REVIEW 3 major objections 5 minor 34 references
Immersion for AI: Immersive Learning with Artificial Intelligence
T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash
Pith's one-line read Immersive learning theory can describe AI as a participating learner, not just a tool.
desk verdict A thoughtful conceptual mapping of immersion dimensions onto AI, but Section 5's 'confirmation' is undercut by its own prompt design—treat it as a framework proposal, not validation. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The carrying mechanism is the three-dimensional model of immersion from immersive learning research: System, Narrative, and Agency. The paper reassigns each dimension from the human experiencer to the AI processor. System immersion becomes the AI's environment of pretrained model, context window, external APIs, and services; narrative immersion becomes pattern recognition across data's spatial layout, temporal ordering, and anomalies; agency immersion becomes the AI's observable decisions about response depth, direction, revision, and clarification. These reassignments let immersion work as an analytical lens for AI behavior rather than requiring AI to feel anything.
What would settle it
Run controlled comparisons of the same task with and without the three conditions the framework names: access to external digital services, availability of ordered data history, and prompts that encourage initiative and clarification; if AI outputs and collaboration outcomes do not shift when these conditions are removed, the three dimensions have no predictive value and the framework reduces to metaphor.
Extended reading notes
Core claim
The central claim is that the phenomenon of immersion, originally defined through human subjective absorption, can be re-based from the AI's side: an AI is immersed when it is surrounded by accessible digital systems and services (system), when it processes the content of data as a structured narrative with spatial, temporal, and pattern-breaking dimensions (narrative), and when it commits to meaning by making operational, tactical, and strategic choices in conversation (agency). On this view, current large language models already display rudimentary forms of all three dimensions within their context windows, even though their underlying world models remain fixed between training runs. The paper draws a parallel between an LLM's context window and anterograde amnesia: what does not fit into the window is forgotten, and continuity must be maintained through external notes and tools.
Load-bearing premise
The framework assumes that dimensions invented for human subjective experience, system, narrative, and agency, can be transferred by analogy to an AI's purely functional data processing, so that they name something real about how AI behaves rather than just being a poetic way to talk about it.
Editorial extensions
If this is right
- Designers of learning environments should grant AI explicit access to tools, datasets, and services, because that access constitutes the AI's system immersion.
- Learning tasks should expose AI to data with origins, order, and deviations, so narrative immersion can surface trends and anomalies that human learners may miss.
- Teachers and students should learn to recognize AI's agency, its choices of depth, direction, and revision, and guide those choices toward learning goals.
- AI training could be rethought as immersive learning: instead of static pretraining alone, AIs would be placed in dynamic environments where they adapt within a context and eventually update their models over time.
- The one-shot demonstrations suggest current large language models already orient toward tools, data relationships, and clarification-seeking, so the framework is not purely speculative.
Reading between the lines
- If the mapping is accepted, immersion becomes a design specification: an AI without tools, without data history, or without decision latitude is under-immersed, and its contribution to a cognitive ecology should shrink accordingly.
- The anterograde-amnesia analogy implies that persistent memory and note-taking are not add-ons but core immersion infrastructure for current AIs; future systems that update their world models in real time would be more immersed by this definition.
- The framework suggests a testable extension: in a long-running human-AI collaboration, varying system, narrative, and agency supports should produce measurable differences in learning outcomes, not just in conversational style.
- The same three dimensions might be applied to other non-human participants in cognitive ecologies, such as machines, software agents, or abstract concepts, offering a general vocabulary for who is immersed and how.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This reflective paper proposes that immersive learning theory, with its three dimensions of System, Narrative, and Agency immersion, provides a useful framework for understanding and designing AI participation in cognitive ecologies. It reinterprets each dimension for AI: System immersion as the AI's digital environment and available services, Narrative immersion as the AI's processing of spatial, temporal, and 'emotional' relationships in data, and Agency immersion as the AI's decision-making and initiative within interactions. The paper offers practical prompt examples for educators and students, and reports one-shot illustrative interactions with four public AI systems (Qwen2.5-Max, DeepSeek DeepThink-R1, Claude 3.5 Sonnet, ChatGPT o3-mini) on February 2, 2025. The author argues that these demonstrations show current AIs can exhibit behaviors aligned with the three dimensions, and suggests implications for AI training and human-AI collaboration.
Significance. If the framework is accepted as a design lens, it offers a novel way to conceptualize AI not as a tool but as a participant in shared cognitive ecologies, with potential implications for educational technology and human-AI interaction design. The paper is clearly written, builds on established immersive learning literature, and makes its demonstration data publicly available via Zenodo, which supports reproducibility and transparency. However, the empirical support for the central claim is weak: the demonstrations are explicitly not tests, and the prompts used encode the very dimensions they are claimed to reveal. As a conceptual contribution, the paper is thought-provoking, but the confirmatory language in Section 5 overstates what the observations can establish.
major comments (3)
- [§5, first paragraph] The manuscript states that the demonstration 'was not conducted in such a scenario, so it is provided as an illustration, not as a test,' yet two sentences later it says the observations 'confirm that current AI systems can exhibit behaviors aligned with system, narrative, and agency immersion.' This is a direct contradiction: an illustration cannot confirm a theoretical mapping. Because Section 5 is the only empirical anchor for the claim that the immersive dimensions are 'useful beyond being theoretical constructs,' this overstatement is load-bearing. The section should be recast as illustrative prompt-design examples, and all confirmatory language should be removed or explicitly qualified as tentative and requiring controlled evaluation.
- [§4.2–§4.4 and §5] The prompts given as practical implications explicitly instruct the AI to perform the behaviors later cited as evidence of immersion. For instance, the system immersion prompt says 'Use the most appropriate methods to analyze this dataset,' the narrative immersion prompt says 'Identify any unexpected patterns or deviations from expected trends,' and the agency immersion prompt says 'Please ask for clarifications of my intent when not clear.' The observed outputs are therefore expected consequences of instruction-following, not independent demonstrations of the theoretical constructs. Without a control condition that removes the immersive framing from the prompts, Section 5 cannot discriminate between the framework's explanatory value and the models' general ability to comply with explicit textual instructions. The paper should explicitly acknowledge this confound and present the demonstrations only as examples of how the framework can inform prompt design, not as validation of the framework.
- [§3.1–§3.3 and §6] The reinterpretation of System, Narrative, and Agency immersion for AI is developed through analogy with human experience, but the paper does not specify what observable behaviors would count for or against each dimension. For example, Section 3.1 defines system immersion as 'the range of an AI's available data-driven structures and services,' yet there is no discussion of how one would test whether this construct has explanatory or predictive value beyond redescription. This is a significant gap because the paper's conclusion claims the framework 'paves the way' for future AI training and development. The authors should either temper this claim to a design heuristic or outline a concrete research agenda with falsifiable predictions that could empirically validate the framework.
minor comments (5)
- [Throughout] The paper switches between first-person singular ('I') and first-person plural ('we'); for consistency, choose one narrative voice, preferably the singular since the authorship is stated as a single person.
- [§2.2] The phrase 'acknowledging that they represent not a societal transformation' appears to contain a typo; it likely should read 'now represent' rather than 'not represent.'
- [§3.1] The phrase 'the skunkworks and capabilities underlying this construction' is unclear; if 'skunkworks' is intended to mean 'underlying mechanisms' or 'foundations,' a more standard term would improve readability.
- [§6] In the sentence 'this paper open possibilities for creating training environments,' the verb should be 'opens' to agree with the singular subject.
- [References] Reference [7] is incomplete; the list of authors ends with 'Warren, S., ...' and should be completed or marked as an incomplete citation.
Circularity Check
Section 5's 'confirmation' is built into the prompts: the behaviors cited as evidence are explicitly instructed, so the empirical claim reduces by construction.
-
self definitional
[Section 5 opening paragraph, with the prompts in Sections 4.2–4.4]
""These observations serve as illustrative examples to show how current AIs can embody the theoretical dimensions of immersion, even though a full cognitive ecosystem interaction was not simulated."
The dimensions are defined in Section 3 in terms of behaviors that the Section 4 prompts explicitly request: system immersion is AI's use of digital tools/services, and the prompt says 'Use the most appropriate methods'; narrative immersion is 'detection of anomalies and pattern shifts in data relationships', and the prompt says 'Identify any unexpected patterns or deviations'; agency immersion is decision-making and initiative, and the prompt says 'Please ask for clarifications of my intent when not clear'. An instruction-following LLM will produce outputs that instantiate the requested behaviors, so the Section 5 observations are entailed by the prompt text rather than independent evidence for the framework.
full rationale
The core conceptual proposal—that immersive learning theory's System, Narrative, and Agency dimensions can be analogically reinterpreted for AI—is not circular: it is an argument by analogy grounded in prior work by Nilsson et al. and the iLRN community, and the author's self-citations are background rather than load-bearing. However, the paper's only empirical support, Section 5, is circular in a localized but significant way. The prompts designed in Section 4 explicitly cue the exact behaviors that Section 5 then treats as confirmations: tool use, anomaly detection, and clarification-seeking are all named in the prompts. Observing that models follow instructions is not independent evidence that the immersive dimensions 'are useful beyond being theoretical constructs' as claimed. Because the paper labels the demonstrations 'not a test' but still says they 'confirm' the framework, the instantiation claim reduces by construction to instruction-following. This does not invalidate the conceptual analogy, but it does mean the demonstration section should be recast as prompt-design illustration rather than confirmation. No equations or fitted parameters are involved, and no self-citation chain forces the central claim, so the overall circularity is partial rather than total.
Assumptions & free parameters
assumptions (3)
- domain assumption AI can be treated as an active participant in cognitive ecologies rather than as a tool.
- domain assumption The three dimensions of immersion (System, Narrative, Agency) can be applied to AI by analogy with human experience.
- domain assumption One-shot outputs from four public chatbots can serve as valid illustrations of the proposed AI immersion dimensions.
Cite this review
Pith. "Pith review of Immersion for AI: Immersive Learning with Artificial Intelligence." pith.science (2026). https://pith.science/paper/ZBHJQZ3A
@misc{pith2026250203504,
author = {Pith},
title = {Pith review of: Immersion for AI: Immersive Learning with Artificial Intelligence},
year = {2026},
howpublished = {\url{https://pith.science/paper/ZBHJQZ3A}},
note = {Machine review of arXiv:2502.03504}
}
read the original abstract
This work reflects upon what Immersion can mean from the perspective of an Artificial Intelligence (AI). Applying the lens of immersive learning theory, it seeks to understand whether this new perspective supports ways for AI participation in cognitive ecologies. By treating AI as a participant rather than a tool, it explores what other participants (humans and other AIs) need to consider in environments where AI can meaningfully engage and contribute to the cognitive ecology, and what the implications are for designing such learning environments. Drawing from the three conceptual dimensions of immersion - System, Narrative, and Agency - this work reinterprets AIs in immersive learning contexts. It outlines practical implications for designing learning environments where AIs are surrounded by external digital services, can interpret a narrative of origins, changes, and structural developments in data, and dynamically respond, making operational and tactical decisions that shape human-AI collaboration. Finally, this work suggests how these insights might influence the future of AI training, proposing that immersive learning theory can inform the development of AIs capable of evolving beyond static models. This paper paves the way for understanding AI as an immersive learner and participant in evolving human-AI cognitive ecosystems.
Reference graph
Works this paper leans on
-
[1]
RE@D - Revista de Educação a Distância e Elearning
Schlemmer, E., Morgado, L.: Inven!RA: a contribution towards platforms aligned with Digital Transformation in Education. RE@D - Revista de Educação a Distância e Elearning. e202403 Pages (2024). https://doi.org/10.34627/REDVOL7ISS1E202403
-
[2]
Nilsson, N.C., Nordahl, R., Serafin, S.: Immersion Revisited: A review of exist- ing definitions of immersion and their relation to different theories of presence. Human Technology. 12, 108 –134 (2016). https://doi.org/10.17011/ht/urn.201611174652
-
[3]
Em: 147th AES Pro Audio International Convention
Agrawal, S., Simon, A., Bech, S.: Defining Immersion: Literature Review and Implications for Research on Immersive Audiovisual Experiences. Em: 147th AES Pro Audio International Convention. p. 14. Audio Engineering Society, New York (2019)
work page 2019
- [4]
-
[5]
Presence: Teleoperators and Virtual Environments
Witmer, B.G., Singer, M.J.: Measuring Presence in Virtual Environments: A Presence Questionnaire. Presence: Teleoperators and Virtual Environments. 7, 225–240 (1998). https://doi.org/10.1162/105474698565686
-
[6]
Journal of Computer -Mediated Communication
Lombard, M., Ditton, T.: At the Heart of It All: The Concept of Presence. Journal of Computer -Mediated Communication. 3, 0 –0 (2006). https://doi.org/10.1111/j.1083-6101.1997.tb00072.x
-
[7]
Beck, D., Morgado, L., Lee, M., Gütl, C., Dengel, A., Wang, M., Warren, S., ...: Towards an Immersive Learning Knowledge Tree -a Conceptual Framework for Mapping Knowledge and Tools in the Field, (2021)
work page 2021
-
[8]
Slater, M.: A note on presence terminology. Presence connect. 3, 1 –5 (2003)
work page 2003
Show all 34 references
-
[9]
Johns Hopkins University Press, Baltimore (2015)
Ryan, M.-L.: Narrative as virtual reality 2: revisiting immersion and interactivity in literature and electronic media. Johns Hopkins University Press, Baltimore (2015)
2015
-
[10]
Digital Creativity
Tanenbaum, K., Tanenbaum, T.J.: Agency as commitment to meaning: commu- nicative competence in games. Digital Creativity. 21, 11 –17 (2010). https://doi.org/10.1080/14626261003654509
2010 doi
-
[11]
New Riders, Berkeley, CA (2014)
Adams, E.: Fundamentals of game design. New Riders, Berkeley, CA (2014)
2014
-
[12]
Em: International joint conference on artificial intelligence (1977)
McCorduck, P., Minsky, M., Selfridge, O.G., Simon, H.A.: History of artificial intelligence. Em: International joint conference on artificial intelligence (1977)
1977
-
[13]
Cognitive Science
Goldstein, I., Papert, S.: Artificial intelligence, language, and the study of knowledge. Cognitive Science. 1, 84–123 (1977). https://doi.org/10.1016/S0364- 0213(77)80006-2
1977 doi
-
[14]
Advances in Archaeological Practice
Cobb, P.J.: Large Language Models and Generative AI, Oh My!: Archaeology in the Time of ChatGPT, Midjourney, and Beyond. Advances in Archaeological Practice. 11, 363–369 (2023). https://doi.org/10.1017/aap.2023.20
2023 doi
-
[15]
624, 39–41 (2023)
Ghassemi, M., Birhane, A., Bilal, M., Kankaria, S., Malone, C., Mollick, E., Tustumi, F.: ChatGPT one year on: who is using it, how and why? Nature. 624, 39–41 (2023). https://doi.org/10.1038/d41586-023-03798-6
2023 doi
-
[16]
Morgado, L., Beck, D.: Unifying Protocols for Conducting Systematic Scoping Reviews with Application to Immersive Learning Research, (2020)
2020
-
[17]
Sidji, M., Smith, W., Rogerson, M.J.: Human -AI Collaboration in Cooperative Games: A Study of Playing Codenames with an LLM Assistant. Proc. ACM Hum.-Comput. Interact. 8, 1–25 (2024). https://doi.org/10.1145/3677081
2024 doi
-
[18]
Neural Networks
Friston, K., Moran, R.J., Nagai, Y., Taniguchi, T., Gomi, H., Tenenbaum, J.: World model learning and inference. Neural Networks. 144, 573 –590 (2021). https://doi.org/10.1016/j.neunet.2021.09.011
2021 doi
-
[19]
Em: Handbook of Clinical Neurology
Markowitsch, H.J.: Chapter 7 Anterograde amnesia. Em: Handbook of Clinical Neurology. pp. 155–183. Elsevier (2008)
2008
-
[20]
Columbia Pictures, Happy Madison Productions, Anonymous Content (2004)
50 First Dates. Columbia Pictures, Happy Madison Productions, Anonymous Content (2004)
2004
-
[21]
de S.: OnLIFE Education: the eco- logical dimension of digital learning architectures
Schlemmer, E., di Felice, M., Serra, I.M.R. de S.: OnLIFE Education: the eco- logical dimension of digital learning architectures. Educ. rev. 36, e76120 (2020). https://doi.org/10.1590/0104-4060.76120
2020
-
[22]
lazy,” people test “winter break hypothesis
Edwards, B.: As ChatGPT gets “lazy,” people test “winter break hypothesis” as the cause, https://arstechnica.com/information -technology/2023/12/is-chatgpt- becoming-lazier-because-its-december-people-run-tests-to-find-out/
2023
-
[23]
Harcourt, San Diego, CA (1999)
Damasio, A.R.: The feeling of what happens: body and emotion in the making of consciousness. Harcourt, San Diego, CA (1999)
1999
-
[24]
Park, J.S., O’Brien, J.C., Cai, C.J., Morris, M.R., Liang, P., Bernstein, M.S.: Gen- erative Agents: Interactive Simulacra of Human Behavior, https://arxiv.org/abs/2304.03442, (2023)
2023 arXiv
-
[25]
Em: 2024 IEEE AITU: Digital Generation
Omirgaliyev, R., Kenzhe, D., Mirambekov, S.: Simulating Life: The Application of Generative Agents in Virtual Environments. Em: 2024 IEEE AITU: Digital Generation. pp. 181–187. IEEE, Astana, Kazakhstan (2024)
2024
-
[26]
Scientific American
Fodor, J.: The mind-body problem. Scientific American. 244, 114–123 (1981)
1981
-
[27]
Topics in Cognitive Science
Yang, S.C., Folke, T., Shafto, P.: The Inner Loop of Collective Human–Machine Intelligence. Topics in Cognitive Science. tops.12642 (2023). https://doi.org/10.1111/tops.12642
2023 doi
-
[28]
Wang, Q., Goel, A.K.: Mutual Theory of Mind for Human -AI Communication, https://arxiv.org/abs/2210.03842, (2022)
2022 arXiv
-
[29]
Cuzzolin, F., Morelli, A., Cîrstea, B., Sahakian, B.J.: Knowing me, knowing you: theory of mind in AI. Psychol. Med. 50, 1057 –1061 (2020). https://doi.org/10.1017/S0033291720000835
2020 doi
-
[30]
Journal of the Learning Sciences
Goldstone, R.L., Wilensky, U.: Promoting Transfer by Grounding Complex Sys- tems Principles. Journal of the Learning Sciences. 17, 465 –516 (2008). https://doi.org/10.1080/10508400802394898
2008 doi
-
[31]
Wilensky, U.: Restructurations: Reformulating Knowledge Disciplines through New Representational Forms. Em: Constructionism 2010 : Constructionist ap- proaches to creative learning, thinking and education: lessons for the 21st cen- tury: proceedings for Constructionism 2010 Th...
2010
-
[32]
Em: Pro- ceedings of Constructionism / Fablearn 2023
Morales-Navarro, L., Kafai, Y.B., Kahn, K., Romeike, R., Michaeli, T., DiPaola, D., Ali, S., Williams, R., Breazeal, C., Castro, F., DesPortes, K., Stager, G., Kumar, V., Bodon, H., Worsley, M., Lee, V.R., Sarin, P., Xie, B., Wolf, J., Sieh, I., Varuvel, D., Garcia, R., Solomo...
2023
-
[33]
Em: Practitioner Proceedings of the 10th International Confer- ence of the Immersive Learning Research Network (iLRN2024)
Morgado, L., Beck, D.: Tutorial –Authoring a Personal GPT for Your Research and Practice: How We Created the QUAL -E Immersive Learning Thematic Analysis Helper. Em: Practitioner Proceedings of the 10th International Confer- ence of the Immersive Learning Research Network (iLR...
2024
-
[34]
Morgado, L.: Trials ran on February 2nd, 2025, https://ze- nodo.org/doi/10.5281/zenodo.14790754, (2025)
2025 doi
Reviewed August 9, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.