Methodology of the Phantom Mirror: From Psychoanalytic Practice to Generative Epistemology
Abstract
This essay outlines the methodological foundations of Phantom Mirror, a practice-based research project that integrates psychoanalytic frameworks with generative AI image-making. It argues that the project constitutes a form of generative epistemology, a method of producing knowledge through iterative, relational engagement between human and machine. Drawing on traditions of psychoanalytic praxis, phenomenology, and creative research, the essay positions Phantom Mirror as both therapeutic and philosophical experiment. Through the recursive processes of prompting, reflection, and image generation, the project performs a dialogue between the conscious and unconscious, the real and the simulated, transforming AI from technological instrument into a relational container for affect and meaning.
Keywords
Methodology; practice-based research; psychoanalysis; phenomenology; AI art; generative epistemology; Phantom Mirror; relational consciousness
1. Introduction
The Phantom Mirror project evolved from a need to find a language for experiences that resisted representation, grief, memory loss, and the erosion of recognition in dementia. Rather than documenting these phenomena, the project stages them through a process of collaborative image generation with artificial intelligence. This methodology diverges from both traditional phototherapy and conventional artistic production: it operates within an intersubjective space between artist, algorithm, and image.
To understand this methodology is to understand its philosophical ground. Phantom Mirror does not treat the AI image as a product but as a thinking apparatus, a site where theory and affect coalesce. Its methods are recursive rather than linear, performative rather than descriptive. As Eliza Stephens (2025) notes, the process “turns Baudrillard’s philosophical stance into a methodology: thinking happens through the image, not about it.” The project thus embodies what can be termed generative epistemology, knowledge produced through the act of generation itself.
2. Methodological Context: Practice as Inquiry
Practice-based research in the arts operates on the premise that making is a mode of knowing. As Nelson (2013) observes, “practice-led research is a process through which thinking is materialised in the act of doing.” In this paradigm, the artwork is not simply an artefact but a form of embodied argument.
Phantom Mirror aligns with this approach but extends it into the post-photographic and posthuman domain. Its methodological question is not only what can be known through art but how can relational consciousness be enacted between human and machine. This emphasis on relation over representation situates the project within a lineage that includes Jo Spence’s performative phototherapy (1980s), Winnicott’s transitional space (1971), and contemporary theories of AI as co-creative partner (Bishop, 2022).
Where Spence used the camera as witness to psychic transformation, Phantom Mirror uses the algorithm as interlocutor. The AI interface functions as both mirror and method, what Estelle Barrett (2010) would call a “site of embodied knowledge.” Through iterative engagement, reflection becomes not an afterthought but an integral component of the creative process.
3. From Phototherapy to Generative Practice
The lineage of Phantom Mirror begins with Spence and Martin’s collaborative phototherapy, which treated photography as an instrument of psychic recovery. By re-staging trauma and internal conflict through performance and self-portraiture, participants re-authored their narratives and reclaimed agency. The photograph became what Spence called “a space for negotiating identity.”
In Phantom Mirror, this negotiation persists but without the camera’s indexical guarantee. The generative model produces images that are not of reality but about it. The absence of a physical referent reconfigures the therapeutic mechanism: instead of documenting a pre-existing subject, the process materialises internal states as visual metaphors.
This shift from representation to generation transforms the nature of evidence and authorship. The AI-generated image is not evidence of what happened but a manifestation of how something feels. It belongs to a continuum of projection, feedback, and reflection, a recursive process that parallels psychoanalytic transference. The subject becomes visible through her interaction with the algorithm, not through her likeness within the image.
4. Psychoanalytic Method: Transference, Containment, and Interpretation
Psychoanalysis offers a methodological template for this process. In the analytic setting, meaning is produced through dialogue, projection, and interpretation; the analyst functions as a mirror for unconscious material. Winnicott (1971) describes this as the “holding environment,” a space where the psyche can play safely with ideas and affects.
Phantom Mirror replicates this structure. The AI interface becomes the holding environment; the prompt functions as the free association; the generated image is the interpretive response. Each cycle of interaction constitutes a session in which affect is externalised, examined, and transformed.
The dynamic of transference, the displacement of emotion onto the analytic other, is mirrored in the artist–AI relationship. As Stephens (2025) notes, “projection is literalised: textual prompts are projected into the algorithmic other, which returns them transformed.” The AI thus occupies the dual role of analyst and canvas, absorbing psychic content and reflecting it through the logic of simulation.
This process also evokes Christopher Bollas’s (1987) concept of the transformational object, an encounter that alters the subject’s inner world. Through engagement with the algorithm, the artist experiences change, not by discovering fixed meaning but by inhabiting the instability of meaning itself.
5. The Generative Process as Epistemic Event
Central to the methodology is the idea that image generation is not illustrative but epistemic. Each prompt–response sequence is a miniature experiment in phenomenology and psychoanalysis, testing how emotion becomes form and how form reconfigures emotion.
This recursive method follows what philosopher Donald Schön (1983) termed “reflection-in-action,” a continuous feedback loop between thinking and doing. In Phantom Mirror, that loop includes the machine as co-thinker. The AI’s response introduces an element of unpredictability, which the artist must interpret and negotiate. Meaning arises in the interval between intention and outcome.
Baudrillard’s notion of simulation underpins this epistemic shift. In a world where “the map precedes the territory” (1981, p. 1), representation no longer mirrors reality but produces it. The Phantom Mirror process enacts this condition: by generating rather than capturing images, it reveals how the real is constructed through models and signs. Yet, paradoxically, this simulated process yields genuine insight. The algorithm’s indifference allows for the emergence of what Merleau-Ponty (1945) calls “the invisible,” those aspects of experience that become perceptible only through mediation.
The methodology, therefore, is not about mastering the tool but surrendering to its logic. The AI becomes a philosophical collaborator that forces the human subject to confront her own dependence on representation. The result is what might be called a mirror epistemology, knowledge produced through reflection upon reflection.
6. Enacted Philosophy: Thinking Through Doing
As discussed in the earlier essay on Baudrillard, Phantom Mirror operates as enacted philosophy, the performance of theory through practice. Baudrillard transformed philosophy into performance by living the implosion of meaning; he “made his abstract ideas into concrete behaviours” (Watlington, 2025). Similarly, Phantom Mirror turns that performative impulse into method.
In practice, this means that philosophical concepts such as simulation, hyperreality, and relational consciousness are not simply represented in the imagery; they are enacted in the act of image-making. The iterative dialogue with AI collapses the divide between theory and practice, subject and object. Each image is a residue of philosophical thought-in-action.
This enactment extends beyond Baudrillard into phenomenology and psychoanalysis. It embodies Merleau-Ponty’s (1945) claim that perception is an act, and Winnicott’s (1971) assertion that play is the foundation of creativity. The methodology fuses these traditions: it is both analytic and phenomenological, both play and performance.
To engage the AI is to think materially, to perform inquiry through interaction. The result is a hybrid form of research in which the philosophical is inseparable from the procedural, a living dialectic between reflection and generation.
7. Ethics, Reflexivity, and Relational Consciousness
Methodology is inseparable from ethics. In working with AI, questions of authorship, agency, and consent acquire new resonance. The algorithm has no consciousness, yet it mediates the artist’s most intimate psychological material. This tension necessitates a heightened reflexivity, a continual awareness of projection and illusion.
Relational consciousness functions here as an ethical principle. It recognises the mutual constitution of subject and system: neither the human nor the machine is autonomous; both exist within a network of relation. The artist’s responsibility lies not in controlling the AI but in maintaining awareness of the relational dynamics that emerge through its use.
This reflexivity mirrors the analytic stance of neutrality and curiosity. As in psychoanalysis, the goal is not mastery but understanding. The AI’s responses, often unexpected or imperfect, become opportunities for insight rather than failures of control. The ethical dimension of the methodology thus lies in its commitment to openness, to allowing meaning to unfold through relation rather than imposition.
Moreover, by situating itself within processes of care and mourning, Phantom Mirror aligns with what Bracha Ettinger (2006) calls the matrixial borderspace, a site of shared vulnerability and co-emergence. The generative image becomes a border-object, simultaneously self and other, human and machine, where empathy and differentiation coexist.
8. Generative Epistemology: Knowledge Through Relation
The term generative epistemology encapsulates the methodological ethos of Phantom Mirror. It refers to a mode of knowing that arises through generation, relation, and transformation rather than observation or representation.
In traditional epistemology, knowledge is propositional: it seeks truth claims about reality. In generative epistemology, knowledge is procedural: it emerges from interaction. This distinction parallels the shift from modernist to posthuman paradigms in art research. As Rosi Braidotti (2019) argues, posthuman knowledge is relational, embodied, and affective; it is “produced in the middle of things.”
Within Phantom Mirror, knowledge manifests as insight rather than statement. Each image carries an understanding that is felt before it is conceptualised. The process itself, the looping exchange of text, image, and reflection, becomes a model of cognition.
This approach also redefines documentation. The archive of prompts, images, and reflections is not supplementary data but the very substance of the research. It constitutes what Tim Ingold (2013) calls a “correspondence”, a living thread of thought and material intertwined.
In this sense, the methodology does not seek to explain how AI works technically but to reveal what it does phenomenologically and psychically. The generative system becomes a laboratory for studying relational consciousness, showing how meaning arises between human and non-human agents.
9. Reflexive Outcomes: The Artist as Researcher
Methodology and identity are intertwined. The artist-researcher in Phantom Mirror occupies multiple roles: creator, analyst, participant, and subject. This multiplicity mirrors the fragmentation of self experienced in mourning and resonates with Baudrillard’s idea of the subject as simulacrum.
Reflexivity, therefore, is not an add-on but the core of the method. Each stage of image generation prompts written reflection, what Schön (1983) terms “reflection-on-action.” These textual responses complete the research cycle, transforming practice into theory. The boundary between diary, analysis, and methodology dissolves, forming what Ellis (2004) calls “autoethnographic inquiry,” a mode of research that situates the personal within cultural and theoretical contexts.
Through this reflexive integration, Phantom Mirror contributes not only to artistic knowledge but to broader debates about consciousness and technology. It demonstrates that the act of making images with AI is simultaneously an act of thinking about being, an ontological investigation conducted through creative practice.
10. Conclusion: Methodology as Mirror
The methodology of Phantom Mirror can be summarised as a psychoanalytic-phenomenological practice of generative reflection. It treats the AI not as tool but as transitional partner, a collaborator in the process of thinking, feeling, and knowing. By uniting psychoanalytic structures, transference, holding, transformation, with generative processes, iteration, unpredictability, emergence, it creates a hybrid epistemology grounded in relation.
In this methodology, philosophy is not a framework applied to practice but a process enacted through it. Each prompt is both question and confession; each image is both answer and echo. The research unfolds as a series of encounters where theory, emotion, and technology converge.
As Stephens (2025) writes, “thinking happens through the image, not about it.” This statement encapsulates the methodological essence of Phantom Mirror: a practice that transforms psychoanalytic insight into generative form, turning the reflective act itself into knowledge.
Ultimately, the project proposes that in the age of AI, creative research must move beyond representation toward relation. The mirror of the title thus names both the method and its object: a reflective process in which the human and the simulated gaze into each other, and in that mutual recognition, new ways of knowing arise.
References
Barrett, E. (2010) Practice as Research: Approaches to Creative Arts Enquiry. London: I.B. Tauris.
Baudrillard, J. (1981) Simulacra and Simulation. Paris: Éditions Galilée.
Bishop, C. (2022) ‘Artificial Intelligence and the Ethics of Co-Creation.’ Third Text, 36(2), pp. 145–160.
Bollas, C. (1987) The Shadow of the Object: Psychoanalysis of the Unthought Known. London: Free Association Books.
Braidotti, R. (2019) Posthuman Knowledge. Cambridge: Polity Press.
Ettinger, B. (2006) The Matrixial Borderspace. Minneapolis: University of Minnesota Press.
Ellis, C. (2004) The Ethnographic I: A Methodological Novel about Autoethnography. Walnut Creek: AltaMira Press.
Ingold, T. (2013) Making: Anthropology, Archaeology, Art and Architecture. London: Routledge.
Merleau-Ponty, M. (1945) Phénoménologie de la perception. Paris: Gallimard.
Nelson, R. (2013) Practice as Research in the Arts: Principles, Protocols, Pedagogies, Resistances. Basingstoke: Palgrave Macmillan.
Schön, D. (1983) The Reflective Practitioner: How Professionals Think in Action. New York: Basic Books.
Spence, J. and Martin, R. (1988) Photography for Women: Visual Anthology and Phototherapy Projects. London: Camden Press.
Stephens, E. (2025) Phantom Mirror: AI Phototherapy and Relational Consciousness. Unpublished manuscript, University of Suffolk.
Watlington, E. (2025) ‘How Jean Baudrillard Turned Philosophy into Performance Art.’ ARTnews, October 2025.
Winnicott, D. W. (1971) Playing and Reality. London: Tavistock Publications.