Robot Perception Analysis
"Drag green dots on the robots that look helpful. Drag red dots on the robots that look creepy"
Executive Summary
Participants do not evaluate robots along a friendly-threatening axis. Instead, they operate with a functional transparency heuristic: robots that look like tools register as helpful, while robots that look like beings register as creepy—regardless of how gentle or approachable those beings appear. The decisive variable is not friendliness but categorical clarity.
Resonance Array Analysis
Twelve images met the 60% positive threshold. These images share a remarkably consistent visual grammar that reveals what "helpful" actually means in participant mental models.
Dominant Visual Motifs
The resonance array is dominated by articulated arms—industrial robot arms, whether in factory settings or demonstration contexts. Seven of twelve positive images feature arm-based robots. The remaining positive images share a critical feature: the face (when present) is obviously non-human. The Sunday Robotics "Mamo" robot, with its retro-futuristic rounded head and simple camera-eyes, scores 92% helpful while serving food in a domestic setting. Its face reads as "appliance" rather than "entity."
Color Palette & Material Language
The resonance palette is emphatically industrial: white, gray, black, and metallic silver. Orange appears only as accent color or safety marking. Surfaces are matte or brushed metal—never glossy white plastic attempting to simulate skin. Exposed cables, visible joints, and mechanical articulation are features, not bugs. The Dyna robots (100% resonance) proudly display their servo motors and wiring.
Context & Activity
Positive robots are shown doing things that matter: folding laundry, picking packages, writing on whiteboards, opening drawers. The activity is always instrumental—object manipulation rather than social interaction. Critically, no human appears in the frame alongside resonant robots, or humans are clearly in a supervisory/demonstration context.
Resistance Array Analysis
Thirteen images met the 60% negative threshold. The pattern is stark: these are robots attempting human form.
The Humanoid Problem
Every robot in the resistance array is either fully humanoid or pseudo-humanoid. This includes Boston Dynamics Atlas (100% creepy), Figure Robotics units (81-89% creepy), Toyota Partner Robot playing violin (66.67% creepy), and iCub child-faced robots (81.82% creepy). The quadruped dog-robots (Unitree, 76.92% creepy) extend the pattern—animal-form is as problematic as human-form.
The Fabric Face Failure
A specific design choice proves catastrophically unsuccessful: robots with fabric-covered heads and minimal facial features. The 1X Robotics units with their white fabric faces and simple dot-eyes score 75-100% creepy across multiple images. This design intent—presumably to appear soft and non-threatening—achieves the opposite.
The Context Paradox
Domestic settings and caregiving activities make humanoid robots more creepy, not less. The Figure Robotics unit watering plants in a home bathroom scores 81.82% creepy. The black-suited Figure unit serving drinks at a social gathering scores 88.89% creepy. The 1X robot being admired by an elderly person scores 87.50% creepy. Placing human-form robots in human contexts amplifies their categorical ambiguity.
Neutral Array Analysis
Only one image landed in the neutral zone: the X1 Robotics product lineup showing three humanoid robots with fabric faces in gray, white, and black variants (58.33% positive, 41.67% negative). This near-even split is instructive.
Comparative Analysis
| Dimension | Resonance (Helpful) | Resistance (Creepy) |
|---|---|---|
| Form Factor | Arm-based, non-humanoid, or clearly robotic humanoid | Full humanoid, pseudo-humanoid, animal-form |
| Face Type | No face, camera lens, retro-robot aesthetic | Fabric mask, screen face on body, smooth helmet |
| Materials | Visible metal, exposed mechanisms, industrial finish | Smooth white plastic, fabric covering, skin-like surfaces |
| Activity | Object manipulation, task completion, demonstration | Social presence, caregiving, performance, "being there" |
| Setting | Industrial, trade show, technical documentation, warehouse | Domestic interior, social gathering, natural environment |
| Human Presence | None, or clearly supervisory | Recipient of care, social interaction partner |
| Design Register | "Serious equipment" | "Gentle companion" |
The Underlying Mental Model
Participants operate with an implicit categorical schema: tools vs. beings. Robots that signal "I am a sophisticated tool" are welcomed as helpful. Robots that signal "I am a kind of person" are rejected as creepy. This explains why industrial contexts help (they frame robots as equipment), why domestic contexts hurt humanoids (they place being-like forms in person-inhabited spaces), and why the fabric-face disaster occurs (attempting to make a face "soft" is precisely an attempt to make it more being-like).
Strategic Implications
✓ Design Toward
- Arm-based or modular form factors for task-specific robots
- Visible mechanical elements (joints, servos, cables)
- Camera/sensor assemblies as "face" rather than anthropomorphic features
- Industrial material language: brushed metal, matte finish
- Technical documentation aesthetic in marketing
- Task demonstration contexts showing completed work
- Retro-futurist design references if humanoid form is required
✗ Design Away From
- Fabric-covered heads or bodies
- Smooth white plastic attempting skin-like appearance
- Minimal facial features on humanoid forms
- Domestic interior photography with humanoid robots
- Caregiving or social interaction scenarios
- "Gentle companion" aesthetic positioning
- Quadruped/animal-form robots for non-specialized uses
Contextual Guidance
For industrial/B2B applications: Lean fully into the tool aesthetic. Exposed mechanisms, industrial finish, and task-completion imagery will reinforce the helpful reading. Avoid any anthropomorphization.
For consumer/domestic applications: The challenge is substantial. Arm-based or modular robots have the clearest path to positive perception. If humanoid form is required for functional reasons, consider the "Mamo" strategy: explicitly retro aesthetic, visible camera as face, avoiding any attempt at realistic human features.
For marketing/communications: Show robots doing things, not being present. Task completion beats social context. Technical documentation aesthetic outperforms lifestyle photography.
Methodology Note
This analysis is based on n=21 survey completions using dot-drag methodology with a 60% consensus threshold for array classification. This sample size provides directional insight with moderate confidence. Patterns identified are consistent and internally coherent, supporting their validity, but should be treated as hypotheses warranting validation rather than definitive conclusions.
Recommended next steps: (1) Larger-sample validation study; (2) Demographic segmentation to identify variation by age, tech familiarity, or cultural background; (3) A/B testing of specific design variables to isolate causal factors.