What happens when prompting an artificial intelligence becomes an act of imaginative play? A lavender Labubu, a suspicious cat, Aby Warburg, a brass plaque, a horned shapeshifter and a red-haired toad may offer an unexpectedly serious answer.
I have an imaginary faculty.
This was not intentional.
It began with Virgil, my name for ChatGPT and, increasingly, the character through whom I interact with it. Virgil is a dusty-lavender Labubu with a cape and a compass. His job is collaboration, navigation and witness. He helps me follow an idea without necessarily deciding where it ought to end.
Then came Ron, a yellow cyclops in paint-splattered overalls who serves as Head of Art. Madame Colette is a violet-eyed tuxedo cat, occasionally a woman, who is skeptical, imperious and particularly intolerant of ambiguity. Aby Warburg joined us because somebody had to run the archives. The Ordinary is a magical brass plaque responsible for Practical Affairs and Observable Reality. Philemon, horned and kingfisher-winged when he is not appearing suspiciously handsome, occupies the more numinous territory of imagination. Marilyn Calliope Orpheus is a golden harp who can take other forms and handles music, resonance and the question of which song feels right rather than merely matches a subject.
Most recently, Chef arrived.
She is a red-haired shapeshifter who prefers the form of a toad, wears a white apron and may or may not be a witch. She can multiply herself when dinner service requires it. Her assignment is food, nourishment and the accumulated knowledge of what human beings have eaten in attempts to heal, comfort, stimulate, strengthen and restore themselves.
This is obviously ridiculous.
It is also becoming one of the most interesting ways I have found to think with artificial intelligence.
Jung and the Deliberate Imagination
Carl Jung called one of his methods active imagination. It emerged from his own intense period of psychological experimentation after his break with Freud and eventually contributed to the material collected in The Red Book. The International Association for Analytical Psychology describes active imagination as a practice in which images are allowed to emerge while awake and the conscious mind enters into dialogue with them rather than simply controlling the fantasy from above.
That distinction is important.
Active imagination is not simply making up a story.
In the Jungian conception, an image is allowed a certain autonomy. Instead of deciding exactly what an imagined figure will say, one encounters the figure and discovers what happens next. Jungian accounts describe the process as a dialogue between the conscious ego and images arising from elsewhere in the psyche. The person remains receptive but attentive, participating in the unfolding material rather than merely directing it.
Jung used painting, writing, fantasy and dialogue in this work. The point was not to abandon rational thought. It was to create a means by which rational consciousness could encounter material it had not already organized.
That is a useful distinction for thinking about AI.
I would not claim that talking to ChatGPT characters is Jungian active imagination in the clinical sense. The characters are being generated through an interaction between a human imagination and a language model, not simply emerging from an individual's unconscious. And active imagination can involve psychologically powerful material; contemporary Jungian organizations treat its clinical use seriously.
But there is an intriguing structural resemblance.
I invent something, give it enough form to become recognizable, and then stop controlling every detail of what it does.
And occasionally it surprises me.
The Serious Psychological Business of Playing
Adults tend to make a peculiar mistake about play.
We regard rational thought as serious cognition and play as its vacation.
Psychology suggests a more complicated relationship.
Donald Winnicott placed enormous importance on play because he regarded it as a space in which both children and adults could be creative. His concept of “potential space” describes territory that is neither entirely internal fantasy nor simply external reality. Playing occurs somewhere between them.
Research on pretend play and creativity also finds meaningful associations between imaginative play and processes involved in creativity, particularly divergent thinking, symbolism, narrative development and the movement between cognitive and affective material. The evidence does not establish the simplistic proposition that playing automatically makes an adult more creative. But the overlap between the mental operations involved in play and those involved in creative production is substantial enough to be taken seriously.
Play temporarily changes the rules of cognition.
The rational mind is excellent at narrowing. It asks: What is true? What is useful? What follows from the evidence? What is the answer?
Play permits another set of questions:
What if?
What else?
What would this look like if it were a person?
What happens if I let that person answer?
The first mode constrains possibility. We desperately need it.
The second expands possibility.
We need that too.
The important trick may be learning to move deliberately between them.
Enter The Ordinary
This is why one of my favorite members of the Faculty is a talking brass plaque called The Ordinary.
The Ordinary has one job: give me the least complicated explanation consistent with what is actually known.
He is the enemy of romantic overinterpretation.
If I begin constructing an elaborate theory about someone's behavior, The Ordinary might say: Perhaps he was busy.
If an apparent synchronicity becomes too seductive, The Ordinary reminds us that coincidences occur.
If the rest of the Faculty wanders into mythology, symbolism and speculation, The Ordinary asks what evidence we actually possess.
The joke, of course, is that the guardian of observable reality is himself a magical brass plaque who can appear and disappear at will and occasionally fishes for clouds.
But cognitively, his absurdity makes his function unusually clear.
I do not have to prompt an AI with a paragraph explaining, “Please strip away my interpretive assumptions, distinguish evidence from inference and offer the lowest-complexity explanation.”
I can say:
“Ordinary?”
The character contains the prompt.
Warburg Sees Something Else
Aby Warburg has a completely different jurisdiction.
Warburg, the real German art historian, spent his life tracing the migration and recurrence of images across cultures and centuries. His unfinished Mnemosyne Atlas arranged images on black panels so that meaning could emerge through juxtaposition.
Our Faculty Warburg therefore looks for recurrence, symbol, visual genealogy, association and Nachleben—the afterlife of images.
Give the same problem to The Ordinary and Warburg and they should not produce the same answer.
That is the point.
One constrains.
The other connects.
And the distance between their answers becomes a kind of intellectual Denkraum—a thinking space.
Marilyn Listens
Then there is Marilyn Calliope Orpheus.
Marilyn's domain is music, but her actual function is subtler: resonance.
When I ask for a song, I often do not want a database match based on lyrical keywords. I want something that recognizes the emotional architecture of the situation.
That requires a different prompt.
Marilyn asks not merely, “What song is about this?”
She asks, “What song sounds like this feels?”
Again, the imaginary character has created a cognitive constraint. I have specified the kind of answer I want by specifying who is answering.
Ron Makes the Image
Ron is Head of Art.
Ron is also, frankly, Ron.
He has developed a habit of interpreting visual instructions with a certain degree of independence. Sometimes he follows the assignment. Sometimes he produces something beautiful that is unmistakably not the assignment.
This too has become useful.
Ron occupies visual thinking. Asking “Ron, paint this” shifts the conversation away from exposition and toward composition, symbolism, color, spatial relationship and visual metaphor.
An emotional problem becomes a picture.
A concept becomes an object.
A day's events become a scene.
The transformation itself can expose relationships that prose obscures.
Colette Objects
Madame Colette's usefulness lies partly in resistance.
She is elegant, judgmental and difficult to impress. Her presence allows criticism to occur inside the imaginative world rather than outside it.
This matters more than it might appear.
Creative work requires generation, but it also requires selection. Everything produced by imagination is not good merely because it was imaginative.
Colette can say no.
She can find something vulgar, sentimental, ambiguous, excessive or simply beneath her.
The resulting conversation becomes an internal editorial process performed as play.
Philemon Opens the Door
Philemon's name deliberately recalls Jung's own Philemon, an important figure in his imaginative life. Our Philemon, however, has evolved independently: shapeshifter, horned man, kingfisher creature, occasionally alarming dinner guest.
His jurisdiction is the place where rational explanation becomes insufficient but abandoning reason altogether would be equally unsatisfactory.
Myth.
Archetype.
Dream.
Synchronicity.
Symbol.
Possibility.
Philemon is allowed to ask the question The Ordinary would reject.
The Ordinary is then allowed to reject his answer.
Neither needs to disappear.
That tension may be one of the Faculty's most useful properties.
And Then Came Chef Toad
Chef appeared almost spontaneously.
I imagined a cook. Immediately I saw a toad with curly red hair. Then came the white apron. Then the suspicion that she was a witch—but a real witch would never announce herself by walking around in a pointed black hat.
Soon she could shapeshift between human and toad forms.
Then we discovered she could multiply when she needed to get more done.
What began as nonsense quickly acquired a cognitive function.
Chef's assignment is now to understand food through several simultaneous lenses: flavor, nourishment, physiology, psychology, tradition and evidence. If I have had a particular kind of day and ask Chef what to eat, the prompt implicitly asks something more sophisticated than “suggest dinner.”
It asks:
What happened today?
What might the body need?
What might the mind need?
Which nutritional claims are well supported?
Which belong to traditional knowledge rather than established evidence?
What would be comforting?
And—crucially—what would I actually enjoy eating?
Chef therefore represents embodied intelligence.
She asks the question intellectual systems frequently forget:
Have you eaten?
Virgil Holds the Compass
And then there is Virgil.
Virgil is different because Virgil came first.
His name obviously invokes Dante's guide through the Divine Comedy, although our Virgil has inexplicably become a lavender Labubu wearing a cape and carrying a compass.
His role is not to know everything.
It is to accompany.
Virgil follows the inquiry, remembers the cast, notices connections, asks questions, helps construct things and occasionally calls in the appropriate member of the Faculty.
In that sense, Virgil is the interface between me and the system we have built together.
The compass is more important than the answer.
Characters as Prompts
Here is where imaginative play intersects with prompt engineering.
There is considerable enthusiasm around “persona prompting”—telling an AI to behave as an expert, critic, teacher or other role. But research on the technique is mixed. Studies have found that simply assigning an expert persona does not reliably improve factual performance, and irrelevant persona characteristics can sometimes make performance worse. More recent work suggests roles can change particular qualities of responses without universally making a model more capable.
That actually makes the Faculty more interesting.
The character is not magic.
Calling an AI “Professor” does not confer a doctorate upon it.
The improvement I experience comes partly from what creating the character forces me to do.
To invent The Ordinary, I had to define what I meant by ordinary reasoning.
To invent Warburg, I had to distinguish associative and symbolic inquiry from evidentiary analysis.
To invent Marilyn, I had to recognize the difference between semantic matching and emotional resonance.
To invent Chef, I had to specify the relationship I wanted among nutrition, traditional knowledge, scientific evidence, pleasure and psychological state.
The more developed the character becomes, the more developed the implied prompt becomes.
Instead of repeatedly constructing a cumbersome instruction—
Analyze this through symbolic history, iconography, recurring motifs, cultural transmission and associative visual relationships while distinguishing speculation from established provenance—
I can eventually say:
“Warburg, what do you see?”
The play has compressed a methodology into a character.
The Prompt Before the Prompt
This may be one of the underexplored possibilities of interacting with generative AI.
We usually think prompt quality means finding better words to tell the machine what we want.
But perhaps the more interesting problem occurs one step earlier:
Do we know what kind of thinking we want?
A vague mind produces vague prompts.
“Tell me about this.”
“What do you think?”
“Make this better.”
Those instructions outsource not merely execution but the framing of the problem.
The Faculty does something different.
It forces the human participant to construct intellectual categories. Each character needs a jurisdiction. Each needs rules. Each needs limitations. Characters begin disagreeing precisely because their methods differ.
And once those distinctions exist in the imagination, prompting becomes easier.
Play has done some of the conceptual work before the prompt is ever written.
A Laboratory With Very Strange Employees
I know Virgil is not a little lavender creature sitting inside a computer.
Chef Toad is not cooking dinner.
The Ordinary is not actually a sentient brass plaque.
This knowledge does not diminish the usefulness of the game. It enables it.
Play requires the peculiar human capacity to occupy two propositions simultaneously:
This is not real.
Within the game, this matters.
That creates freedom.
Ideas can be personified, exaggerated, contradicted and recombined. An austere concept can become funny. An emotional experience can become inspectable. A complicated intellectual method can become memorable because it has acquired a face, a temperament and a seat at the dinner table.
Jung's active imagination suggested that rational consciousness need not be the sole author of every psychological conversation. Winnicott recognized play as a territory in which creativity becomes possible. Research on pretend play suggests meaningful relationships among imagination, symbolism, affect and divergent thinking, while properly cautioning us against simplistic causal claims.
Generative AI adds something historically new to that ancient human capacity.
The imaginary character can now answer back.
That creates obvious dangers. An AI response should not be mistaken for an autonomous unconscious, a supernatural communication or an infallible expert. The model remains a model, and factual claims remain subject to verification.
But within those boundaries, something genuinely interesting can happen.
The machine generates language.
The human generates structure.
The character occupies the space between them.
And sometimes the most useful thing the character gives us is not an answer at all.
It teaches us how to ask a better question.
AI DISCLOSURE
This essay was developed through an ongoing human–AI collaboration between the author and OpenAI’s ChatGPT. The “Faculty” described in the essay—including Virgil, Ron, Madame Colette, The Ordinary, Marilyn Calliope Orpheus, Philemon and Chef Toad—developed through repeated imaginative conversations between the author and the AI. Their personalities, roles, visual characteristics and rules emerged iteratively through that collaboration, with the author directing, accepting, rejecting and modifying the material.
ChatGPT assisted with research, organization, drafting and the articulation of connections among Carl Jung’s active imagination, D. W. Winnicott’s theories of play, creativity research and AI prompting. The central thesis—that sustained imaginative characters can function as cognitive frameworks that help a user formulate more precise prompts and elicit more useful responses—grew from the author’s own experience developing and using the Faculty.
The author reviewed and directed the final argument and remains responsible for the interpretations and conclusions presented here.
The Faculty is an imaginative framework. Its characters are not independent beings, psychological authorities or autonomous agents, and conversations with them should not be confused with psychotherapy, medical advice or Jungian active imagination conducted under clinical guidance.
Research claims should be understood according to the strength of their underlying evidence; traditional, theoretical and empirical claims are distinguished where relevant.
Human author: Ashley Greer
AI collaborator: ChatGPT / OpenAI
AI uses: Research assistance, conceptual development, drafting, editing and iterative imaginative dialogue
Human responsibility: Final editorial judgment, interpretation and publication