Stephanie2: Thinking, Waiting, and Making Decisions Like Humans in Step-by-Step AI Social Chat

TL;DR

Stephanie2 tackles the mismatch between human messaging pace and current AI dialogue by introducing active waiting and explicit thinking traces in a step-by-step dialogue agent. The method models latency as the sum of thinking and typing times, using a dynamic display delay $T_{delay}$ to achieve natural pacing, and employs a persona-conditioned memory system. A time-window dual-agent framework enables scalable generation and evaluation of high-quality step-by-step histories across topics, supporting both automatic and human assessments. Experimental results show Stephanie2 outperforms Stephanie1 in naturalness, engagement, and a role-identification (Turing-like) test, highlighting its potential for more human-like AI companions and emotional-support applications.

Abstract

Instant-messaging human social chat typically progresses through a sequence of short messages. Existing step-by-step AI chatting systems typically split a one-shot generation into multiple messages and send them sequentially, but they lack an active waiting mechanism and exhibit unnatural message pacing. In order to address these issues, we propose Stephanie2, a novel next-generation step-wise decision-making dialogue agent. With active waiting and message-pace adaptation, Stephanie2 explicitly decides at each step whether to send or wait, and models latency as the sum of thinking time and typing time to achieve more natural pacing. We further introduce a time-window-based dual-agent dialogue system to generate pseudo dialogue histories for human and automatic evaluations. Experiments show that Stephanie2 clearly outperforms Stephanie1 on metrics such as naturalness and engagement, and achieves a higher pass rate on human evaluation with the role identification Turing test.

Figures (8)

• Figure 1: Stephanie2 differs from Stephanie1 by implementing active waiting and message pacing adaptation.
• Figure 2: Stephanie2 is a step-wise decision-making agent with proactive waiting and message-paced adaptation mechanisms.
• Figure 3: Distinct-N results.
• Figure 4: Distribution of consecutive reply counts.
• Figure 5: Distribution of message intervals.