EchoGuard: An Agentic Framework with Knowledge-Graph Memory for Detecting Manipulative Communication in Longitudinal Dialogue
arXiv:2603.04815v1 Announce Type: new Abstract: Manipulative communication, such as gaslighting, guilt-tripping, and emotional coercion, is often difficult for individuals to recognize. Existing agentic AI systems lack the structured, longitudinal memory to track these subtle, context-dependent tactics, often failing due to limited context windows and catastrophic forgetting. We introduce EchoGuard, an agentic AI framework that addresses this gap by using a Knowledge Graph (KG) as the agent's core episodic and semantic memory. EchoGuard employs a structured Log-Analyze-Reflect loop: (1) users log interactions, which the agent structures as nodes and edges in a personal, episodic KG (capturing events, emotions, and speakers); (2) the system executes complex graph queries to detect six psychologically-grounded manipulation patterns (stored as a semantic KG); and (3) an LLM generates targeted Socratic prompts grounded by the subgraph of detected patterns, guiding users toward self-discov
arXiv:2603.04815v1 Announce Type: new Abstract: Manipulative communication, such as gaslighting, guilt-tripping, and emotional coercion, is often difficult for individuals to recognize. Existing agentic AI systems lack the structured, longitudinal memory to track these subtle, context-dependent tactics, often failing due to limited context windows and catastrophic forgetting. We introduce EchoGuard, an agentic AI framework that addresses this gap by using a Knowledge Graph (KG) as the agent's core episodic and semantic memory. EchoGuard employs a structured Log-Analyze-Reflect loop: (1) users log interactions, which the agent structures as nodes and edges in a personal, episodic KG (capturing events, emotions, and speakers); (2) the system executes complex graph queries to detect six psychologically-grounded manipulation patterns (stored as a semantic KG); and (3) an LLM generates targeted Socratic prompts grounded by the subgraph of detected patterns, guiding users toward self-discovery. This framework demonstrates how the interplay between agentic architectures and Knowledge Graphs can empower individuals in recognizing manipulative communication while maintaining personal autonomy and safety. We present the theoretical foundation, framework design, a comprehensive evaluation strategy, and a vision to validate this approach.
Executive Summary
This article introduces EchoGuard, an agentic AI framework designed to detect manipulative communication in longitudinal dialogue. EchoGuard utilizes a Knowledge Graph as its core episodic and semantic memory, enabling the tracking of subtle, context-dependent manipulation tactics. The system employs a structured Log-Analyze-Reflect loop, which includes user interaction logging, graph-based detection of six psychologically-grounded manipulation patterns, and the generation of targeted Socratic prompts. This framework holds promise in empowering individuals to recognize manipulative communication while maintaining personal autonomy and safety. The approach has significant implications for the development of agentic AI systems and the prevention of emotional manipulation. A comprehensive evaluation strategy is proposed, and the authors present a vision for further validation.
Key Points
- ▸ EchoGuard is an agentic AI framework designed to detect manipulative communication in longitudinal dialogue.
- ▸ The system utilizes a Knowledge Graph as its core episodic and semantic memory.
- ▸ EchoGuard employs a structured Log-Analyze-Reflect loop to detect manipulation patterns.
Merits
Strength
EchoGuard's use of a Knowledge Graph enables structured, longitudinal memory to track subtle, context-dependent manipulation tactics.
Strength
The system's Log-Analyze-Reflect loop provides a structured approach to detecting manipulation patterns.
Demerits
Limitation
The evaluation strategy proposed in the article may be limited in scope, and further testing is necessary to validate the effectiveness of EchoGuard in real-world scenarios.
Limitation
The reliance on a Knowledge Graph may introduce scalability issues as the volume of user interactions increases.
Expert Commentary
The EchoGuard framework demonstrates a promising approach to addressing the complex issue of manipulative communication in longitudinal dialogue. By utilizing a Knowledge Graph as its core memory, EchoGuard is able to track subtle, context-dependent manipulation tactics in a way that existing agentic AI systems cannot. The structured Log-Analyze-Reflect loop employed by the system provides a clear and effective approach to detecting manipulation patterns. While the proposed evaluation strategy may be limited in scope, further testing of EchoGuard is necessary to validate its effectiveness in real-world scenarios. The article's focus on agentic AI development highlights the importance of creating systems that can navigate complex social interactions. Ultimately, EchoGuard has the potential to empower individuals to recognize manipulative communication and prevent emotional manipulation in various contexts.
Recommendations
- ✓ Further testing of EchoGuard in real-world scenarios is necessary to validate its effectiveness.
- ✓ The authors should consider expanding the evaluation strategy to include a broader range of user interactions and contexts.