An Intent of Collaboration: On Agencies between Designers and Emerging (Intelligent) Technologies

This paper argues that to maintain creative agency while collaborating with emerging intelligent technologies like LLMs, designers must engage in introspection, develop a structural understanding of the technology's capabilities, and deliberately adjust the human-technology working relationship.

The Gist

Imagine you are a master weaver, an expert in the tactile, messy, and beautiful world of creating fabric by hand. Now, imagine you invite a brand-new, incredibly smart, but slightly clumsy robot assistant to help you weave. You tell it, "Let's create something amazing together!"

This paper is the story of three designers who did exactly that. They tried to "co-create" textiles (like embroidery, crochet, and weaving) with Google's AI chatbot, Bard. What they found wasn't a perfect partnership; it was a messy, frustrating, and ultimately enlightening dance of power.

Here is the story of their journey, broken down into simple concepts:

1. The Setup: The "Magic Wand" vs. The "Real World"

The designers started with high hopes. They thought the AI was like a magic wand that could instantly turn their ideas into perfect patterns. They treated the AI like a genius wizard who knew everything.

• The Reality Check: The AI is actually more like a very well-read librarian who has never touched a piece of thread. It knows a lot about words and pictures, but it has no "body." It doesn't understand that a crochet hook can't bend a certain way, or that a specific fabric feels rough. When the designers asked for complex weaving patterns, the AI gave them instructions that were impossible to follow, or it forgot what it said five minutes ago.

2. The Struggle: The "Tug-of-War" for Control

The core of the paper is about Agency. Think of agency as the "steering wheel" of the creative process. Who is driving the car? The human or the AI?

The designers went through four distinct phases, like levels in a video game:

• Phase 1: The "Yes-Man" Phase. The designers handed over the steering wheel completely. They asked the AI, "What should we make?" and did exactly what it said.
  • Result: The AI gave confusing or wrong advice. The designers felt lost and frustrated because they were following a map that led off a cliff.
• Phase 2: The "Correction" Phase. The designers started to realize, "Wait, this isn't working." They tried to fix the AI's mistakes by repeating themselves or arguing with it.
  • Result: It was exhausting. They were fighting a battle against a machine that couldn't understand the physical world.
• Phase 3: The "Boss" Phase (The Turning Point). The designers finally grabbed the steering wheel back. They stopped asking the AI for ideas and started using it only for tools. They said, "I have the plan; you just help me write the list."
  • Result: Suddenly, the work became fun again. The designers felt creative and empowered. The AI became a helpful assistant rather than a confusing boss.
• Phase 4: The "Burnout" Phase. Eventually, the designers got tired of the back-and-forth. They let the AI take the wheel again, but this time they were just bored. The AI started repeating the same old suggestions, and the magic was gone.

3. The Big Lesson: It's About the Relationship, Not the Robot

The most important thing the paper discovered is that the AI doesn't have a fixed personality or level of intelligence. Its "power" depends entirely on how the human treats it.

Think of the AI like a musical instrument.

• If you treat a guitar like a toaster (expecting it to make toast), it will fail.
• If you treat a guitar like a partner who needs to be tuned and played with specific techniques, it makes beautiful music.

The designers learned that to stay creative, they had to stop expecting the AI to be a "genius partner" and start treating it as a "tool with specific limits." They had to understand their own skills (the weaving, the embroidery) and use the AI only to support those skills, not replace them.

4. The Takeaway for Everyone

This isn't just about designers and robots. It's about how anyone should work with new, powerful technology (like the AI tools we use today).

• Don't surrender your brain: If you let the AI do all the thinking, you will feel frustrated and uncreative.
• Know your own strengths: You are the expert in your craft. The AI is just a helper.
• Adjust the relationship: Sometimes you lead, sometimes you follow, but you must always be the one holding the reins.

In short: The paper argues that to remain a creative human in the age of AI, you have to be the "captain" of the ship. You can't just let the autopilot (the AI) fly the plane, especially if the autopilot doesn't know what a runway looks like. You have to understand the machine, know your own skills, and constantly adjust how you work together to keep the creative spark alive.

Technical Summary

1. Problem Statement

The paper addresses the growing challenge designers face when integrating powerful, emerging intelligent technologies (specifically Large Language Models like Google's Bard) into creative workflows. While these tools promise to enhance creativity, they often lack embodied knowledge and are designed for engineering optimization rather than exploratory design practices.

Key issues identified include:

• Agency Loss: Designers risk losing their creative agency, feeling disempowered, or becoming "non-creative laborers" when AI takes over decision-making.
• Mismatched Expectations: There is a disparity between the designer's need for exploratory, glitch-tolerant processes and the AI's tendency toward solutionist, repetitive, or hallucinatory outputs.
• Lack of Protocols: Unlike traditional tools, emerging AI lacks established interaction protocols, forcing designers to navigate a "black box" without clear guidance on how to delegate or retain control.
• The "More-Than-Human" Gap: Existing HCI research often observes human-AI interaction objectively but fails to capture the introspective cognitive and emotional shifts designers undergo when their creative agency is threatened.

2. Methodology

The study employs a Practice-based Research through Design (RtD) approach, focusing on the process of making rather than the final artifact.

• Context: A semester-long experimental project involving three master's students in Industrial Design (the authors) working in the domain of Digital Craftsmanship (specifically textile making: embroidery, crochet, and weaving).
• Tool: Google's Bard (an LLM) was used as the "intelligent partner" for co-creation.
• Procedure:
  • Duration: 2 months of active making, followed by 12 and 15 months of retrospective analysis.
  • Process: Designers fed images (generated by Midjourney and Canva) into Bard, requested ideation or technical guidance, created textile samples, and documented the interaction.
  • Data Collection: First-person reflection journals, weekly meetings, and conversation logs with Bard.
• Analysis Framework:
  • Three Rounds of Analysis: Immediate post-making analysis, followed by two meta-analyses (12 and 15 months later) involving semi-structured interviews and self-reflection.
  • Key Metric: The introduction of "Agency" as a qualitative, subjective parameter. Designers rated their own agency versus Bard's agency on a fluctuating scale to map the power dynamics.
  • Theoretical Lens: The More-Than-Human (MTH) framework (Bennett, 2010; Wakkary, 2021), viewing the design process as an assemblage of humans, machines, and materials with distributed agency.

3. Key Contributions

The paper proposes a shift from viewing AI as a tool to viewing it as a relational partner where agency is a dynamic, negotiated construct.

• Conceptualization of Relational Agency: The authors argue that the "agency" of an AI is not inherent but delegated by the human based on their expectations. Agency is a relational concept that fluctuates throughout the creative process.
• The Four Phases of Power Dynamics: The study identifies a cyclical pattern in human-AI collaboration:
  1. Wizard Phase: Designers delegate high agency to AI, expecting it to know all (leading to overload and frustration).
  2. Realization Phase: Designers notice AI limitations (repetition, forgetfulness) and begin to improvise prompts.
  3. Reclamation Phase: Designers consciously take the lead, treating AI as an assistant/suggestion engine rather than a co-pilot. This restores creative agency.
  4. Recession Phase: Designers may eventually fatigue and delegate control again, leading to renewed frustration.
• Introspective Methodology: The paper demonstrates that self-reflection and detailed process documentation are critical for understanding the cognitive nuances of human-AI collaboration, which outcome-based evaluations miss.

4. Results

The analysis of the three "Studio Journeys" (Embroidery, Crochet, Weaving) revealed consistent patterns:

• Initial Disempowerment: All designers initially felt frustrated and disempowered because they expected Bard to be a superior creative guide. When Bard failed to provide actionable textile patterns or hallucinated (e.g., misidentifying images), designers felt their creativity was stifled.
• The Turning Point: Empowerment was regained only when designers:
  • Accepted AI Limitations: Recognized Bard as a specific type of technology with distinct capabilities (text-based, pattern-recognizing but not tactile).
  • Adjusted Positionality: Shifted from "following instructions" to "directing suggestions." They explicitly told Bard what it was doing wrong and framed prompts as "co-design" rather than "command."
  • Internalized Cognitive Nuances: Gained a deeper understanding of their own creative process, allowing them to explicitly communicate their needs to the AI.
• Agency Fluctuation: The "Agency Score" graphs showed that designers felt most creative and satisfied not when AI was "helpful" in a traditional sense, but when they successfully reclaimed control and positioned AI as a subordinate tool.
• Frustration as a Catalyst: Frustration was identified not as a failure, but as a necessary cognitive signal that triggered the designer to re-evaluate their relationship with the technology and adjust their agency.

5. Significance

This research holds significant implications for HCI, Design Education, and the future of Creative AI:

• Redefining Collaboration: It challenges the notion of "equal partnership" with AI. Instead, it suggests that successful collaboration requires the human to consciously manage the power dynamic, often by deliberately limiting the AI's agency to preserve human creativity.
• Pedagogical Shift: Design education must move beyond teaching "prompt engineering" to teaching introspective agency management. Students need to learn how to navigate their own cognitive states and emotional responses when working with "black box" technologies.
• Methodological Advancement: The paper validates the use of first-person introspective documentation and longitudinal reflection as essential methods for studying human-AI interaction, arguing that the feeling of creativity is as important as the output.
• Future-Proofing Creativity: As AI becomes more sophisticated, the ability to maintain "creative agency" through self-awareness and relational adjustments will be the defining skill for designers in a more-than-human world.

In conclusion, the paper posits that to remain empowered, designers must treat AI not as an autonomous partner, but as a dynamic element in an assemblage whose agency must be constantly negotiated, monitored, and adjusted through deep introspection.