Organizations: The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, Guangdong, China · Tsinghua University, Beijing, China · Beijing Institute of Technology, Beijing, China · Sichuan University, Chengdu, Sichuan, China · Institute of Software, Chinese Academy of Sciences, Beijing, China · The Hong Kong University of Science and Technology, Hongkong, China
Designing affective behaviors for animal-inspired social robots often relies on intuition and personal experience, leading to fragmented outcomes. To provide more systematic guidance, we first coded and analyzed human-pet interaction videos, validated insights through literature and interviews, and created structured reference cards that map the design space of pet-inspired affective interactions. Building on this, we developed MojiKit, a toolkit combining reference cards, a zoomorphic robot prototype (MomoBot), and a behavior control studio. We evaluated MojiKit in co-creation workshops with 18 participants, finding that MojiKit helped them design 35 affective interaction patterns beyond their own pet experiences, while the code-free studio lowered the technical barrier and enhanced creative agency. Our contributions include the data-informed structured resource for pet-inspired affective HRI design, an integrated toolkit that bridges reference materials with hands-on prototyping, and empirical evidence showing how MojiKit empowers users to systematically create richer, more diverse affective robot behaviors.
Figures & tables
Figure 1. Overview of the MojiKit. (a) MojiKit integrates a zoomorphic robot prototype (MomoBot), design reference cards (human-, animal-centric, and environmental context), and a behavior control studio, to support emotion-oriented interaction design. (b) Affective interaction scenarios co-created in workshops, organized into three levels of realizability: Directly realizable with the current prototype (Companion Resting, Face-Cupping Touch), Enabled by moderate enhancements (Welcome-home Greeting, Amused by Robot Play), and Extended distributed morphologies (Tail-Wagging Greeting, Fluffy Resting). Figure shows the MojiKit framework and example design outcomes. Panel (a) illustrates three components: the zoomorphic robot prototype “MomoBot,” the Behavior Control Studio, and the Design Reference Cards, organized into human-centric, animal-centric, and environmental modules. Panel (b) depicts six affective interaction scenarios co-created in workshops: (i) companion resting, (ii) face-cupping touch, (iii) welcome-home greeting, (iv) amused by robot play, (v) tail-wagging greeting, and (vi) fluffy resting. Scenarios are positioned along a continuum from directly realizable with the prototype, to requiring moderate enhancements, to extended distributed morphologies.
Figure 2. Design process of the reference cards: from video-based deductive coding (V1.0), to literature integration on animal behaviors and emotions (V2.0), to refinement via interviews with pet owners, resulting in the finalized Design Reference Cards V3.0 . Figure illustrates the design process of the reference cards in three stages. Left: data collection and deductive coding of 65 human–pet interaction videos, producing Structured Data V1.0 with explicit behaviors and implicit affective elements across human, pet, and environment. Center: integration of 24 ethological studies, adding generalizable animal behaviors and emotion mappings, resulting in Structured Data V2.0. Right: validation and refinement through interviews with five pet owners, producing Reference Cards V3.0 organized into three modules, eight cards: human intentions/behaviors/interpretations, cat–dog behavior–emotion mappings, and environmental context.
Figure 3. Design Reference Cards comprising three modules (eight cards in total) of human–pet interaction. Example: the “Human Interaction Intentions” card, detailing motives such as human-initiated actions, passive responses, and non-interactive intent. A two-panel reference card for human-pet interaction design. Left: Color-coded sections categorize animal-centric (e.g., cat or dog emotional expressions), human-centric (e.g., interaction intentions), and environmental cards. Right: Detailed breakdown of human intentions, including proactive interaction, passive response to animal-initiated actions, non-interactive scenarios (e.g., refusal/neglect), and routine behaviors like preparing food or spatial transitions.
Figure 4. The structure of MomoBot. (a) Exploded view of MomoBot’s internal components with servomotors. (b) Structural cutaway showing the plush exterior and embedded mechanisms. (c) Motion ranges of ears, head, limbs, and tail across pitch, rotation, lifting, flexion, wagging, and curling. A three-panel figure illustrating the design and motion capabilities of MomoBot, an animal-inspired robot. Left: Exploded view highlighting components including servomotors, head, ears, body, tail, and limbs. Center: Integrated internal and external structure. Right: Range of motion visualizations for ears (pitch/rotation), head (pitch/rotation), limbs (lifting/flexion), and tail (wagging/curling), with angular degrees annotated.
Figure 5. Action editing and execution in MomoBot. (a) Motion sequences created in the Behavior Control Studio are (b) sent via Wi-Fi to the ESP32 and relayed as serial commands to the Arduino Mega, (c) which drives 16 servomotors. The Studio supports (d) timeline editing, (e) code editing, and (f) preset actions. A multi-panel figure describing the "Behavior Control Studio" for editing and communicating motion sequences to the MomoBot robot via WiFi. (a) Main studio interface for sequencing actions. (b) Hardware components: ESP32, Arduino MEGA 2560, batteries, and a customized expansion board. (c) Diagram of 16 servomotors distributed across MomoBot's body. The lower half of the diagram is Alternative editing methods: (d) Timeline editor with drag and double-click interactions, (e) Code editor with timeline synchronization, (f) Preset command library with 15 pre-built actions.
Group/ID
Gen
Age
Background
Pet Exp.
Animal Interaction Summary
G1
P01
F
26
Interaction Design
N/A
Occasional park encounters; touch/observation
P02
F
27
Interaction Design
Dog (7 yrs)
Daily care: feeding, grooming, play, health
G2
P03
F
26
Biomedical Eng. & Robotics Design
Dog (3+ yrs)
Regular care: feeding, grooming, play
P04
F
25
Urban & Rural Planning Design
N/A
Occasional visits; play/observation
G3
P05
M
31
Information Art & Design
Dog (5 yrs)
Deep care; frequent play; behavioral familiarity
P06
F
29
Interaction Design
N/A
Frequent pet cafés/friends’ pets; casual interaction
Table 1. Participant Demographics and Animal Interaction Experience. A table titled ``Participant Demographics and Animal Interaction Experience'' with 18 participants (P01-P18) across 9 groups (G1-G9). Columns list participant ID, gender (F/M), age, professional background (primarily design-related fields), pet ownership experience (years or N/A), and a summary of their animal interaction habits (e.g., daily care, occasional observation).
Figure 6. Co-design workshop.(a) Participants co-creating with MojiKit during the workshop sessions. (b) Examples of affective interaction patterns co-created by participants, including (i) face-to-face nuzzling, (ii) emotional embracing, (iii) face-cupping gesture, and (iv) resting on the shoulder. Photos are used with informed consent. A two-panel figure from co-creation workshops. (a) Four example groups of participants collaboratively designing with MojiKit around a table, using plush toys in a well-lit indoor environment. (b) Four affective interaction pattern examples developed during the workshop: (i) face-to-face nuzzling, (ii) emotional embracing, (iii) face-cupping gesture, and (iv) resting on the shoulder, each shown with participant photos (faces blurred for privacy) and corresponding MomoBot robotic motion demonstrations.
Figure 7. Quantitative evaluation results of MojiKit. (a) Creativity Support Index (CSI) ( Cherry and Latulipe, 2014 ) assessing six dimensions of creative support, (b) Design Reference Cards evaluation focusing on efficiency, inspiration, and clarity, and (c) MomoBot & Behavior Control Studio usability scale covering feedback comprehension, expressiveness, creative exploration, and adaptability. Three horizontal stacked bar charts comparing quantitative evaluation results across three HCI systems. (a) CSI User Experience Scale: Six metrics (Enjoyment, Exploration, Expressiveness, Immersion, Collaboration, and Results Worth Effort) show score distributions (1-5), with Collaboration having the highest 5-score proportion (69.4%, dark green). (b) Design Reference Cards: Three metrics (Information Retrieval, Inspiration, Comprehensibility), where Comprehensibility achieves the highest 5-score rate (55.6%). (c) MomoBot & Control Studio: Four metrics (Feedback, Expression, Creativity, Adaptability), with Adaptability showing 55.6% at 5-score. All charts use a green-to-yellow color gradient for scores (1=light yellow, 5=dark green) with percentage labels.
Category
Count (n)
Percentage (%)
Examples
Human Intent
Greeting / Reunion
7
20.0%
Running to greet; playful salutation
Affection / Comfort
7
20.0%
Sleeping together; cuddling; soothing
Play / Teasing
6
17.1%
Toy play; playful provocation
Attention-seeking
5
14.3%
Responding to name; seeking attention
Training / Instruction
3
8.6%
Sit command; learning tricks
Table 2. Summary statistics of 35 validated interaction patterns collected from the workshop. A four-section statistical table summarizing 35 validated human-robot interaction patterns from co-design workshops. Section 1 (Human Intent): Eight categories including Greeting/Reunion (20.0%), Affection/Comfort (20.0%), and Play/Teasing (17.1%) with behavioral examples. Section 2 (Trigger Types): Four trigger categories dominated by Human Action (60.0%) including speech/touch/gesture. Section 3 (Robot Behavior): Eight response types including Head-turn/Nodding (20.0%), Approach/Move closer (17.1%), and Tail-wagging (14.3%). Section 4 (Affective Meaning): Five affect categories with Positive-Seeking behaviors (57.1%) as most prevalent.
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 8. Human-Pet Interaction Process Analysis Series Cards: Human Interaction Intentions, Human Behavioural Activities, Human Interpretive Processes. Figure showing three ``Human-centric'' cards from the Human-Pet Interaction Process Analysis Series Cards. The cards are stacked vertically. The top card is titled Human Interaction Intentions and organizes human-side intentions around interacting with animals into multiple labeled sections with example bullet points. The middle card is titled Human Behavioural Activities and categorizes observable human behaviors during human–animal situations (e.g., direct contact, non-contact, object-mediated, vocal, and nonverbal behaviors) in several column-like panels. The bottom card is titled Human Interpretive Processes and summarizes how people interpret animals’ signals, grouping interpretations such as inferred traits/abilities, relationship perception, motivation, and emotional state. All three cards share a consistent visual layout with headings and boxed lists.
Figure 9. Human-Pet Interaction Process Analysis Series Cards: Environment. The figure shows an ``Environment'' card from the Human–Pet Interaction Process Analysis Series Cards. The card focuses on environmental factors hypothesized to affect animals during interaction. It presents a structured list of environmental factors organized into three sections: environmental acoustics (such as human voices, household sounds, and outdoor noise), others or other animals (such as family members moving, vocalizations, or strangers approaching), and external light and shadow (such as unexpected changes in lighting). The information is arranged in labeled columns with example items listed under each category.
Figure 10. Human-Pet Interaction Process Analysis Series Cards: Animal Behavior and Emotional Meaning - A Kitten’s Guide, Animal Emotions and Corresponding Behaviors. The figure shows two ``Animal (Cat)'' cards from the Human–Pet Interaction Process Analysis Series Cards, displayed one above the other. The top card is titled Animal Behavior and Emotional Meaning: A Kitten’s Guide and organizes common kitten body cues and postures by body parts, including body posture, ears, eyes, tail, limbs, whiskers, mouth, head, and composite motion, with example behaviors and associated emotional meanings listed in labeled boxes. The bottom card is titled Animal Emotions and Corresponding Behaviors: A Kitten’s Guide and maps emotional states (such as fear, play/joy, pain, contentment, and anger) to observable kitten gestures across body parts and behaviors. Both cards use structured panels, headings, and bullet lists to present relationships between animal behaviors and emotional interpretations.
Figure 11. Human-Pet Interaction Process Analysis Series Cards: Animal Behavior and Emotional Meaning - A Puppy’s Guide, Animal Emotions and Corresponding Behaviors. The figure shows two ``Animal (Dog)'' cards related to puppies from the Human–Pet Interaction Process Analysis Series Cards, displayed one above the other. The top card is titled Animal Behavior and Emotional Meaning: A Puppy’s Guide and organizes common puppy body cues and actions by body parts, including face, eyes, muzzle, ears, tail, limbs, body posture, and actions, with example behaviors and associated emotional meanings listed in labeled sections. The bottom card is titled Animal Emotions and Corresponding Behaviors: A Puppy’s Guide and maps emotional states (such as fear, pain or distress, submission or anxiety, joy or play, anger or aggression, contentment, and anticipation) to observable puppy behaviors across body parts and actions. Both cards use structured panels, headings, and bullet lists to present relationships between animal behaviors and emotional interpretations.
ID
Human Intent
Robot Behavior / Feedback
Robot Affective Meaning
G1-1
Attract attention
Head-turn toward sound, return after confirmation
Responsive, attentive
G1-2
Reject unwanted touch
Head shaking, trunk movement, leaving
Impatient, avoiding
G1-3
Desire for petting
Roll onto back, chin raised, C-shape posture
Feeling needed, happy
G1-4
Reward
Rush forward, cling with paws, licking
Affectionate gift, pampered
G2-1
Respond to call
Tail wagging, walking over
Happy, eager to play
G2-2
Companionship
Approaching, ear and head movement
Relaxed, positive
Appendix
Table 3. Condensed summary of 35 validated interaction patterns, showing intent, robot behavior, and affective meaning. A comprehensive longtable summarizing 35 validated human-robot interaction patterns organized by participant group (G1-G9) and pattern ID. Three main columns show: (1) Human Intent categories (e.g., Attract attention, Desire for petting, Training, Companionship), (2) Corresponding Robot Behaviors (e.g., Head-turn toward sound, Rolling onto back, Tail wagging, Lying beside), and (3) Robot Affective Meanings (e.g., Responsive, Happy, Excited, Affectionate). Patterns demonstrate diverse interaction contexts, including greeting, play, discipline, and companionship.
Large language models (LLMs) are increasingly used for verbal interaction in social robots, yet prompt design in human-robot interaction (HRI) remains underspecified. As a result, robots may present hallucinated capabilities, unclear behavioural boundaries, and misleading personas. This paper develops a framework for prompt design in LLM-based robots and introduces a structured prompt template comprising eight functional components through which robot behaviour can be specified, bounded, and adapted. The framework is grounded in a review of prior LLM-based HRI work and complemented by survey and discussion data from HRI experts gathered at the Robo-Identity workshop at IEEE RO-MAN 2025 (N=27). The qualitative findings highlight limited legibility of robot personality, the need for user adaptation, and strong ethical concerns about safety, deception, and governance. Based on these findings, we present prompting guidelines accompanied by proof-of-concept template as a structured design and reporting aid for HRI research. We argue that prompt design should be treated as a socio-technical problem rather than a minor implementation detail, requiring explicit capability boundaries, transparent behavioural assumptions, and context-sensitive safeguards to support reliable and interpretable HRI.
Ashita Ashok, Franziska Babel, Patrick Holthaus +7
Robotics Research Lab, RPTU Kaiserslautern, Kaiserslautern, Germany · Human-Centered Systems, Linköping University, Linköping, Sweden · Robotics Research Group, University of Hertfordshire, Hatfield, UK +7
When designing robots' nonverbal behaviors, many researchers have turned to arts-based insights, such as Disney's Animation Principles. Yet, while these principles bear key insights into the design of like-life characters, their application to robot design is inherently limited, in part because animation is not constrained by real-world physics, and in part because animation principles focus on low level animation mechanics and not high-level design considerations for physically embodied, interactive characters. In contrast, little attention has been paid to art forms like puppetry, despite their long history of exploring morphological, behavior, and interaction design of physically embodied, interactive characters. As such, in this work we leverage puppetry texts and practicing puppeteers' expert knowledge knowledge to derive a set of puppetry principles with key insights for robot design. As we show, these insights go beyond -- and uniquely complement -- the prior insights provided by theater, dance, and animation.
Claire Lewis, Sawyer Collins, Alyssa Hanson +2
Colorado School of Mines, Golden, Colorado, USA · California Sate University San Bernadino, San Bernadino, California, USA
Design-oriented HRI is increasingly interested in robots as long-term companions, yet many designs still assume a fixed form and a stable set of functions. We present an ongoing design research program that treats modularity as a designerly medium - a way to make long-term human-robot relationships discussable and material through co-design. Across a series of lifespan-oriented co-design activities, participants repeatedly reconfigured the same robot for different life stages, using modular parts to express changing needs, values, and roles. From these outcomes, we articulate PAS (Personalization-Adaptability-Sustainability) as a human-centered lens on how people enact modularity in practice: configuring for self-expression, adapting across transitions, and sustaining robots through repair, reuse, and continuity. We then sketch next steps toward a fabrication-aware, community-extensible modular platform and propose evaluation criteria for designerly HRI work that prioritize expressive adequacy, lifespan plausibility, repairability-in-use, and responsible stewardship - not only usability or performance.
Lingyun Chen, Qing Xiao, Zitao Zhang +2
Luddy School of Informatics, Computing, and Engineering, Indiana University Bloomington, Bloomington, IN, USA · Human-Computer Interaction Institute, Carnegie Mellon University, Pittsburgh, PA, USA