Dyadic Manipulation Model Explained

• The Dyadic Manipulation Model is a formal framework that defines dyadic interactions using a finite action set, including a null option, resulting in six distinct relational types.
• It establishes a rigorous state space and taxonomy, mapping elementary interactions to classic social coordination concepts such as Equality Matching, Market Pricing, Authority Ranking, and Communal Sharing.
• The model is applied to empirical social data through algorithmic inference techniques, enabling quantitative analysis of dyadic relations and supporting performance metrics like precision and recall.

A Dyadic Manipulation Model is a formal framework for representing and classifying the possible stable patterns of social coordination and exchange between two agents, each of whom can choose to engage in one of multiple well-defined actions (including “no action”) toward the other. This abstraction, originally formulated by Favre & Sornette, situates itself at the intersection of mathematical sociology, discrete relational modeling, and theoretical social psychology, notably providing a principled justification of Relational Models Theory’s four celebrated relational types, as well as their asocial and null limiting cases (Favre et al., 2013).

1. Formal Architecture: State Space and Elementary Interactions

Let $S = \{s_1, \ldots, s_N\}$ be a finite set of $N$ possible social actions. Each agent in the dyad (labelled $A$ and $B$) can act by selecting any $a, b \in S \cup \{\varnothing\}$, where $\varnothing$ is a distinguished “null” action (do-nothing). The full state space for a single interaction is thus $$E_N = (S \cup \{\varnothing\}) \times (S \cup \{\varnothing\})$$, of cardinality $(N+1)^2$.

An elementary dyadic interaction is specified by a pair $(a, b) \in E_N$, interpreted as “$A$ does $a$ to $B$ and $B$ does $b$ to $A$.” Empirical data over time yields a multiset $R \subseteq E_N$ corresponding to the observed interactions between $A$ and $B$, possibly with multiplicities.

A (simple) relationship is $|R| = 1$, just a single observed $(a, b)$, while composite relationships are unions of such events across time, i.e., arbitrary subsets of $E_N$.

2. Taxonomy of Relationship Types: Six Exhaustive Atomic Categories

The Dyadic Manipulation Model identifies exactly six mutually disjoint, exhaustive “atomic” relationship types, classified according to logical properties of the paired actions:

Category	Canonical Elements	Relational Models Theory Mapping
Equality Matching (EM)	$R = \{(x, x)\}$, $x \in S$	Equality Matching
Null Interaction	$R = \{(\varnothing, \varnothing)\}$	Null
Market Pricing (MP)	$R = \{(x, y), (y, x)\}$, $x \neq y \in S$	Market Pricing
Authority Ranking (AR)	$R = \{(x, y)\}$, $x \neq y$	Authority Ranking
Communal Sharing (CS)	$R = \{(x, \varnothing), (\varnothing, x)\}$, $x \in S$	Communal Sharing
Asocial (Unilateral)	$R = \{(x, \varnothing)\}$ or $\{(\varnothing, x)\}$, $x \in S$	Asocial/Unilateral

Each category is specified by conditions on $R$. For example, “EM” corresponds to both agents performing the identical non-null action; “MP” to reciprocal, role-exchange actions with $x \neq y$; “AR” to unilateral directed cross-flux with no reciprocation. Categories are provably exhaustive and mutually exclusive for $|R| \leq 2$ (Favre et al., 2013).

3. Theoretical Justification: Mapping to Relational Models Theory (RMT)

The four coordinated categories (EM, MP, AR, CS) provide an exact formal mapping to the four fundamental RMT relational models postulated by Fiske: Equality Matching, Market Pricing, Authority Ranking, and Communal Sharing. The asocial and null categories are limiting cases—either the absence of coordinated behavior or strict non-coordination (Favre et al., 2013).

The model demonstrates that these six forms—arising from the logic of pairwise action, in/exclusion of $\varnothing$, and symmetry/reciprocity constraints—are the only atomic possibilities for dyadic coordination (when each party either matches, differs, or abstains from action). This strict exhaustiveness implies that RMT’s four coordinated models are complete at the level of elementary dyadic structures.

4. Generalization to Arbitrary Action Sets and Composite Relationships

For any $N$ (number of actions), the atomic classification persists: despite the exponential growth of $E_N$ (size $(N+1)^2$), only six shape-types arise at the atomic level. Composite relationships (sequences or unions of elementary interactions) can be represented as vectors giving the counts of each atomic pattern observed (Favre et al., 2013).

A classification function $C: 2^{E_N} \to \{1, \ldots, 6\}$ assigns each $R$ (set or multiset of interactions) to its atomic category (or categories, for composites). This approach affords modularity and practicality for empirical dyadic data, allowing for the extension to, e.g., digital interaction networks, negotiation transcripts, or controlled laboratory dyads.

5. Application to Dyadic Social Data and Algorithmic Inference

To operationalize the Dyadic Manipulation Model on empirical data:

• All observed $(a, b)$ action pairs are aggregated into $R \subseteq E_N$, possibly with frequencies.
• For each action $x \in S$, the following tests are applied (in order): search for MP (reciprocal exchange, $x \neq y$), AR (unilateral directed, non-reciprocated cross-flux), EM (identical non-null actions), CS (bilateral but not alternating), Asocial (unilateral, no reverse), Null (neither acts).
• These criteria directly yield inferred relational category assignments, which can be analyzed as predictors, compared to ground truth human coding, or tracked over time.
• Metrics such as precision, recall, cross-validation, or clustering purity are used for performance analysis.

For example, market pricing is indicated if both $(x, y)$ and $(y, x)$ are present in approximately balanced numbers, ideally alternating in time; authority ranking by persistent asymmetrical roles (Favre et al., 2013).

6. Mathematical Proof of Exhaustiveness and Uniqueness

Favre & Sornette’s key result is a formal proof (constructive and via case analysis) that exactly six atomic relationship-types exist under the logical space of $(a, b) \in E_N$ for $N$ arbitrary. Disjointness is immediate from the logical character of the case distinctions, and exhaustiveness is obtained by enumerating all possible identity, difference, and null combinations for $a$ and $b$ (Favre et al., 2013).

The crucial corollary is that no further “elementary” coordinated dyadic relationship exists beyond those formalized by the four RMT types + asocial/null. This constitutes a theoretical closure property, supporting the RMT framework from first principles.

7. Illustrative Examples and Further Implications

Applied instances include:

• EM: $(\text{invite}, \text{invite})$ → Dinner reciprocity.
• CS: $$\{(\text{gift}, \varnothing), (\varnothing, \text{gift})\}$$ → Gift exchange sequences.
• AR: $(\text{pay}, \text{service})$ only, never $(\text{service}, \text{pay})$ → Payment-for-service hierarchy.
• MP: $$\{(\text{apples}, \text{bananas}), (\text{bananas}, \text{apples})\}$$, roughly alternating, → Barter.

This atomic approach provides a compact symbolic calculus for the combinatorics of dyadic coordination, enables formal analysis and classification of observed relationship data, and theoretically constrains the universe of dyadic social forms (Favre et al., 2013).

Further research directions include the extension to triadic/multiparty settings, continuous action spaces, and dynamic network inference, but the atomic closure at the dyadic level is established rigorously within this formalism.