Artists' Views on Robotics Involvement in Painting Productions

Technical Summary: Artists' Views on Robotics Involvement in Painting Productions

Problem Statement
As robotic technologies evolve, their integration into artistic creation presents a significant area of inquiry. While existing research has extensively explored what robots can contribute to art—ranging from algorithmically generated paintings to autonomous brushstrokes—there is a paucity of empirical data regarding how professional artists experience creating with these systems. Specifically, little is known about the lived experience of artists engaging in long-term, co-creative interactions with autonomous robotic agents compared to human partners. This study addresses the gap in understanding the social, cognitive, and creative dynamics of human-robot collaboration (HRC) within the specific context of abstract painting.

Methodology
The study employed a longitudinal, within-subjects experimental design involving eight professional abstract artists. The research protocol consisted of the following components:

• Participants: Eight professional artists (ranging in experience from 12 to 50+ years) were recruited from the Ann Arbor and Detroit metropolitan areas.
• Experimental Conditions: Each participant engaged in six painting sessions over three weeks. The sessions were divided into two conditions:
  1. Human-Human (H-H): Three sessions collaborating with another human artist.
  2. Human-Robot (H-R): Three sessions collaborating with an autonomous robotic arm.
  • The order was fixed (H-H followed by H-R) with at least one day of separation between sessions to allow for reflection.
• Robotic System: The study utilized a UF850 6-DOF robotic arm mounted on a vertical easel. The system architecture, built on Robot Operating System (ROS) 2, included:
  • Hardware: A custom 3D-printed palette, water container, and cleaning rag holders. A mechanical brush-swapping mechanism (V3.1) inspired by retractable pens allowed for tool-free brush changes (though changes were manually initiated by the artist in this study).
  • Control & Interface: Artists interacted via voice commands ("start," "pause") processed through AssemblyAI's speech-to-text API and a zero-shot classification model. A Tkinter-based visual interface provided real-time status feedback.
  • Artistic Framework: The robot was programmed to execute pre-generated stroke sequences based on an abstract artistic framework inspired by Bruno Munari and Vasilij Kandinskij. The robot drew shapes (circles, squares, triangles) associated with specific colors (blues, reds, yellows).
  • Stroke Generation: Strokes were not generated in real-time. Instead, the system used the CoFRIDA framework and an InstructPix2Pix model (fine-tuned on a Kaggle abstract art dataset) to generate target images. The FRIDA image-to-action planning process then converted these targets into discrete 3D waypoint sequences for the robot.
• Data Collection: Following the six sessions, participants underwent semi-structured exit interviews. These interviews explored the nature of interaction, perceived agency, creative flow, and social engagement.
• Analysis: The interview transcripts were analyzed using reflexive thematic analysis to identify emergent patterns across the eight participants.

Key Contributions
This work offers one of the first empirical investigations into the lived experiences of professional artists collaborating with an autonomous painting robot. The study contributes:

1. Comparative Insight: A direct comparison of H-H and H-R collaborative dynamics, highlighting distinct experiential differences.
2. Stakeholder Perspective: A focus on the perspectives of domain experts (artists) as key stakeholders in the development of robotic systems for the arts, moving beyond end-user functionality to address social dynamics and creative agency.
3. Longitudinal Engagement: An approach that prioritizes repeated, long-term interaction over one-off encounters, allowing for a deeper understanding of how artists adapt to and interpret robotic behavior over time.
4. Multidisciplinary Framework: The integration of robotics, artistic practice, and qualitative research methods to study human-robot co-creation.

Results
The thematic analysis revealed five primary themes:

1. Comparing Collaboration Types: H-H interactions were described as intuitive, dialogic, emotionally engaging, and socially rich, characterized by mutual adaptation and negotiation. In contrast, H-R interactions were perceived as more reflective, playful, and self-directed. Artists reported feeling greater creative freedom with the robot due to the absence of emotional demands and social pressure, though this came with a sense of "sterility."
2. Perceived Agency: Artists oscillated between viewing the robot as a passive tool and an autonomous collaborator. Many expressed a desire for "mixed-initiative" systems where the robot could also take the lead or offer suggestions. Some framed the robot's behavior metaphorically, comparing it to a child painting or a cart on a track, noting its technical limitations but also its non-judgmental nature.
3. Creative Flow: The robot's impact on creative flow was multifaceted. For some, the robot disrupted their rhythm, causing frustration due to a lack of communication or predictability. For others, the robot acted as a catalyst, introducing constraints that forced new strategies and inspired unexpected aesthetic decisions. The robot was often seen as a "complementary person" that disrupted habitual patterns.
4. Social Interaction: Participants expressed mixed feelings about verbal interaction; some welcomed the idea of a talking robot, while others found it unsettling. There was a consistent desire for multimodal communication (e.g., gestures, pointing) to enhance collaboration.
5. Painting Techniques: Artists critiqued the robot's limited expressive range, noting its repetitive, predictable movements and lack of organic variation (e.g., pressure sensitivity, continuous lines). However, these limitations also sparked imagination, with artists envisioning future capabilities such as modular tool attachments (knives, airbrushes), adaptive computer vision, and the ability to emulate specific artistic styles.

Significance and Claims
The paper claims that while human-robot collaboration lacks the emotional reciprocity and social richness of human-human interaction, it serves as a productive space for reflection, inspiration, and experimentation. The study posits that artists are not passive users but active meaning-makers who construct narratives around the robot's actions, adapting their practice to incorporate robotic gestures.

The authors emphasize that their findings highlight the value of long-term engagement and a multidisciplinary approach in HRI. By foregrounding the desires and needs of artistic stakeholders, the research aims to inform the future design of robotic systems that are more attuned to the nuances of creative expression. The work concludes that the "game" or "fun" aspect of H-R collaboration, alongside its technical limitations, creates a unique environment where constraint can become a generative force for creativity.