Cognitive Science Artificial Intelligence

Computational Creativity

At a glance: Theoretical framework founded by Margaret Boden (1990) to analyze creativity as a modelable process, not as a mystery. It distinguishes three types of creativity and proposes a key distinction between novelty for oneself (P-creativity) and novelty for humanity (H-creativity) — a distinction directly useful in therapeutic contexts.

Theoretical framework

This resource is grounded in a computationalist perspective (creativity as an analyzable, modelable process) and a cognitive one (creativity as a mental operation, not a mysterious gift). This framework, drawn from cognitive psychology and artificial intelligence, does not claim to exhaust the creative phenomenon — it illuminates its mechanisms. → See other perspectives

Why this concept matters

Generative AI tools (Midjourney, Suno, ChatGPT) produce images, music, and texts. Patients use them, sometimes in therapeutic contexts. But is AI "creative"? This question often traps clinicians in a sterile philosophical debate.

The computational creativity framework allows us to move past this impasse:

1. Refocus on the patient: in therapy, what matters is not whether AI is "truly" creative, but whether the process is therapeutically productive for the person
2. Qualify what AI does: have a precise vocabulary (exploration, transformation, combination) to analyze creative patient-AI interactions
3. Assess therapeutic value: the P/H-creativity distinction shows that novelty for the patient is the only thing that matters clinically

The Three Types of Creativity (Boden)

Margaret Boden identifies three fundamental creative processes, of increasing complexity. Each has different implications when AI enters the equation.

1. Combinational creativity

Combining familiar ideas in novel ways: metaphors, analogies, unexpected collages. This is the most common form of everyday creativity.

AI example: A patient types a few words into Suno AI and gets a song mixing rap and cello — a combination they would never have imagined alone.

2. Exploratory creativity

Systematically exploring the possibilities of a "conceptual space" — a style, a genre, a set of rules. Like a jazz musician exploring all possible variations of a harmonic progression.

AI example: Generating the same patient phrase in 5 different musical styles (classical, electronic, folk, metal, jazz). The patient discovers that their words "sound" different depending on the context — an exploration of their emotional space.

3. Transformational creativity

Modifying the fundamental rules of the conceptual space itself. This is the rarest and most profound form: it produces results so surprising that they redefine the domain. Picasso moving from figurative to cubism.

In therapy: When a patient, through an AI-assisted creative process, changes their very way of conceiving their problem — not just a new solution, but a new framework for thinking.

P-creativity and H-creativity: The Key Distinction

Boden distinguishes two levels of novelty:

P-creativity

Psychological creativity — the idea is new to the person who produces it. This is subjective creativity, the one that matters in therapy.

H-creativity

Historical creativity — the idea is new to all of humanity. Mozart, Einstein, Picasso. This level is not relevant in clinical work.

Why this is liberating for clinicians: The question "is AI truly creative?" concerns H-creativity — can it produce something new for humanity? The debate is fascinating but irrelevant in therapy.

On the other hand, the question "does this AI process produce something new for this patient?" — that's P-creativity, and it's clinically measurable. A patient who has never put their emotions into music and does so via Suno AI experiences authentic P-creativity, regardless of the "historical" originality of the result.

Illustrative Clinical Case

Karim, 34, in art therapy for a depressive episode. Traditional mediations (painting, clay modeling) set him up for failure: "I can't draw, I have no talent." The therapist suggests using Suno AI.

Karim types a few sentences about how he feels. The AI generates a song. He is surprised: "That's weird, those are my words but it doesn't sound like I imagined." The therapist suggests varying the style (exploratory creativity). In the folk version, Karim recognizes himself. In the metal version, he laughs: "That's my anger, right there."

Over the following sessions, Karim starts modifying his texts based on the results, exploring emotions he had not verbalized. He moves from a passive stance ("I can't create") to an active one ("I choose what the AI does with my words").

Reading through Boden's framework: the AI provides combinational creativity (assembling text + musical style) and exploratory creativity (traversing the space of styles). For Karim, the result is authentic P-creativity: he had never put his emotions into musical form. It doesn't matter that the song isn't a masterpiece (H-creativity) — what matters is that Karim accessed a form of self-expression he didn't believe was possible.

In Practice for Clinicians

Distinguish process from product: what matters is not the artistic quality of the result, but what the co-creation process with AI mobilized in the patient (emotions, representations, insights).
Use variation as a tool: generate multiple versions of the same content (styles, tones) to explore the patient's emotional space. This is exploratory creativity in service of self-discovery.
Welcome surprise: the unpredictability of AI (it never produces exactly what we expect) can be therapeutically productive — it confronts the patient with the unexpected and invites them to take a stance.
Stay focused on P-creativity: if a colleague or patient objects that "AI isn't really creative," remember that the therapeutic criterion is novelty for the patient — not for art history.

Points of Caution

This framework does NOT claim that:

AI is "truly" creative in the human sense — the debate remains open
All use of generative AI is therapeutic — the intention, setting, and clinician's guidance remain decisive
Creativity reduces to a computational process — emotional, embodied, and relational dimensions go beyond this framework

Limitations to keep in mind:

Ease vs. creative effort: if AI makes the process "too easy," the therapeutic benefit of creative struggle may be lost. The challenge is finding the right balance.
Agency: the patient must remain the author of their process. If AI "does everything," the P-creativity belongs to the AI, not the patient.
Tool dependency: the therapeutic goal is not for the patient to become an expert user of Suno AI, but for them to access their own capacities for expression and representation.

Other Perspectives

Computational creativity belongs to a Western analytical tradition. Other perspectives enrich our understanding of the creative process in therapy.

Phenomenology: Embodied creativity

For Merleau-Ponty, creativity cannot be reduced to a cognitive process: it engages the body, perception, gesture. Modeling creativity computationally necessarily loses its embodied dimension.

For clinicians: A reminder that AI-mediated creation does not offer the sensory engagement of clay modeling or painting — a dimension to compensate for.

Winnicott: Transitional space

The transitional object (blanket, drawing, melody) is "neither me nor not-me." Generative AI could occupy this intermediate zone: what it produces is neither entirely the patient's nor entirely the machine's.

For clinicians: Patient-AI co-creation as a new transitional space — a psychodynamic framework complementary to Boden's cognitive approach.

Romantic critique: The creative mystery

For the heirs of Romanticism, reducing creativity to "mechanisms" is an impoverishment. Inspiration, intuition, "genius" escape all modeling. Boden herself acknowledged that her framework does not exhaust the phenomenon.

For clinicians: Some patients experience creation as a quasi-mystical experience. This lived experience is clinically relevant even if it doesn't fit within the computationalist framework.

Boden's framework is an analytical tool, not a total theory of creativity. Its clinical value is to provide a precise vocabulary for accompanying patients in their creative uses of AI — not to reduce creativity to an algorithm.

This Concept in Our Tool Pages

Boden's framework applies directly to generative AI tools used in therapeutic contexts.

Suno

Combinational and exploratory creativity in action — music generation as a field of observation