Ekova: A Personality-Support Agent for Self-Discovery Dialogue

A recent paper on arXiv (2608.00478) presents Ekova, an AI agent focused on personality support (PS) in conversations, setting it apart from conventional emotional support (ES) systems. The authors contend that while ES seeks to reduce immediate emotional pain, PS is intended to assist users in gaining cognitive clarity and articulating themselves, without serving as counseling or clinical intervention. The research is organized into three components: initially, they introduce DSD, a Chinese dataset for self-discovery PS comprising 8,590 samples from real longitudinal interactions across five minimal units: Coach, Warm, Tsukkomi, Real, and Gonzo. Next, they develop DeepSupport, a multi-persona PS system utilizing OrthoTune, which includes style-specific adapters and a style-consistency regularizer. Finally, they integrate the five personas into Ekova, a persistent agent with a cohesive cross-session memory layer. This paper marks a significant shift from ES to PS in AI dialogue systems, particularly regarding self-discovery and personality support, and introduces innovative datasets and training methodologies. Although a publication date is not provided, the arXiv ID indicates a submission in 2026. Researchers in AI, psychology, and human-computer interaction may find this work particularly relevant as it sets a new direction for dialogue systems beyond mere emotional assistance.