💻 computer science

Why That Robot? A Qualitative Analysis of Justification Strategies for Robot Color Selection Across Occupational Contexts

This paper qualitatively analyzes over 4,000 user justifications for robot color selection across occupational contexts, revealing that while functional reasoning is dominant, users often unconsciously align their choices with racial and occupational stereotypes, a bias that shifts toward de-racialized machine-centric reasoning as robots become more anthropomorphic.

Original authors: Jiangen He, Wanqi Zhang, Jessica K. Barfield

Published 2026-04-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jiangen He, Wanqi Zhang, Jessica K. Barfield

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are hiring a new employee for a job, but instead of a human, you are picking a robot. You have a menu of options: some look like shiny metal, some look like they have human skin, and those with skin come in different shades (light, medium, brown, dark).

This paper is like a detective story that asks: "Why did you pick that specific robot?"

The researchers didn't just look at which robot people picked; they analyzed thousands of written explanations to understand the reasons behind the choices. They found that while people think they are making logical, practical decisions, their choices are often secretly guided by the same racial stereotypes we have about real humans.

Here is the breakdown of their findings using simple analogies:

1. The "Excuse Maker" (Functionalism)

The most common reason people gave for their choice was practicality. They said things like, "I picked the brown robot because it won't show dirt," or "I picked the white robot because it looks clean."

The Metaphor: Think of this as a suit of armor. People put on a "practical" excuse to hide their true feelings.
The Reality: The researchers found that people were using "practicality" as a cover story.
- When picking a robot for a construction site, people chose darker colors saying, "It hides the dirt."
- When picking a robot for a hospital, people chose lighter colors saying, "It looks sterile."
- The Twist: These reasons perfectly matched our real-world stereotypes about race (e.g., associating darker skin with manual labor and lighter skin with professional/clean environments). The "practical" reason was just a polite way of saying, "This robot looks like the type of person we expect to see doing this job."

2. The "Prime Time" Effect

The study also tested what happens when you show people a picture of a human worker before they pick a robot.

The Experiment: If you show a picture of a Latino construction worker, do people pick a brown robot? If you show a Black athlete, do they pick a dark robot?
The Result: Yes. Seeing the human "primed" (prepared) their brains to pick the matching robot color.
The Magic Trick: Even though their choices changed based on the picture, their explanations did not. They still used the same "practical" excuses ("It hides dirt," "It looks strong"). It's like a magician changing the card in your hand while you insist you still picked the same card. The bias happened in the background, but the explanation stayed the same.

3. The "Mirror" That Doesn't Reflect Everyone

Usually, we think people like things that look like them (like seeing a mirror image). The study found this is true for some groups but not all.

White and Asian participants: When they picked a robot that matched their skin tone, they said, "I feel a connection," or "It feels friendly." They used emotional reasons.
Black participants: When they picked a robot that matched their skin tone, they did not say, "It feels like me." Instead, they doubled down on practical reasons: "It looks strong," "It looks capable."
The Analogy: Imagine a mirror that only reflects warmth for some people. For Black participants, the "mirror" didn't trigger feelings of warmth or friendship; it triggered a need to prove competence. This suggests that in a society with deep-seated biases, seeing someone who looks like you doesn't always feel like a hug; sometimes it feels like you have to prove you belong.

4. The "Uncanny Valley" of Race

The study looked at how "human-like" the robots were.

Low Human-Likeness: If the robot looks like a toaster or a metal box, people feel safe picking any color. They say, "I picked silver because it looks like metal."
High Human-Likeness: As the robot starts to look more like a real human (with a face and skin), people get nervous about race.
The Reaction: When the robot looks too human, people often panic and switch to "Machine-Centric" colors (like silver or teal) to avoid the race issue entirely.
The Metaphor: It's like trying to wear a costume. If the costume is a simple hat, you don't worry. But if the costume is a full-body suit that looks exactly like a person, you suddenly worry about whether you are "getting the race right." To avoid the awkwardness, people just put on a silver suit instead.

The Big Takeaway

The paper concludes that designing robots is not just about aesthetics; it's about ethics.

If we let people pick robot colors based on their "gut feelings" and "practical excuses," we will accidentally build a workforce of robots that reinforces old, unfair stereotypes (e.g., all the construction robots are brown, all the doctors are white).

The Lesson for Designers:

Don't trust "practical" requests blindly. When someone says, "I need a brown robot to hide dirt," ask if they are actually relying on a stereotype.
Be careful with "human" skin. Giving a highly realistic robot a specific skin tone is a loaded decision. Sometimes, a "neutral" metal finish is the only way to avoid racial bias.
One size does not fit all. Assuming that everyone will feel a warm connection to a robot that looks like them is wrong. Systemic bias changes how different groups interact with these machines.

In short: We can't just paint robots and hope for the best. We have to paint them with intention, knowing that every color carries a history.

1. Problem Statement

As robots increasingly integrate into the workforce, Human-Robot Interaction (HRI) faces a critical challenge: implicit social biases influence how users select and perceive robots. While previous research has documented that users exhibit racial biases in robot selection (e.g., preferring lighter-skinned robots for healthcare and darker-skinned robots for construction), there was a significant gap in understanding why and how users rationalize these choices.

The core problem addressed is the disconnect between selection behavior and explicit justification. Users often select robots based on unconscious racial stereotypes but provide "rational," task-based explanations that mask these biases. The study aims to uncover the underlying reasoning frameworks, the role of situational factors (task context, racial priming), and how demographic backgrounds and robot anthropomorphism modulate these justifications.

2. Methodology

The study employed a mixed-methods qualitative approach, analyzing data from two prior experiments (Study 1 and Study 2) involving 1,038 participants and 4,146 open-ended justifications.

Experimental Design:
- Contexts: Four professional scenarios: Construction, Hospital, Home Tutoring, and Sports.
- Variables:
  - Robot Appearance: Six color options (Light, Medium, Brown, Dark skin tones; Silver, Teal non-skin tones) across five levels of human-likeness.
  - Priming (Study 2): Racial priming was manipulated by embedding images of human professionals in the task scenarios. "Stereotype-consistent" primes (e.g., Latino in construction) were contrasted with "Stereotype-inconsistent" primes (e.g., White in construction).
- Task: Participants selected a robot for a specific scenario and provided an open-ended written justification for their choice.
Data Analysis Pipeline:
1. Codebook Development: An inductive thematic approach was used to create a multidimensional coding scheme. The process involved stratified sampling, open coding, and axial coding to reach theoretical saturation.
2. AI-Assisted Coding: Given the dataset size, the authors utilized Google's Gemini API to classify responses into a structured schema.
3. Validation: The AI coding was validated against human coders (three independent raters). The inter-rater reliability between human consensus and AI achieved a Cohen's $\kappa$ of 0.73 at the main category level, confirming the reliability of the automated analysis.
4. Statistical & Thematic Analysis: The study combined frequency distributions, Pearson's $\chi^2$ tests, and Cramér's $V$ effect sizes with targeted thematic analysis of specific sub-categories.
Coding Scheme (5 Main Categories):
- C1. Functionalism: Task-fit, durability, cleanliness, safety.
- C2. Psych/Affective: Approachability, trust, engagement.
- C3. Machine-Centric: Neutrality, de-racialization, materiality.
- C4. Preference/Evasion: Personal aesthetic, random choice.
- C5. Identity/Social: Stereotype alignment, identity projection.

3. Key Contributions

Mapping Justification Frameworks: The study provides the first comprehensive taxonomy of how users rationalize robot color selection, revealing that Functionalism is the dominant strategy (52.2%) but is often used as a post-hoc rationalization for bias.
Unmasking "Aversive Racism" in HRI: It demonstrates that bias often operates beneath conscious rationalization. Users adapt practical rationales (e.g., "hiding dirt," "sterility") to align with established racial and occupational stereotypes, effectively masking the bias under the guise of utility.
Intersectional Analysis of Demographics: The research challenges the universality of "racial mirroring." It reveals that while White and Asian participants exhibit affective resonance with color-congruent robots, Black participants predominantly rely on Functionalism rather than affective warmth when selecting dark-skinned robots, suggesting systemic biases alter the nature of identity projection.
The "Anthropomorphism Threshold": The study identifies a critical design tension where increasing human-likeness causes users to retreat from functional reasoning toward Machine-Centric de-racialization (choosing metallic/non-skin tones) to avoid the ethical complexity of assigning race to highly human-like agents.

4. Key Results

Dominance of Functionalism: 52.2% of all justifications fell under Functionalism. However, the content of these justifications shifted based on the robot's color and the task:
- Construction: Darker robots were justified by "durability" and "hiding dirt."
- Hospital: Lighter robots were justified by "cleanliness" and "sterility."
- Sports: Darker robots were linked to "athleticism" and "strength."
Impact of Racial Priming: While racial priming significantly shifted which robot was chosen (aligning with stereotypes), it had a weak effect on the type of justification used. Users maintained standard affective or task-related reasoning even when their choices were clearly stereotype-driven.
Demographic Interactions:
- Black/African American participants showed the highest rates of Functionalism (59.6%) and the lowest Machine-Centric reasoning. When selecting Dark robots, they justified choices based on competence/performance rather than affective connection.
- White and Asian participants showed elevated Psych/Affective and Preference reasoning when selecting robots matching their own skin tone (Light and Medium, respectively).
Effect of Human-Likeness: As robots became more anthropomorphic (Level 1 to Level 5):
- Functionalism decreased (54.5% $\to$ 44.3%).
- Machine-Centric reasoning nearly doubled (8.6% $\to$ 15.8%), driven by a desire for "Neutrality" and "De-racialization."
- Users increasingly avoided skin-tone colors for highly human-like robots, opting for metallic finishes to "erase" race.

5. Significance and Design Implications

The paper argues that robot appearance is not merely an aesthetic choice but a socio-technical intervention with ethical consequences.

Critique of "Functional" Requests: Designers must critically interrogate requests for "practical" colors (e.g., brown for construction), as these often mask implicit biases. Accepting them blindly risks hardcoding societal prejudices into the workforce.
The Anthropomorphism Threshold: Assigning human skin tones to high-fidelity robots inevitably triggers racial categorization. To achieve a truly neutral agent, designers should consider explicitly non-human finishes (e.g., metallic silver) rather than relying on skin tones.
Beyond "One-Size-Fits-All" Customization: Simply offering diverse skin tones (like digital emojis) does not guarantee universal trust or affective bonding. Systemic biases mean that marginalized groups (specifically Black users in this study) may not experience the same "affective mirroring" as dominant groups. Design strategies must account for the socio-historical context in which the robot operates.

Conclusion: The study reveals that users frequently use functional reasoning to rationalize bias. As robots become more human-like, users experience "racial anxiety," leading them to de-racialize the machine. Future HRI design must move beyond surface-level customization to address the deep-seated structural biases that influence human-robot selection.