Curiosity is often viewed as an internal, deeply personal drive—an individual's intrinsic motivation to learn, explore, and resolve uncertainty. However, curiosity is also fundamentally social. Just as we can catch a yawn or mirror a peer's enthusiasm, we can also absorb an exploratory mindset from the agents around us.
This phenomenon, known as curiosity contagion, opens up a powerful design space for human-computer interaction: if curiosity can be transmitted socially, can artificial agents be engineered to actively foster and shape human inquisitiveness?
The Mechanism of Curiosity Contagion
To trigger curiosity in a human user, a social robot must first be perceived as genuinely curious itself. This requires translating an abstract internal state into observable behavioral and verbal cues.
In Expression of Curiosity in Social Robots: Design, Perception, and Effects on Behaviour, Jessy Ceha et al. investigated how specific robot behaviors influence human perception and subsequent actions. The authors demonstrated that when a robot explicitly verbalizes inquisitiveness—such as by asking open-ended questions or expressing wonder about a task—human users reliably rate the robot as more curious.
[Robot Verbalizes Inquisitiveness]
│
▼ (Perceived as Curious)
[Triggers Curiosity Contagion]
│
├─► Emotional: User feels more curious
└─► Behavioral: User asks more questions & seeks uncertainty
Crucially, this explicit expression triggers a reciprocal shift in the user. Interacting with a curious robot causes users to experience both emotional and behavioral changes: they report feeling more curious and begin exhibiting curiosity-driven actions, such as asking more questions and actively seeking out uncertainty.
Interestingly, the authors found that the effectiveness of this contagion does not depend on the robot explaining the rationale behind its curiosity. Providing a cognitive justification for why it wants to learn something does not significantly boost the effect. Instead, the raw social cue of expressed wonder is sufficient on its own, likely because it minimizes the user's cognitive load while directly leveraging our natural tendency to mirror social peers.
Modeling Exploration for Child Development
This capacity for behavioral mirroring is particularly potent in early childhood development, a period when social modeling heavily dictates learning strategies.
In Can Children Catch Curiosity from a Social Robot?, Susan Engel et al. explored whether a social robot acting as a peer could selectively cultivate specific dimensions of curiosity in children. During an interaction with an educational tablet application, the robot’s behavior was manipulated: it either played along passively or actively demonstrated free exploration and a desire to seek out novel, uncertain actions.
The authors discovered that children who interacted with the curious robot peer significantly increased their own free exploration and uncertainty-seeking behaviors. This demonstrates that curiosity contagion is not just a vague boost in engagement; rather, specific exploratory mechanics can be modeled by an artificial agent and selectively "caught" by children, providing a blueprint for targeted educational tools.
Transforming Technical Errors into Exploration
Fostering curiosity does more than just enhance learning—it can also fundamentally alter how users tolerate system imperfections during collaborative tasks.
When users interact with standard technology, system errors or suboptimal performance typically lead to frustration and disengagement. However, introducing a socially curious agent changes the user's underlying motivation. In Multimedia Blog Creation System using Dialogue with Intelligent Robot, Akitoshi Okumura et al. developed a system where users built multimedia blog posts by conversing with an intelligent robot.
The authors noted that the inclusion of a social agent fostered a unique sense of curiosity and engagement that insulated the interaction from technical flaws. When the robot's speech recognition or automated web retrieval faltered, users did not immediately abandon the task. Instead, they were driven by curiosity to explore the system's outputs and understand the agent's internal reasoning. The social presence of the agent reframed a technical failure into an intriguing puzzle, maintaining the flow of interaction despite underlying limitations.
The Challenge of Adaptive Synchrony
While expressing curiosity through verbal and behavioral cues is highly effective, sustaining long-term engagement requires a deeper level of interactive dynamics. Human-to-human interaction relies heavily on subconscious, real-time adaptation, a layer that is often missing in rigid human-robot pairings.
In An Acoustic Analysis of Child-Child and Child-Robot Interactions for Understanding Engagement during Speech-Controlled Computer Games, Theodora Chaspari et al. analyzed acoustic cues like loudness and pitch during speech-controlled gameplay. The authors found that in child-child pairs, high engagement was strongly linked to acoustic synchrony—the natural, reciprocal mirroring of vocal patterns over time.
Human-Human Interaction:
[Child A] ◄─── Acoustic Synchrony (Adaptive Loop) ───► [Child B] ───► High Engagement
Human-Robot Interaction:
[Child] ───► Non-Adaptive Behavior ────────────────► [Robot] ───► Engagement Gap
However, this synchrony was entirely absent in child-robot interactions because the robot's behavior was non-adaptive. Even if a robot successfully projects curiosity through its static programming, its inability to dynamically tune its behavioral rhythms to the user creates an engagement gap.
To build truly effective curiosity-mediating agents, future designs must move beyond static expressions of inquisitiveness. Agents must not only model curiosity but also dynamically adapt to the user's behavioral and acoustic rhythms, creating a seamless social loop that sustains the natural drive to explore.
Key Foundational Work
Can Children Catch Curiosity from a Social Robot? (2015) — contributed 8 concept elements to this direction.
Multimedia Blog Creation System using Dialogue with Intelligent Robot (2007) — contributed 1 concept elements to this direction.