Rapport Management Theory

• Rapport Management Theory is defined as a framework that analyzes interpersonal communication by managing illocutionary, relational, and self-presentation domains.
• It categorizes politeness into true politeness, impoliteness, and mock politeness, offering a detailed insight into the dynamics of social interactions.
• The theory underpins computational models for annotating and simulating social exchanges, enhancing cross-cultural AI and agent-based simulations.

Rapport Management Theory (RMT) articulates the pragmatics of interpersonal communication as fundamentally oriented toward the management of rapport—harmony, social distance, and personal identity—within social interaction. Developed by Spencer-Oatey (2008), RMT decomposes every communicative act into simultaneous operations over three distinct but interacting domains: illocutionary management (managing the force and appropriateness of speech acts), relational management (calibrating power, distance, and rank), and self-presentation management (projecting social identity and image). Central to RMT is the proposition that (im)politeness phenomena arise from the strategic balancing of these domains. These include conventional politeness, overt impoliteness, and complex forms such as mock politeness (i.e., surface-level civility masking negative intent). Formal and computational models extend RMT to enable annotation, detection, and simulation of these strategies, notably in cross-cultural AI/LLM and agent-based environments (Zhang et al., 3 Feb 2026, Bölöni et al., 2018).

1. Theoretical Structure of Rapport Management

RMT frames communication as a multidimensional negotiation, where meaning and social effect are determined not merely by semantic content but by the indexical management of three core domains:

• Illocutionary Management: Oversight of the illocutionary force and contextually appropriate action selection (e.g., requests, apologies) in accordance with genre and convention.
• Relational Management: Regulation of relative social distance, power hierarchies, and institutional roles between interlocutors.
• Self-presentation Management: Maintenance of personal or collective face, social reputation, and alignment with group identities.

Each communicative event is thus a vector in the space spanned by these domains, with the expressive and interpretive weight distributed according to genre norms, role expectations, and sociocultural constraints. RMT thereby integrates classical face theory with finer-grained pragmatic and interactional analyses (Zhang et al., 3 Feb 2026).

2. Classification of Politeness Phenomena

RMT underpins a three-way categorization of communicative strategies, as formalized in recent computational-linguistic research:

• True Politeness ($P$): Linguistic or behavioral strategies that are contextually appropriate and which serve to enhance or maintain interpersonal harmony. Operationally, $P$ is marked by alignment between polite surface markers and the expectations or presuppositions of the communicative context.
• Impoliteness ($I$): Strategies causing threat to face or negative rapport impact; marked by contextually inappropriate, offensive, or face-damaging expressions.
• Mock Politeness ($M$): A subtype of impoliteness, characterized by a pragmatic mismatch: polite external form (markers or lexemes) accompanied by impolite intent or interpretable effect, typically uncovered through inference or context violation.

Formally, let $U = \{u_1, u_2, …\}$ denote the set of utterances, $C$ the vector of context parameters (roles, setting, distance and power metrics), and $L = \{P, I, M\}$ the category labels. The functional classifier $f: U \times C \rightarrow L$ is driven by cues both at the semantic-surface and pragmatic-inference levels. For $M$ (Mock Politeness), key indicators are external mismatch (polite surface, impolite contextual fit) or internal mismatch (juxtaposition of polite and impolite content within an utterance) (Zhang et al., 3 Feb 2026).

3. Formalization and Operational Criteria

Although RMT is primarily a theory of pragmatic interaction, recent work embeds its insights into computational pipelines for annotation and model training, using minimal formal structures:

• Surface-form Politeness Score: $surface\text{-}form: U \rightarrow \mathbb{R}_+$, quantifying the presence of markers such as honorifics, mitigating particles, or formulaic lexemes.
• Implicature-intensity: $$implicature\text{-}intensity: U \times C \rightarrow \mathbb{R}_+$$, representing the inferred impoliteness or negative stance given context.

A mock-politeness decision rule is approximated by

$$u \in M \iff surface\text{-}form(u) > \tau_1 \wedge implicature\text{-}intensity(u, C) > \tau_2$$

where thresholds $\tau_1, \tau_2$ are not specified but operationalized via human-annotated mismatch criteria. The taxonomy can be summarized as $$L = \{\text{TruePoliteness}, \text{Impoliteness}, \text{MockPoliteness}\}$$. For Mock Politeness:

$$\text{MockPoliteness} = \mathrm{MP}_{\mathrm{ext}} \cup \mathrm{MP}_{\mathrm{int}}$$

with $$\mathrm{MP}_{\mathrm{ext}} = \{(u,C)\mid \mathrm{PoliteForm}(u) \wedge \neg \mathrm{PoliteContext}(u,C)\}$$ and $$\mathrm{MP}_{\mathrm{int}} = \{u\mid u=[p_\text{segment}; i_\text{segment}]\}$$ (Zhang et al., 3 Feb 2026).

4. Illustrative Classification and Annotation Techniques

Dataset annotation under the RMT paradigm leverages explicit cues and context valuation, as shown below:

Label	Marker/Feature Example	Contextual Criterion
True Politeness ($P$)	honorifics (“皇上”, “嫔妾”); formal titles	Surface and context in harmony; power/distance aligned
Impoliteness ($I$)	direct insults, blunt imperatives	No mitigating/pragmatic compensation
Mock Politeness ($M$)	polite surface + “（笑）” or sarcasm; formulaic apology	Surface politeness with discordant context or intent

Annotators identify both surface markers and the pragmatic consistency/mismatch, relying on RMT’s insistence that politeness is indexed to social distance and power, but with outcomes modulated by genre and institutional norms. Pragmatic inference, often nontrivial, is required to detect veiled or ironic negative stance behind formally polite expressions (Zhang et al., 3 Feb 2026).

5. Computational-Modeling Extensions: CSSM Formalism

In computational implementations (e.g., agent-based social simulations), “politeness” and its variants are operationalized through Culture-Sanctioned Social Metrics (CSSMs). Each CSSM is a five-argument metric

$CSSM(C, M, SA, PA, EA)$

where $C$ is culture, $M$ is the social value (e.g., Politeness), $SA$ is the subject agent, $PA$ is the perspective agent, and $EA$ is the estimator. The value assigned reflects $EA$’s estimate of how $PA$ perceives $SA$’s standing in $M$ under $C$ (Bölöni et al., 2018).

Politeness metrics (both internal and public) are dynamically updated following logistic action-impact functions (AIFs), such as

$\Delta p_\text{pub} = F_\text{pub}(x, y, b, ...)$

$$\Delta p_\text{int} = s \cdot F_\text{pub}(x, y, b, ...)$$

where $s$ is a “sincerity factor” distinguishing genuine ($s \approx 1$) from mock ($0 < s \ll 1$) politeness. Divergence between public and internal politeness reveals simulated mock politeness. In planning settings, agents may maximize weighted sums of CSSMs, selecting high-impact, low-sincerity actions when public display outweighs internal conviction (Bölöni et al., 2018).

6. Applications and Significance in Linguistic Technology

RMT and its computational extensions provide a foundational schema for:

• LLM Benchmarking and Prompt Engineering: Constructing datasets and evaluation metrics that disambiguate true from mock politeness, enabling robust pragmatic comprehension in state-of-the-art LLMs (e.g., GPT-5.1, DeepSeek) under varied prompting conditions.
• Annotation and Corpus Creation: Explicit operationalization of mismatch and context gives rise to high-quality labeled data in cross-cultural and multilingual settings.
• Agent-Based Social Simulation: Modeling agent rationality and decision-making where social harmony, reputation, and sincerity trade-offs influence strategy, plan selection, and interactional outcomes.

A plausible implication is that future pragmatic AI architectures require not only semantic analysis but also explicit modeling of sincerity and context-driven mismatches to approach human-like communicative competence. RMT’s integration into annotation, simulation, and interface design supports nuanced analyses of civility, irony, and covert aggression in both human and artificial discourse (Zhang et al., 3 Feb 2026, Bölöni et al., 2018).