not necessarily symmetric) induce weakly balanced networks, while symmetric irreflexive collusive relations yield strongly balanced frames. These results provide further evidence of the close connection between collusive relations and stable configurations. As a final contribution, Section 5 provides a modal characterization of collusive relations. Following the approach of [ 13 , 11 ] , the proposed axiom gives rise to an inference rule in a labeled sequent calculus. We show that the resulting cut-free system can be used to establish key properties of balanced networks. In particular, we demonstrate that previous characterizations of balance in [ 14 ] can be derived within our system. This provides a new proof-theoretic tool for reasoning about balance in networks. 2 Quadrangular Relations This section explores some properties of relations that will be fundamental for the results in the forthcoming sections. Along this paper we will assume an arbitrary set X . Given a relation R\subseteq X\times X , we write xRy whenever (x,y)\in R . We say that x\in X is a source or initial element if its indegree is zero, i.e. \forall y\in X.\neg yRx . Typical properties of binary relations include: • reflexivity : \forall x\in X.xRx , and irreflexivity (or anti-reflexivity ): \forall x\in X.\neg xRx • seriality (or totality ): \forall x\in X.\exists

Read the complete text