AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance
Open accessThe DOI matches its Crossref record; 16 of the 16 references it cites were found in a registry or an index; 6 of the 21 sources its claims cite were checked for support.
Abstract
As AI becomes a prism through which we reflect, see, and make sense of the world, the way we create creative, transcendent experiences around AI can shape our relationship to it. Drawing inspiration from the ritual structures of Spiritualist séances and creative art-making séances of Hilma af Klint, we present reflections from a series of participatory experiments we called AI Séances. These gatherings brought together artists, technologists, and spiritual practitioners to engage with generative models in contexts shaped by ritual, randomness, and collaborative interpretation. We found that creative production with AI can yield transcendent user experiences (TUX), different communities bring distinct interpretive lenses to AI outputs, and increased technical control can paradoxically diminish serendipity and transcendence. Through our experiences, we suggest that reclaiming interpretive agency over AI outputs in the creative and spiritual context, rather than treating models as machines that produce answers, opens up new avenues for critical and creative engagement with these technologies and is critical to preserving our humanity. The AI Séance offers a model for human-centered interaction with generative systems where magic lies not in the machine’s capabilities, but in our collective ability to create meaning.
Overview
CoChat Summary
TL;DR
The authors recount a two-year artistic-research practice, “AI Séance,” that stages collective text-to-image generation as a divination-like ritual rather than an answer-giving oracle. They argue that generative AI can support transcendent user experience (TUX) when participants co-interpret outputs, but that too much system control diminishes that potential.
Key findings
- Ritual and creative production using generative AI can yield a “transcendent user experience” (TUX): in the first séance, four participants generated 35 images over 2.5 hours and reported the experience as transcendent due to serendipitous, emergent imagery.
- Different communities bring different interpretive lenses: at a Goldsmiths symposium with over 40 participants, prompts and interpretations shifted toward occult themes, while at Harvard Divinity School’s Interspecies Dialogues, herbalist and animist participants steered prompts toward plants, fungi, trees, and roots.
- Too much technological determinism hinders transcendence: adding an LLM “cognitive” layer to synthesize prompts increased control and fidelity but became boring and produced “AI slop”; reverting to direct text-to-image prompting restored the “unexpected jumpiness and lack of control” the authors identify as core to TUX.
- The practice reframes AI as a hermeneutic technology: outputs are treated as prompts for collective reflection, not as privileged answers from a medium or machine.
Methods
- Qualitative, practice-based artistic research: the authors, a group of computer scientists and spiritual creatives, recount a series of AI Séances over roughly two years, using Stable Diffusion via Replicate and group co-interpretation.
- Séance 1 (November 2022): four participants around a table, inspired by Hilma af Klint’s art-making séances, generated 35 images over 2.5 hours. Séance 2 (June 2023): demonstration at Goldsmiths’ “21st Century Magic and Spirituality in Media and Culture” symposium with over 40 participants, plus a séance at Harvard Divinity School’s Interspecies Dialogues Conversation Group.
- Séance 3 (summer residency at Stochastic Labs, Berkeley): used an antique bingo cage to randomize model random seeds and prompt-contributor order, and tested an LLM “cognitive” layer for prompt synthesis before reverting to a pure text-to-image stack.
Limitations
- Stated: the authors discuss limits to the divination analogy—generative AI outputs come from vast but particular output spaces shaped by often unknown training data and methods; traditional divination systems have defined rules, abstracted symbols, and conventional guidebooks, whereas text-to-image outputs are more open-ended and lack discrete symbols or a guidebook.
- Evident: the findings rest on the authors’ subjective recounts and small, self-selected workshop groups rather than a controlled study or validated measure of TUX.
- Evident: the paper does not report systematic participant data, comparison conditions, or long-term outcomes beyond the authors’ own artistic follow-up work.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures






Verification
Claims the paper makes
5 Valid · 2 Not valid · 13 Not checked · 51 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“For one, the text or image outputs from generative AI come from vast but particular output spaces, shaped and constrained by the (often unknown) source data and methods used for training and post-tuning these systems (Ugander and Epstein, 2024).”
§ AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance > Discussion
What the sources say
- Supportedcites Ugander and Epstein, 2024 source
The 'Reclaiming Randomness in the Age of Generative AI' section (para. on sampling) states that generative models are high-dimensional probability distributions whose text (GPT-4) and image (Midjourney) outputs are sampled realizations, and immediately after that that 'the space of possibility contained within these generators does not represent the full "Library of Babel," but rather one peculiar corner filtered through both individual and collective subjectivities' — i.e. vast but particular, shaped and constrained output spaces. The same passage ties the constraint to 'the annotations codified in the training data' and notes that 'technical communities and the art world do not have clear answers' to whose values these reflect, which carries the claim's '(often unknown)' qualifier at paraphrase level rather than as an explicit statement about undisclosed data. The Introduction adds the post-tuning element: fine-tuning methods such as RLHF have 'refined their probability distributions' so that maximum-likelihood output now beats randomized output. (LLM, conf=0.82)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Recent work has made explicit the ways in which generative AI may similarly be a medium of randomness (Ugander and Epstein, 2024), which hints at the possibility of using generative AI for cybermancy, divination with computers.”
§ AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance > Séance 3: randomness and how technological determinism leads away from TUX
What the sources say
- Supportedcites Ugander and Epstein, 2024 source
In “Reclaiming Randomness in the Age of Generative AI,” the authors describe generative AI outputs as samples from probability distributions and say users can experience their randomness by repeating a prompt. The subsequent possibility of cybermancy is the citing authors’ inference, not a claim attributed to this source. (LLM, conf=0.93)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Looking back, our prompts were strongly inspired by the physical surroundings of the 'séance'.”
§ AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance > Introduction
boostedstates a causeSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Participants craft their own understandings of the output, much as one might uncover hidden connections in the frameworks of Tarot or I Ching.”
§ AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance > Discussion
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We found that this intervention dramatically increased the level of control we had over the image output, allowing us to explore the worlds of images we created with a degree of fidelity akin to moving through levels of a video game.”
§ AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance > Séance 3: randomness and how technological determinism leads away from TUX
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
25 more claims
Create a free account to see all 71 claims and to verify the 4 references still open.
- Probably not a substantial claim. 26
- It may state two separate claims as one. 21
Citations
Cites
Generative artificial intelligence enhances creativity but reduces the diversity of novel content
Doshi, A., & Hauser, O. (2023) · doi:10.2139/ssrn.4535536 · Cited by 1,350
Art and the science of generative AI
Epstein, Z., Hertzmann, A., the Investigators of Human Creativity, Akten, M., Farid, H., Fjeld, J., Frank, M. R., Groh, M., Herman, L., Leach, N., et al. (2023) · Science · doi:10.1126/science.adh4451 · Cited by 788
Self-citationSituating methods in the magic of Big Data and AI
Elish, M. C., & boyd, D. (2018) · Communication Monographs · doi:10.1080/03637751.2017.1375130 · Cited by 409
The impact of generative artificial intelligence on socioeconomic inequalities and policy making
Capraro, V., Lentsch, A., Acemoglu, D., Akgun, S., Akhmedova, A., Bilancini, E., Bonnefon, J.-F., Brañas-Garza, P., Butera, L., Douglas, K. M., et al. (2024) · PNAS Nexus · doi:10.1093/pnasnexus/pgae191 · Cited by 371
Enchanted Determinism: Power without Responsibility in Artificial Intelligence
Campolo, A., & Crawford, K. (2020) · Engaging Science, Technology, and Society · doi:10.17351/ests2020.277 · Cited by 254
Goldstein, J. A., Sastry, G., Musser, M., DiResta, R., Gentzel, M., & Sedova, K. (2023) · arXiv · doi:10.48550/arxiv.2301.04246 · Cited by 152
AI Companions Reduce Loneliness
De Freitas, J., Uğuralp, A. K., Uğuralp, Z., & Puntoni, S. (2024) · doi:10.2139/ssrn.4893097 · Cited by 80
Let Us Say What We Mean: Towards Operational Definitions for Techno-Spirituality Research
Buie, E. A. (2019) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3290607.3310426 · Cited by 44
Greenwood, S. (2020) · Routledge eBooks · doi:10.4324/9781003086864 · Cited by 39
Conjuring algorithms: Understanding the tech industry as stage magicians
Nagy, P., & Neff, G. (2024) · New Media & Society · doi:10.1177/14614448241251789 · Cited by 26
Research works printed but not linked: 4
- Braude A (2001) Radical Spirits: Spiritualism and Women’s Rights in Nineteenth-Century America. Bloomington, IN: Indiana University Press.
- Davis E (2015) Techgnosis: Myth, Magic, and Mysticism in the Age of Information. Berkeley, CA: North Atlantic Books.
- Kelly L (2024) Could congress leverage AI to help restore faith in US Democracy? Tech Policy Press 18 September. Available at: https://techpolicy.press/could-congress-leverage-ai-to-help-restore-faith-in-us-democracy
- Tambiah SJ (1990) Magic, Science and Religion and the Scope of Rationality. Cambridge: Cambridge University Press.
Cited by
Forage: Understanding LLM-facilitated sensemaking of conversation data
Schroeder, H., Beeferman, D., Detwiller, M. E., Dimitrakopoulou, D., & Roy, D. K. (2026) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3772318.3791640
Self-citationPonochevnyi, N., Kim, Y., Brudno, M., & Kuzminykh, A. (2026) · Designing Interactive Systems Conference · doi:10.1145/3800645.3813033 · Cited by 1
“Artists may treat prompts themselves as creative artifacts, deliberately work with unexpected model behaviors, or use stochasticity and collective interpretation to support discovery, reflection, or transcen-dence [9, 49, 53, 61].”
Metadata
- Authors
- Schroeder, H., Smith, A., & Epstein, Z.
- Affiliations
- MIT Media Lab (US)
- Queen Mary University of London (GB)
- Massachusetts Institute of Technology (US)
- Corresponding author
- Schroeder, Hope
- Venue
- European Journal of Cultural Studies
- Publisher
- SAGE Publications
- Published
- 2025-07-17
- ISSN
- 1367-5494 (ISSN-L)
- 1460-3551
- OpenAlex
- W4415776280
- Cited by
- 3
- References
- 12
- Citations by year
- 2026: 3
- FWCI
- 13.85
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by-nc
- creativecommons.org/licenses/by-nc/4.0/ (unspecified)
- journals.sagepub.com/page/policies/text-and-data-mining-license (tdm)
- Locations
- European Journal of Cultural Studies (publishedVersion, cc-by-nc)
- DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc)
- Funders
- Stochastic Labs for Ziv Epstein and Amy Smith’s summer residencies working on AI Séances in 2024
- Topics
- Social Sciences › Arts and Humanities › Visual Arts and Performing Arts › Art, Technology, and Culture
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Digital Media and Visual Art
- Keywords
- generative AI, spiritualism, séances, Hilma af Klint, ritual, randomness, serendipity, collaborative interpretation, human-centered AI, participatory art, human-AI interaction, meaning-making
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-11-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance. https://app.cochat.ai/papers/10.1177/13675494251351223
- The paper
- Schroeder, H., Smith, A., & Epstein, Z. (2025). AI Séance: Recounts from designing artificial intelligence for transcendence, interpretive lenses and chance. European Journal of Cultural Studies. https://doi.org/10.1177/13675494251351223
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.