Affective and Conversational Predictors of Re-Engagement in Human-Robot Interactions -- A Student-Centered Study with A Humanoid Social Robot

Humanoid social robots are increasingly present in daily life, making sustained user engagement a critical factor for their effectiveness and acceptance. While prior work has often examined affective evaluations or anthropomorphic design, less is known about the relative influence of dynamic conversational qualities and perceived robot characteristics in determining a user’s intention to re-engage with Large Language Model (LLM)-driven social robots. In this study, 68 participants interacted in open-ended conversations with the Nadine humanoid social robot, completing pre- and post-interaction surveys to assess changes in robot perception, conversational quality, and intention to re-engage. The results showed that verbal interaction significantly improved the robot’s perceived characteristics, with statistically significant increases in pleasantness ($p<.0001$) and approachability ($p<.0001$), and a reduction in creepiness ($p<.001$). However, these affective changes were not strong and unique predictors of users’ intention to re-engage in a multiple regression model. Instead, participants’ perceptions of the interestingness ($β=0.60$, $p<.001$) and naturalness ($β=0.31$, $p=0.015$) of the robot’s conversation emerged as the most significant and robust independent predictors of intention to re-engage. Overall, the results highlight that conversational quality, specifically perceived interestingness and naturalness, is the dominant driver of re-engagement, indicating that LLM-driven robot design should prioritize engaging, natural dialogue over affective impression management or anthropomorphic cues.

💡 Research Summary

This paper investigates which factors most strongly predict users’ intention to re‑engage with a Large Language Model (LLM)‑driven humanoid social robot. Sixty‑eight university students (46 female, 22 male, ages 18‑34) participated in a controlled study at the University of Geneva. Each participant completed pre‑interaction surveys measuring three affective robot characteristics—pleasantness, creepiness, and approachability—using a 7‑point Likert scale. Participants then engaged in a ten‑minute open‑ended conversation with the Nadine humanoid robot, which is powered by an LLM to generate natural language responses. After the interaction, the same affective measures were repeated, and additional questionnaires assessed perceived conversational quality along three dimensions: naturalness, human‑likeness, and interestingness. A final scale captured the participants’ intention to re‑engage with the robot.

Statistical analysis showed that verbal interaction significantly increased the robot’s perceived pleasantness (p < .0001) and approachability (p < .0001) while reducing perceived creepiness (p < .001). However, when these affective changes were entered into a multiple regression model predicting re‑engagement intention, they did not emerge as significant independent predictors. In contrast, perceived conversational quality demonstrated strong predictive power. Specifically, interestingness yielded the largest standardized coefficient (β = 0.60, p < .001), followed by naturalness (β = 0.31, p = 0.015). Human‑likeness was not a significant predictor. These findings indicate that users’ willingness to interact with the robot again is driven more by how engaging and fluid the dialogue feels than by changes in affective impressions of the robot’s appearance or demeanor.

The authors discuss the implications for HRI design, arguing that while affective improvements (e.g., making the robot seem more pleasant) can enhance short‑term likability, they are insufficient to sustain long‑term engagement. Instead, designers should prioritize developing LLM‑based dialogue systems that deliver interesting, context‑appropriate, and naturally flowing conversations. The study also challenges the prevailing emphasis on anthropomorphism; merely making a robot sound or look more human‑like does not guarantee continued user interaction if conversational quality is lacking.

Limitations include the homogeneous student sample, the short interaction duration, and the lack of longitudinal tracking of actual re‑engagement behavior. Future work should explore diverse demographic groups, real‑world deployment contexts, and longer‑term usage patterns to validate and extend these findings.

Overall, this research provides empirical evidence that conversational quality—particularly perceived interestingness and naturalness—is the dominant driver of re‑engagement intentions in human‑robot interaction, offering clear guidance for the next generation of socially intelligent robots.