Avoiding Conflict: When Speaker Coordination Does Not Require Conceptual Agreement

Area of Science:

• Computational Linguistics
• Cognitive Science
• Sociolinguistics

Background:

• The socialization hypothesis posits that shared environments lead to shared concepts, a prerequisite for successful communication.
• Distributional semantic models (DSMs) offer a computational approach to investigate semantic similarity and conceptual alignment.

Purpose of the Study:

• To challenge the socialization hypothesis by exploring conceptual alignment using DSMs.
• To investigate whether shared concepts are a necessary prerequisite for effective communication.

Main Methods:

• Utilized distributional semantic models (DSMs) trained on distinct corpora to represent lexical meaning.
• Modeled inter-subject coordination as dimensionality selection via Singular Value Decomposition (SVD) on DSM similarity matrices.
• Employed alignment-based similarity metrics to quantify conceptual alignment across semantic spaces.

Main Results:

• Demonstrated that compatibility, not mere agreement, is key in alignment-based conceptual similarity.
• Showed that unique individual experiences (idiosyncrasy) can facilitate, rather than hinder, semantic coordination.
• Highlighted the distinct roles of singular vectors in spreading linguistic information within DSMs.

Conclusions:

• The socialization hypothesis may not be a necessary prerequisite for successful communication.
• Effective communication can be viewed as a cooperative act of avoiding conflict rather than maximizing agreement.
• Idiosyncratic experiences contribute to, rather than detract from, communicative success.