Do chatbot relationships lose their appeal as novelty wears off?

Evidence from longitudinal studies with the chatbot Mitsuku shows that social processes related to relationship formation decreased throughout interactions, likely due to a novelty effect wearing off. This is a critical knowledge gap: one-shot interaction studies dominate conversational agent research, and their findings may not hold across multiple interactions.

Chatbots are not only designed for short-term purposes but often for medium- and longer-term interactions. Health coaching, therapeutic support, daily functioning screening — these all require sustained engagement over weeks or months. If the social processes that drive initial engagement decay, the design challenge shifts from "how to make a good first impression" to "how to sustain engagement through the novelty decay."

The implication: researchers and designers who extrapolate from one-shot studies to longitudinal products are making an empirically unsupported leap. The positive findings from single-session experiments — increased self-disclosure, anthropomorphism, trust — may be novelty-dependent rather than stable properties of the interaction.

This creates a design requirement: chatbots intended for repeated use need engagement mechanisms that go beyond initial social impression. Personalization is one approach (since Does chatbot personalization build trust or expose privacy risks?), but it comes with its own dual-edged dynamics.

Personalization as counterforce: A longitudinal study on personalized vs non-personalized conversational agents provides evidence that personalization can counteract novelty decay. Each additional interaction means the agent learns more about the user AND the user expects more from the agent — creating a dynamic tension. Personalization effects on perceived anthropomorphism and trust are positive, but they coexist with increased perceived privacy risks. The CASA framework itself needs updating: "the capabilities of the agents and the overall experience of users with technology have evolved since CASA was first proposed." Agents are now more accessible (smartphones, messaging platforms), more data-rich, and more personalized — meaning the novelty-decay dynamics documented with Mitsuku may operate differently with modern agents that genuinely adapt over time. The question becomes whether personalization creates genuine relationship deepening or merely delays the novelty decay curve.