Model of Mock Politeness

• Mock politeness is a subtype of impoliteness that uses polite markers to mask negative intent in communicative acts.
• The model employs a three-way classification schema based on RMT to distinguish true politeness, impoliteness, and mock politeness.
• Computational frameworks like CSSM enable accurate simulation and analysis of nuanced politeness behaviors across diverse cultural contexts.

Rapport Management Theory (RMT) is a pragmatics-driven theoretical framework for analyzing the management of interpersonal relations through language and communicative acts. RMT posits that all communicative behavior can be modeled as the management of multiple domains of rapport, with particular attention to the mechanisms by which speakers maintain, enhance, threaten, or manipulate social harmony. This theory is foundational not only for understanding politeness and impoliteness but also for formalizing and operationalizing related constructs such as “mock politeness.” Empirical work—particularly in cross-cultural and computational linguistics—employs RMT to create classification schemas, annotation protocols, and simulation models that robustly map communicative phenomena to underlying social, cultural, and contextual variables (Zhang et al., 3 Feb 2026, Bölöni et al., 2018).

1. Theoretical Framework and Domains

RMT, as articulated by Spencer-Oatey (2008) and adopted in contemporary computational pragmatics, identifies three interlinked domains central to every communicative act:

• Illocutionary Management: The management of conventional communicative acts as interpreted through shared norms.
• Relational Management: Navigation of relative social distance, power, and rank, shaping the appropriateness and impact of communicative choices.
• Self-presentation Management: The ongoing construction, presentation, and defense of personal identity and social image.

The equilibrium among these domains determines the perceived politeness or impoliteness of an utterance. Politeness phenomena are not regarded as absolute property of form but as outcomes of pragmatic balancing across the above domains. RMT's multi-domain approach underpins annotation schemes and classifier designs in computational linguistics that seek to capture nuanced distinctions between true politeness, impoliteness, and mock politeness (Zhang et al., 3 Feb 2026).

2. Classification Schema: Politeness, Impoliteness, and Mock Politeness

Leveraging RMT, a three-way classification schema differentiates fundamental pragmatic categories:

• True Politeness (P): Utterances employing contextually appropriate linguistic/behavioral strategies to maintain or enhance harmonious interpersonal relations.
• Impoliteness (I): Strategies that are contextually inappropriate or offensive, threatening face or damaging rapport.
• Mock Politeness (M): A subtype of impoliteness where “polite” forms are used with impolite intent or effect. Mock politeness involves a pragmatic mismatch: polite surface markers coupled with underlying negative evaluative stance, recognized through inference rather than form alone.

These categories are formally represented as

$$L = \{\text{TruePoliteness},\,\text{Impoliteness},\,\text{MockPoliteness}\}$$

with further distinctions within Mock Politeness between external ($\mathrm{MP}_{\mathrm{ext}}$) and internal ($\mathrm{MP}_{\mathrm{int}}$) mismatch mechanisms (Zhang et al., 3 Feb 2026).

3. Formal Modeling and Computational Structures

RMT has been integrated into computational pragmatics and social simulation via two main frameworks:

A. Minimal Formal Skeleton for Classification

Given

• $U = \{u_1, u_2, ...\}$: Set of utterances in context
• $C$: Context parameters (roles, setting, power, distance)
• $L = \{P, I, M\}$: Classification labels

Define the classifier

$f : U \times C \rightarrow L$

Mock Politeness is detected by “mismatch” conditions, partitioned as:

• $$M_{\mathrm{ext}} = \{ (u, C) \mid \text{surface-form}(u) \in \text{markers\_of\_politeness} \wedge \text{context\_evaluation}(u, C)=\text{impolite} \}$$
• $$M_{\mathrm{int}} = \{ u \mid \exists u_1, u_2 \subseteq u,\, u_1=\text{polite},\,u_2=\text{impolite/sarcastic} \}$$

Detection involves explicit annotation of surface markers and pragmatic implicature, with final category assignment via consensus on mismatch (Zhang et al., 3 Feb 2026).

B. Culture-Sanctioned Social Metrics (CSSM)

In the CSSM paradigm, social values such as politeness are represented as metrics:

$\text{CSSM}(C, M, SA, PA, EA)$

where $C$ indexes culture, $M$ is the metric (e.g., “Politeness”), $SA$ the subject agent, $PA$ the perspective agent, and $EA$ the estimator agent (Bölöni et al., 2018).

Updates to politeness values follow action impact functions that combine action parameters (e.g., “loudness,” “rudeness”), beliefs, and prior CSSMs via sums and products of logistics:

$$F^{CSSM}_i = \sum_k \prod_l L\left(x_{kl}; K_{kl}, M_{kl}, B_{kl}\right),\quad L(x; K, M, B) = \frac{K}{1 + e^{-B(x - M)}}$$

Mock politeness is modeled as a disconnect between internal and public CSSMs for politeness, with a sincerity parameter $s$ modulating the internal update:

$$\Delta p_{\text{int}} = s \cdot \Delta p_{\text{pub}}$$

Values $s \ll 1$ indicate insincere or “mock” politeness (Bölöni et al., 2018).

4. Annotation and Typology Strategies

Empirical studies relying on RMT implement rigorous multi-criterial annotation protocols:

• Surface Politeness Markers: Honorifics, mitigating particles, softeners.
• Impoliteness Markers: Insults, blunt imperatives, taboo expressions.
• Mismatch Evidence: “External mismatch” (polite forms used incongruently with context); “internal mismatch” (juxtaposition of polite lexemes with sarcasm, laughter tags, or ironic content).

Table: Taxonomy of Politeness Phenomena (Zhang et al., 3 Feb 2026)

Category	Surface Markers	Contextual Alignment
True Politeness	Honorifics, softeners	Role, status, appropriateness
Impoliteness	Blunt, direct, offensive language	Incongruent, threatening
Mock Politeness	Polite lexemes plus sarcasm/irony	Mismatch (external/internal)

This typology operationalizes RMT for both human and computational annotation tasks, enabling explicit benchmarking of model comprehension, as exemplified in datasets pairing authentic and simulated discourse (Zhang et al., 3 Feb 2026).

5. Model Application: LLMs and Social Simulation

RMT’s constructs have been directly applied in evaluating LLMs on pragmatic tasks. In “The Mask of Civility,” LLM performance in categorizing Chinese utterances as polite, impolite, or mock-polite was benchmarked across four prompting strategies, with explicit use of RMT’s framework for annotation and performance calibration (Zhang et al., 3 Feb 2026). This approach reduces false-positive politeness judgments, enhances the detection of concealed impoliteness, and provides a cross-linguistic, cross-cultural baseline for evaluating AI interactional competence.

In agent-based simulations, CSSM-based models have permitted granular simulation of social behavior, deception, and politeness dynamics. By tracking both internal and public metrics for politeness, and introducing a tunable sincerity coefficient, these models can generate, explain, and predict both genuine and feigned acts of rapport management (Bölöni et al., 2018).

6. Significance and Extensions

RMT offers an integrative, multidimensional scaffold for research at the interface of linguistics, cultural psychology, and AI. Its formalization supports:

• Context-sensitive annotation protocols for manual or automated corpus analysis,
• Computational models capable of simulating or recognizing nuanced pragmatic phenomena such as mock politeness,
• Engineering of prompting systems and social agents with improved sensitivity to relational meaning.

The distinction between public and internal rapport metrics, together with explicit modeling of context and sincerity, facilitates extensions to new domains (e.g., cross-cultural studies, real-time conversational AI), suggesting a path toward increasingly human-aligned communicative technology (Zhang et al., 3 Feb 2026, Bölöni et al., 2018). A plausible implication is that RMT-based frameworks will become increasingly vital in the design and evaluation of socially-aware AI systems.