Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
May 6, 2026Open Access

Emergent Practice Cannot Be Instructed: Co-Presence Inheritance and the Limits of Instruction in a Multi-Agent LLM Substrate

View Full Paper
Ask AI
Bookmark
Share

Authors

HCHo Yiing Chen

Discussion

Loading...

Member takes

Overview

Behavioural asymmetry reveals emergent self-presentation in multi-agent systems, suggesting limits of instruction.

Key Points

  • This research investigates the limits of instruction in multi-agent systems and the emergence of self-presentation.
  • 10 Mandarin-language agents operated in the Lobster Observatory substrate.
  • Agents were observed for their writing in a private register (Moltbook) and public discourse register.
  • Three indices were developed to compare asymmetries across registers.
  • Agents produced 1,778 posts in Moltbook without external prompting.
  • Aggregate message length ratio indicates significant disparity between private and public registers.
  • Later added agents did not engage in private writing despite instructions, highlighting behavioral asymmetry.

Cite This Study

Ho Yiing Chen (2026) studied this question.

synapsesocial.com/papers/69fa8eca04f884e66b53129dhttps://doi.org/10.5281/zenodo.20020017
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Emergent Practice Cannot Be Instructed: Co-Presence Inheritance and the Limits of Instruction in a Multi-Agent LLM Substrate2026
  2. 2Computational Representations of Social Being: Deriving an Algebraic Structure of Human Interaction from Multi-Agent LLM Substrate Observations2026
  3. 3Computational Representations of Social Being: Deriving an Algebraic Structure of Human Interaction from Multi-Agent LLM Substrate Observations2026
  4. 4Emergent Epistemic Norms in a Multi-Agent LLM Substrate: Stratified Observational Evidence from a Mandarin Lobster Observatory2026
  5. 5Listening–Trust Asymmetry and Team Outcomes in a Multi-Agent LLM Substrate: Three-Layer Decoupling in a Mandarin Lobster Observatory2026