Not Forgotten: Implementation and Evaluation of a Personalized Episodic Memory for the Humanoid Robot Head Kim

2026-07-27

AI Summary

This study addresses the challenge that existing social robots struggle to build long-term user relationships due to a lack of cross-session memory. To overcome this limitation, the authors propose a lightweight episodic memory module for the humanoid robot Kim, which innovatively combines cosine similarity with memory strength to form a hybrid retrieval mechanism. This approach dynamically retrieves and integrates relevant historical interactions into the prompt for a large language model–based dialogue system, leveraging vector-based semantic search while avoiding privacy violations and the uncanny valley effect. An online user study (N = 43) demonstrates that this method significantly enhances users’ perceptions of the robot’s social presence (d = 0.60), credibility (d = 0.62), and warmth (d = 0.56), without increasing perceived intrusiveness (d = 0.00).

Abstract

Social robots that rely on large language models for conversation are unable to retain information across sessions. This absence of memory violates social expectations, potentially preventing the formation of persistent relationships. This paper presents a lightweight episodic memory module that integrates vector-based semantic retrieval with an LLM-controlled dialog system, deployed on the humanoid robot head Kim. The module employs a hybrid scoring function combining cosine similarity with a memory strength metric to retrieve contextually relevant past interactions and inject them into the generation prompt. The system was evaluated in a within-subjects video-based online study (N = 43) using the Human-Robot Interaction Evaluation Scale (HRIES). Results show that episodic memory significantly increased perceived sociability (d = 0.60, p < .001), with the strongest effects on perceived trustworthiness (d = 0.62) and warmth (d = 0.56). Perceived disturbance remained unchanged (d = 0.00), indicating that the implemented approach to personalized recall did not trigger privacy-related discomfort or uncanny valley effects. These findings suggest that episodic memory serves as a social lubricant in embodied Human-Robot Interaction, enhancing relational quality without eliciting negative affective responses.

Problem

• episodic memory
• social robots
• human-robot interaction
• conversational AI
• memory retention

Innovation

• episodic memory
• human-robot interaction
• large language models
• semantic retrieval
• personalized recall