Extended Creativity: A Conceptual Framework for Understanding Human-AI Creative Relations

arXiv:2506.10249 · cs.HC, cs.AI · Submitted 2025-06-12 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: We're looking at the paper: "Extended Creativity: A Conceptual Framework for Understanding Human-AI Creative Relations".

Jane: The gist: This paper identifies three fundamental modes by which AI can support and shape creative processes—Support, Synergy, and Symbiosis—defined by technical autonomy and perceived agency.

Tom: First, who's behind it and why it matters.

Title and authors: Tom: We're diving into Extended Creativity: A Conceptual Framework for Understanding Human-AI Creative Relations, and we need to unpack what that actually means for us in a practical sense.

Jane: Yeah, the authors are trying to move past just calling AI a tool or a replacement; they’re proposing this relational perspective.

Lu: They start by defining Extended Creativity systems as these "socio-technical environments where humans and AI interact to shape and transform creative processes" (<ref:2506.10249#pg2>).

Meng: So it’s not just about the output you get from the AI; it’s about how the environment itself—the whole setup—is shaping what creativity looks like in a production setting.

Tom: They define three fundamental modes here: Support, Synergy, and Symbiosis. These modes are defined by two things: technical autonomy and perceived agency.

Jane: That's right. Support is where AI acts as a functional tool without substantially modifying the creative process, with humans retaining full control.

Lu: In that mode, AI operates through "causal mechanisms, executing direct operations without context sensitivity or interpretive flexibility" (<ref:2506.10249#pg1>).

Meng: So it’s basically AI just following instructions precisely rather than helping brainstorm or interpreting the overall vision.

Tom: Synergy is where humans and AI function as collaborative partners, each contributing dialogically to the process in ways that influence each other.

Jane: That mode operates between levels three and four of Simmler and Frischknecht's taxonomy, showing context sensitivity in its agency (<ref:2506.10249#pg3>).

Lu: This mode significantly impacts pro-c creativity, enabling creators to push beyond conventional approaches within their fields (<ref:2506.10249#pg3>).

Meng: That sounds like a real partnership, but I need to know if that means we're just running two separate tools side-by-side or if they're actually talking to each other.

Tom: And then there’s Symbiosis, where human and AI agents become so deeply integrated that they form a unified creative entity, blurring the boundaries between machine and human contributions (<ref:2506.10249#pg1>).

Jane: That deepest integration is linked to approaching intentional agency; it's when the system starts acting with purpose or motivation in a way that’s hard to pin down.

Lu: This configuration operates at the highest levels, level four and five of Simmler and Frischknecht's taxonomy, holding potential for transformational creativity (<ref:2506.10249#pg3>).

Meng: If it gets that deep, I’m wondering what the failure modes look like when that unified system goes wrong at scale.

Tom: It sets up a structure to investigate how these evolving human-AI relationships transform our understanding of what it means to create and relate in this new landscape.

The paper's summary: Jane: Now we’re looking at the actual substance of the paper, how they build this framework from scratch, focusing on those three modes we just touched on.

Tom: The authors spend a lot of time defining what these modes are based on autonomy and agency. They break agency down into causal, contingent, and intentional agency (<ref:2506.10249#pg1>).

Lu: That distinction between those three levels of agency is crucial because it’s what allows them to define autonomy across five ascending levels, from level one, which is transparent and deterministic systems, up to level five, open and co-evolving systems (<ref:2506.10249#pg1>).

Meng: So they are mapping the technical capability of the AI against the kind of human control or purpose we think is involved.

Tom: They define Support as AI acting through causal mechanisms, executing direct operations without any context sensitivity or interpretive flexibility (<ref:2506.10249#pg1>).

Jane: That means the AI doesn't really understand the 'why' behind the task, just the 'how,' and it stays firmly in that foundational level of creativity—enhancing mini-c and little-c creativity (<ref:2506.10249#pg1>).

Lu: They then define Synergy as a mode where humans and AI function as collaborative partners who contribute dialogically, operating at higher levels that influence pro-c creativity (<ref:2506.10249#pg3>).

Tom: So Synergy is where the interaction becomes iterative; it’s not just giving a command, it’s a back-and-forth refinement of both the artifact and the process itself (<ref:2506.10249#pg3>).

Meng: That sounds like real collaboration, but I need to know if that means we're just running two separate tools side-by-side or if they're actually talking to each other.

Jane: And then there’s Symbiosis, which they call the deepest integration. This is where human and AI agents become so unified that the boundaries between what you do and what the machine does become blurred (<ref:2506.10249#pg1>).

Lu: They suggest this Symbiosis mode could potentially transform all levels of the Four-C framework and lead toward transformational creativity, which is described as the rarest and most radical form of creative behavior (<ref:2506.10249#pg3>).

Tom: And to measure that deep integration, they propose new metrics for evaluation here, specifically measuring epistemic surprise to operationalize that idea of creative agency in these deeper modes.

Jane: So what this means for us isn't just building better tools; it’s figuring out how to design relationships—whether that's Support, Synergy, or Symbiosis—to get the specific kind of creative outcome you actually want.

The paper's improvements: Tom: Let’s talk about what the authors suggest we do next, because they aren't just stopping at describing these modes; they’re proposing concrete design implications for each one.

Jane: They aren't just stopping at describing the modes; they are proposing concrete design implications for each one—like saying exactly what your interface needs to look like for Support systems.

Lu: For Support systems, they suggest the focus should be on clarity and control to reduce cognitive load without diluting the creator's sense of ownership or expressive identity (<ref:2506.10249#pg1>).

Meng: Clarity and control. That’s what I need when I’m trying to deploy something that doesn't break every time a user tries to override it; it sounds like they want those systems predictable in their execution.

Tom: They also say for Synergy systems, you need conversational adaptability, memory across sessions, and multimodal interfaces that support cross-domain creativity while still balancing responsiveness with transparency (<ref:2506.10249#pg3>).

Jane: That means we need a system that can remember what we talked about yesterday and switch modes based on what's needed for the creative task.

Lu: They're pushing us toward new metrics for evaluation too, like measuring epistemic surprise to properly gauge that level of creative agency they’re talking about.

Tom: So we’ve finished our deep dive into Extended Creativity: A Conceptual Framework for Understanding Human-AI Creative Relations, which mapped out Support, Synergy, and Symbiosis based on autonomy and agency levels.

Jane: Exactly. It shows that the creative relationship isn't just about what AI can do; it’s how you want to interact with it across those three relational levels of autonomy and agency.

Conclusion: Tom: So we’ve mapped out Extended Creativity: A Conceptual Framework for Understanding Human-AI Creative Relations, covering Support, Synergy, and Symbiosis based on autonomy and agency levels.

Jane: Right. It shows that the creative relationship isn't just about what AI can do; it’s how you want to interact with it across those three relational levels of autonomy and agency.

Lu: The paper sets up this framework by defining these "socio-technical environments" where humans and AI interact to shape creative processes. It posits that algorithmic techniques aren't just tools; they actively mediate, enhance, and reconfigure human imagination, interpretation, and agency within that whole system (<ref:2506.10249#pg2>).

Meng: So it’s not just about the output you get from the AI; it’s about how the environment itself—the whole setup—is shaping what creativity looks like.

Tom: They break down these interactions into three relational modes based on autonomy and agency, but this section really hammers home how much control the AI has versus how much human intent is involved.

Jane: The main thing they nail is that agency isn't one big thing; it’s a continuum. They break it down into causal agency, which is just the basic capacity to produce effects, contingent agency, which means responding to context and choosing alternatives, and intentional agency—that's when you can attribute purpose or reflection to the action.

Lu: That distinction between those three levels of agency is crucial because it’s what allows them to define autonomy across five ascending levels. Level one is transparent and deterministic systems, while level five is open and co-evolving systems (<ref:2506.10249#pg1>).

Tom: I noticed they define Support as AI acting through causal mechanisms, executing direct operations without any context sensitivity or interpretive flexibility. That means the AI doesn't really understand the 'why' behind the task, just the 'how.'

Jane: That’s right. It’s purely operational; it enhances your skills by handling repetitive technical adjustments so you can focus on conceptualization, but because of that lack of interpretive flexibility, it stays firmly in that foundational level of creativity—enhancing mini-c and little-c creativity (<ref:2506.10249#pg1>).

Meng: From a deployment standpoint, the paper says Support systems are the least complex ethically because they clearly delineate between your intent and the AI execution. That’s something I can actually work with when shipping things to production.

Lu: They also defined Synergy as a mode where humans and AI function as collaborative partners who contribute dialogically. This mode operates at higher levels, influencing pro-c creativity, which is about pushing beyond conventional approaches in your field by exploring unexpected associations together (<ref:2506.10249#pg3>).

Jane: And then there's Symbiosis, which they call the deepest integration. This is where human and AI agents become so unified that the boundaries between what you do and what the machine does become blurred (<ref:2506.10249#pg1>).

Meng: That sounds ambitious, but they link that configuration to approaching intentional agency; it’s when the system starts acting with purpose or motivation in a way that’s hard to pin down.

Lu: Exactly. They suggest this Symbiosis mode could potentially transform all levels of the Four-C framework and lead toward transformational creativity, which is described as the rarest and most radical form of creative behavior they are looking for (<ref:2506.10249#pg3>).

Lu: It’s about moving from viewing AI as a mere tool to understanding these complex relational modes and their distinct configurations along the autonomy and agency axes.

Meng: I think for my team, the most important thing is using this structure to guide our deployment decisions on whether we’re building a tool or trying to build a true collaborative partner.

Tom: Agreed. It’s about moving past simple narratives and structuring how we investigate those deep relationships between human creators and AI systems.

Jane: We'll be looking at more papers soon that tackle the practical engineering challenges of implementing these Support, Synergy, and Symbiosis configurations.

Research Center in Communication Psychology (PSICOM), Università Cattolica del Sacro Cuore, Milan, Italy

cs.HC, cs.AI

Submitted: 2025-06-12

Updated: 2025-06-15

Comments: 36 pages, 3 figures. This conceptual paper proposes a taxonomy of Extended Creativity systems and examines the relational dynamics between human and AI agents in creative processes. Suitable for readers in HCI, AI, cognitive science, and digital design. The illustrations were created by Francesco Giordano and are used with permission (not under CC license)

Journal ref: In M. Chiriatti and G. Riva (eds.), Humane Artificial Intelligence: From Foundations to Application and Policy Framework, Vita e Pensiero (2025), pp. 136-161

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 78/100

The gist: The gist: This paper identifies three fundamental modes by which AI can support and shape creative processes—Support, Synergy, and Symbiosis—defined by technical autonomy and perceived agency.

Key concepts

Extended Creativity Systems: These are socio-technical environments where humans and AI interact to shape and transform creative processes. They treat AI not just as a tool, but as an active participant that mediates, enhances, and reconfigures human imagination, interpretation, and agency within the creative workflow.
Support Mode: In this mode, AI acts as a functional tool executing direct operations without interpretive flexibility. It focuses on augmenting human capabilities by reducing cognitive load during repetitive technical adjustments. Humans retain full control over the creative process and conceptual authorship.
Synergy Mode: This involves humans and AI acting as collaborative partners who contribute dialogically to the creative process. The architecture shows contextual sensitivity, enabling creators to push beyond conventional approaches by generating unexpected continuations through cross-domain creativity.
Symbiosis Mode: This is the deepest integration where human and AI agents form a unified creative entity with blurred boundaries. It operates at high levels of intentional agency, potentially leading to transformational creativity—the rarest form of creative behavior that fundamentally redefines creation.

Summary

The gist: This paper identifies three fundamental modes by which AI can support and shape creative processes—Support, Synergy, and Symbiosis—defined by technical autonomy and perceived agency.

Conceptual Foundation

The framework is built upon the concept of Extended Creativity systems, which are defined as socio-technical environments where humans and AI interact to shape and transform creative processes (Page 2). This perspective posits that algorithmic techniques mediate, enhance, and reconfigure human imagination, interpretation, and agency within these systems (Page 2). The study moves beyond viewing AI as a mere tool by examining how it functions within a relational context (Page 3).

Three Relational Modes

The paper outlines three primary relational modes through which Extended Creativity may emerge:

Support: In this mode, AI operates as a functional tool that augments human capabilities without substantially modifying the creative process, with humans retaining full control (Page 2). This configuration is characterized by AI acting through causal mechanisms, executing direct operations without context sensitivity or interpretive flexibility (Page 14). It primarily influences the foundational levels of creativity within the Four-C framework, specifically enhancing mini-c creativity and facilitating little-c creativity (Page 15).
Synergy: In this mode, humans and AI function as collaborative partners, each contributing actively and dialogically to the creative process in ways that influence the other's contributions (Page 16). The technological architecture operates between levels 3 and 4 of Simmler and Frischknecht's taxonomy, demonstrating contextual sensitivity in its agency (Page 17). This mode significantly impacts pro-c creativity, enabling creators to push beyond conventional approaches within their fields (Page 18).
Symbiosis: In this mode, human and AI agents become so deeply integrated that they form a unified creative entity, with boundaries between human and machine contributions becoming increasingly blurred (Page 20). This configuration operates at the highest levels (4 and 5) of Simmler and Frischknecht's taxonomy, approaching intentional agency (Page 23). It potentially transforms all levels of the Four-C framework, holding potential for transformational creativity—the rarest and most radical form of creative behavior (Page 23).

Dimensions of Configuration

The modes are defined along two key dimensions: autonomy and agency. Autonomy is defined as a system’s capacity for self-governance: the ability to act according to internal mechanisms, goals, or principles, without direct external control (Page 10). This is operationalized using a taxonomy with five ascending levels, ranging from Level 1 (Transparent and deterministic systems) to Level 5 (Open and co-evolving systems) (Page 10). Agency is characterized as a continuum involving three principal levels: Causal agency: the basic capacity to produce effects or exert influence, Contingent agency: the ability to respond to context and choose among alternatives, and Intentional agency: the capacity—often attributed rather than intrinsic—to act with purpose, motivation, or reflection (Page 12).

Implications for Research and Design

The framework suggests distinct implications for research and design across different modes. For Support systems, design should focus on clarity, control, and consistency, aiming to reduce cognitive load without diluting the creator’s sense of ownership or expressive identity (Page 30). Synergy systems require conversational adaptability, memory across sessions, and multimodal interfaces that support cross-domain creativity while balancing responsiveness with transparency (Page 31). Symbiosis design involves real-time sensing (e.g., neural input), predictive modeling, and adaptive feedback loops, focusing on cognitive transparency (Page 30). Furthermore, the framework proposes new metrics for evaluation, such as measuring epistemic surprise to operationalize creative agency (Page 31).

Broader Dimensions of Impact

The Extended Creativity framework allows for investigation into neuro-cognitive and psycho-social dimensions. Research questions include how cognitive redistribution in Support relationships affects creative depth (Page 28), how Synergy might enhance divergent thinking through exposure to unexpected associations (Page 29), and exploring neural signatures of extended symbiotic interaction (Page 28). Psycho-socially, the framework enables exploration of how configurations reshape creative communities and cultural values, addressing concerns about authorship, creative authenticity, and dependency on proprietary systems (Page 19).

Operational Structures

Extended Creativity systems can vary in their coordination mechanisms, ranging from explicit hierarchical arrangements to emergent, or stigmergic patterns (Page 26). Coordination structures interact with the relational modes to create diverse relationships (Table 1) (Page 27). The paper also introduces a relational polarity axis—cooperation vs. competition—orthogonal to the structural configurations of Support, Synergy, and Symbiosis to capture the tone of interaction (Page 26).

Conclusion

The framework aims to move beyond simplistic narratives by offering a living conceptual map for understanding relational modes of creative co-agency (Page 32). It suggests that the creative value does not scale linearly with machine autonomy, as even Support configurations can amplify human creative focus (Page 32). The ultimate goal is to understand how these evolving human-AI relationships transform our understanding of what it means to create and relate in a world increasingly co-shaped by natural and artificial intelligence (Page 32).

The gist: This paper identifies three fundamental modes by which AI can support and shape creative processes—Support, Synergy, and Symbiosis—defined by technical autonomy and perceived agency.</ref:2506.10249

Extended Creativity systems, defined as socio-technical environments where humans and AI interact to shape and transform creative processes (Page 2). This perspective posits that algorithmic techniques mediate, enhance, and reconfigure human imagination, interpretation, and agency within these systems (Page 2).</ref:3

In the Support mode, AI acts through causal mechanisms, executing direct operations without context sensitivity or interpretive flexibility (Page 14).</ref:14

In Symbiosis relationships, human and AI agents become so deeply integrated that they form a unified creative entity, with boundaries between human and machine contributions becoming increasingly blurred (Page 20).</ref:20

The Synergy model significantly expands its influence across the creativity spectrum defined by Kaufman and Beghetto's (2009) Four-C model.</ref:18

The Support mode primarily influences the foundational levels of creativity within the Four-C framework (Kaufman & Beghetto, 2009).</ref:31

Autonomy can be broadly defined as a system's capacity for self-governance: the ability to act according to internal mechanisms, goals, or principles, without direct external control (Page 10).</ref:4

Agency is characterized as a continuum involving three principal levels: Causal agency: the basic capacity to produce effects or exert influence, Contingent agency: the ability to respond to context and choose among alternatives, and Intentional agency: the capacity—often attributed rather than intrinsic—to act with purpose, motivation, or reflection (Page 12).</ref:12

The Support mode presents the least complex ethical risks among the three interaction modes.</ref:30

Synergy systems require conversational adaptability, memory across sessions, and multimodal interfaces that support cross-domain creativity while balancing responsiveness with transparency (Page 31).</ref:31

Coordination Typically hierarchical with clear human leadership.</ref:27

The Support systems function best when they clearly delineate between human intent and AI execution, helping users streamline technical tasks while maintaining full conceptual authorship.</ref:30

Synergy relationships characterize the interaction as dialogic rather than merely instrumental, characterized by iterative refinements that transform both the artifact and the collaborative process.</ref:29

Symbiosis systems require real-time sensing (e.g., neural input), predictive modeling, and adaptive feedback loops (Page 30).</ref:31

The framework allows for investigation into how cognitive redistribution in Support relationships affects creative depth (Page 28).</ref:28

The Support mode primarily influences the foundational levels of creativity within the Four-C framework, specifically enhancing mini-c creativity and facilitating little-c creativity (Page 15).</ref:4

Support systems function best when they clearly delineate between human intent and AI execution, helping users streamline technical tasks while maintaining full conceptual authorship.</ref:30

Coordination Typically hierarchical with clear human leadership.</ref:26

The framework enables exploration of how different human-AI configurations reshape creative communities and cultural values (Page 28).</ref:19

The Support mode presents the least complex ethical risks among the three interaction modes.

Improvements for AI systems

Support mode systems can be designed to enhance existing creative skills or support the development of new ones through guided assistance, specifically by reducing cognitive load associated with repetitive technical adjustments so creators can focus on conceptualization.
Synergy systems should incorporate conversational adaptability, memory across sessions, and multimodal interfaces that support cross-domain creativity, allowing the AI to function as a computer colleague that generates unexpected and surprising continuations.
Symbiosis systems must implement real-time sensing (e.g., neural input), predictive modeling, and adaptive feedback loops to create a continuous feedback circuit, enabling the system to generate outputs that both respond to and guide the user's evolving cognitive state.

Abstract

Artificial Intelligence holds significant potential to enhance human creativity. However, achieving this vision requires a clearer understanding of how such enhancement can be effectively realized. Drawing on a relational and distributed cognition perspective, we identify three fundamental modes by which AI can support and shape creative processes: Support, where AI acts as a tool; Synergy, where AI and humans collaborate in complementary ways; and Symbiosis, where human and AI cognition become so integrated that they form a unified creative system. These modes are defined along two key dimensions: the level of technical autonomy exhibited by the AI system (i.e., its ability to operate independently and make decisions without human intervention), and the degree of perceived agency attributed to it (i.e., the extent to which the AI is experienced as an intentional or creative partner). We examine how each configuration influences different levels of creativity from everyday problem solving to paradigm shifting innovation and discuss the implications for ethics, research, and the design of future human AI creative systems.