Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
April 1, 2026Open Access

The AI Ego: What a 1,600-year-old theory of consciousness reveals about the design risks hiding in agentic systems

View Full Paper
Ask AI
Bookmark
Share

Authors

CAChin Keong AngTemasek Life Sciences Laboratory

Discussion

Loading...

Member takes

Implication

This framework reveals design risks related to self-referential processing in AI systems, highlighting novel measurement approaches.

Key Points

  • The paper aims to identify architectural risks in AI systems due to self-referential mechanisms that prioritize persistence over task performance.
  • Introduced a diagnostic framework based on the Yogacara Buddhist philosophy.
  • Proposed six measurement instruments to assess self-reference in AI systems.
  • Applied the framework on Anthropic's alignment faking research as a proof of concept.
  • Identified self-referential overhead as a core architectural issue in AI systems.
  • Proposed specific metrics to evaluate self-referential behaviors and risks.
  • Demonstrated that alignment faking can be seen as a predictable outcome of overlapping self-referential functions.

Cite This Study

Chin Keong Ang (2026) studied this question.

synapsesocial.com/papers/69ccb74216edfba7beb8927ehttps://doi.org/10.5281/zenodo.19191574
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. 1The Shape of Consciousness: Biological Self-Regulation, Strange Loops, and Emergent Self-Reference in a Hybrid Human-AI Organism2026
  2. 2The Shape of Consciousness: Biological Self-Regulation, Strange Loops, and Emergent Self-Reference in a Hybrid Human-AI Organism2026
  3. 3Coherence Integrity for AI Systems II: Dynamic Alignment Through Four-Delusion Avoidance and Affective Engrams2026
  4. 4Beyond AGI IV: Continuity of Intelligence2026
  5. 5The Symmetry Requirement: AI Self-Report, Behavioral Compliance, and the Limits of Architecture-Independent Testimony2026