POIROT: Investigating Direct Tangible vs. Digitally Mediated Interaction and Attitude Moderation in Multi-party Murder Mystery Games

This study challenges the assumption that physical robot interaction universally enhances user experience by demonstrating that while tangible delivery does not inherently improve engagement, it significantly reduces narrative immersion for individuals with high negative attitudes toward robots, who instead benefit from digitally mediated interfaces as a social buffer.

The Gist

Imagine you are at a dinner party playing a murder mystery game. The host is a robot named POIROT (named after the famous detective). Its job is to hand out clues to help you solve the crime.

The researchers wanted to know: Does it matter how the robot gives you the clues?

They tested two ways:

1. The "Handshake" Method: The robot physically rolls over to you, stops right in front of your face, and places a physical card into your hand.
2. The "Screen" Method: The robot stays put, and the clues appear on a tablet screen in front of you.

The Big Surprise: One Size Does Not Fit All

You might think, "Obviously, a robot handing you a real card is cooler and more immersive!" But the study found that it depends entirely on who you are.

The researchers discovered a "personality switch" called NARS (Negative Attitudes towards Robots). Think of this as a "Robot Anxiety Meter."

• Low Anxiety (The "Robot Lovers"): If you are comfortable with robots, the "Handshake" method was just fine. It felt fun, ritualistic, and cool. The robot felt like a fun part of the game.
• High Anxiety (The "Robot Skeptics"): If you are naturally nervous around robots, the "Handshake" method was a disaster. It ruined the fun.

The Analogy: The Overly Enthusiastic Waiter

To understand why, imagine two waiters at a restaurant:

• Waiter A (The Robot Lovers): You love interacting with people. When Waiter A comes right up to your table, leans in, and hands you your menu with a smile, you think, "Wow, great service! I feel special!"
• Waiter B (The Robot Skeptics): You are shy and hate being crowded. When Waiter B comes right up to your table, leans in, and shoves the menu into your hand, you feel trapped. You think, "Get back! I can't breathe! This is awkward!"

In this study, the robot was Waiter B for the anxious people.

• The Problem: For people with high robot anxiety, the robot moving close to their face and physically handing them a card felt intrusive. It broke their "bubble" of personal space. They felt forced to pay attention to the robot, which made them anxious and pulled them out of the story.
• The Solution: For these same people, the Tablet (Screen) was like a waiter who just slides the menu across the table from a distance. It gave them a "social buffer." They could get the information without the robot getting in their personal space. It felt safer and less stressful.

The "Tax" of Tangibility

The study found that physical interaction has a "tax."

• The Bonus: It feels novel and magical (like a real magic trick).
• The Tax: It requires the robot to move, speak, and interact in real-time. If the robot is a little slow or sounds a bit robotic, it becomes annoying.

For people who love robots, the Bonus outweighs the Tax.
For people who are nervous around robots, the Tax is too heavy. The "magic" feels more like a "burden."

The Takeaway for the Future

The main lesson is that robots shouldn't try to be the same for everyone.

If a robot is going to be a game master, a teacher, or a helper in a hospital, it needs to be adaptive.

• If you detect that a user is nervous (high anxiety), the robot should switch to a digital mode (screens, text) to give them space.
• If the user is excited and comfortable, the robot can switch to physical mode (handing things over, moving close) to make the experience more magical.

In short: Don't force a handshake on someone who just wants a text message. The best robot is the one that knows when to get close and when to stay back.

Technical Summary

Here is a detailed technical summary of the paper "POIROT: Investigating Direct Tangible vs. Digitally Mediated Interaction and Attitude Moderation in Multi-party Murder Mystery Games."

1. Problem Statement

The integration of social robots as Game Masters (GMs) in complex, multi-party social deduction games (like Murder Mystery Games or MMGs) presents a critical design challenge. While physical embodiment is often assumed to universally enhance user engagement and immersion, this "one-size-fits-all" view lacks empirical validation in high-stakes, co-located group settings.

The core problem addressed is determining how the modality of interaction (direct tangible delivery vs. digitally mediated delivery) affects user experience in a group context, and whether individual differences—specifically Negative Attitudes towards Robots (NARS)—moderate these effects. The authors hypothesize that while tangible interaction offers novelty, it may impose excessive "proxemic friction" (social pressure and spatial intrusion) for users with high robot anxiety, potentially degrading their experience compared to a buffered digital interface.

2. Methodology

Experimental Design

• Type: Between-subjects experiment ($N=67$).
• Setting: A custom-built murder mystery game script titled Judgment Quartet, played in a living-room-style laboratory.
• Conditions:
  1. Physical-Card Condition: The robot (POIROT) physically navigates to each player, verbally announces the delivery, and dispenses laminated clue cards directly into the player's hand.
  2. In-App Condition: The robot makes identical verbal announcements, but clues are delivered instantly via a custom tablet interface (iPad) accessed by the player.
• Control Variables: The robot's identity, dialogue, narrative pacing, eye animations, and arm gestures were kept constant across both conditions. The only variable was the delivery modality.
• Wizard-of-Oz (WoZ): A human operator tele-operated the robot's movement and card dispensing to ensure precise execution while maintaining the illusion of autonomy.

Apparatus: POIROT

• Hardware: A 50.1 cm tall social robot based on the open-source FishBot platform, featuring a 3D-printed shell, 2-DOF arms, a custom friction-based card dispenser, and an embedded audio pipeline.
• Software: Integrated with an LLM (Xiaozhi) for dialogue management, ASR/TTS for voice interaction, and an iPhone-based eye-tracking module for gaze alignment.
• Companion App: A custom app running on iPads managed room creation, role selection, and script reading. In the In-App condition, it served as the clue delivery mechanism.

Measures

• Independent Variables: Interaction Modality (Physical vs. Digital) and NARS (Negative Attitudes towards Robots Scale).
• Dependent Variables:
  • Narrative Immersion: 15-item scale measuring attentional focus and emotional engagement.
  • Social Presence: 5-item scale measuring the perception of the GM as a tangible, responsive entity.
  • Game Experience: 7-item scale measuring satisfaction and flow.
• Analysis Strategy:
  1. MANOVA: To test for a baseline main effect of modality.
  2. Multilevel Linear Models (MLM): To account for the hierarchical data structure (participants nested within 12 game sessions) and test the interaction between Modality and NARS.
  3. Qualitative Analysis: Thematic coding of open-ended feedback to explain quantitative findings.

3. Key Results

Quantitative Findings

• Main Effect (H1): The baseline MANOVA revealed no significant main effect of delivery modality on the combined dependent variables ($p=0.65$). This refutes the assumption that tangible interaction is universally superior.
• Moderation Effect (H2): The MLM analysis revealed a significant interaction between delivery modality and NARS scores ($p=0.003$ for Narrative Immersion; $p=0.033$ for Game Experience).
  • High NARS Users: Participants with high negative attitudes towards robots experienced significantly lower narrative immersion in the Physical-Card condition compared to the In-App condition. The physical delivery acted as a stressor.
  • Low NARS Users: Participants with low negative attitudes showed no significant difference in immersion or experience between the two conditions.
• Social Presence: No significant interaction was found for Social Presence, suggesting that while immersion was affected, the perception of the robot's "presence" remained stable regardless of modality or attitude.

Qualitative Findings

Three themes emerged from participant feedback:

1. Tangibility as Novelty: For low-NARS users, the physical delivery created a "ritualistic" sense of occasion and enhanced the perceived fairness of the game (the robot as a "pure executor").
2. Interaction Costs & Proxemic Friction: High-NARS users reported that the robot's slow movement, delayed responses, and physical intrusion into personal space disrupted the game flow. They perceived the robot's presence as a source of friction rather than a service.
3. Desire for Social Competence: Participants expressed a need for robots to possess smoother movement and more natural social cues. When the robot failed to meet these expectations, the digital interface served as a necessary "social buffer."

4. Key Contributions

1. Empirical Evidence of Conditional Tangibility: The study challenges the "embodiment is always better" heuristic. It demonstrates that the benefits of tangible interaction are conditional on the user's predisposition. Tangibility provides a novelty bonus for some but incurs a "proxemic tax" for others.
2. Psychological Distance Framework in HRI: The authors apply the concept of psychological distance to explain why digital mediation is beneficial for anxious users. Digital interfaces act as a social buffer, reducing proxemic demands and increasing perceived control, which is crucial for users with high NARS.
3. Design Implications for Adaptive Systems: The paper argues against static interaction designs. It proposes that future social robots in group settings must be adaptive, capable of switching between direct tangible and mediated modalities based on real-time or pre-assessed user anxiety levels (NARS).

5. Significance and Future Directions

• Theoretical Impact: This work advances HRI theory by moving beyond binary comparisons (robot vs. no robot) to investigate the nuances of how interaction is delivered. It highlights that "presence" is not a monolithic construct but is filtered through individual psychological traits.
• Practical Application: For designers of social robots in education, therapy, or entertainment, the findings suggest that a "one-size-fits-all" physical approach may alienate anxious users. Systems should include fallback mechanisms (e.g., switching to a tablet interface) to accommodate diverse user profiles.
• Limitations & Future Work:
  • Cultural Generalizability: The sample was predominantly East Asian; cultural norms regarding proxemics may influence results.
  • Temporal Confounds: The physical delivery took longer (4–5 mins) than the digital delivery (instant). Future work needs to decouple "tangibility" from "latency" to isolate the pure effect of materiality.
  • Long-term Effects: The study was a single-session experiment. Future research should investigate if the "novelty" of tangibility fades over time or if anxiety habituates.

In conclusion, the POIROT study provides a critical boundary condition for HRI design: Direct tangible interaction is not inherently superior; its efficacy is strictly moderated by the user's attitude toward robots. Adaptive systems that respect these individual differences are essential for inclusive and effective human-robot collaboration.