Working papers › I’ll be your mirror · 2026
Disaggregating technological affordance from developed competence in AI-augmented practice
Commentary on human-AI engagement has concentrated on cognitive offloading, on productivity and output quality, and on AI literacy as prerequisite competence. This paper joins a fourth, emerging conversation: what practitioners learn through sustained, intentional engagement with generative AI. It reflects on the making of ten children’s picture books, and separates what the technology contributed from what the practitioner actually learned. The title borrows from the 1967 Velvet Underground song written by Lou Reed, for the paper’s central image: the generative system as a diagnostic mirror.
From Section 4 of the paper: a note of proportion is warranted. The third column documents competencies that were beginning to develop, not necessarily competencies that were fully acquired; the language of the table necessarily codifies what were, in practice, tentative and uneven gains.
From the paper: the Poltu Books are a series of ten children’s picture books for ages two to six, located in settings throughout India, featuring Poltu, an Indian lion-tailed macaque, and a small recurring cast of animal companions. The project involved a complete production cycle from concept to publication, required competencies across multiple domains (writing, visual composition, design, production and self-publishing), and was undertaken by a practitioner whose prior expertise was analytical rather than creative, making the trajectory of skill development more visible.
The practitioner’s primary motivation was not the adoption of generative AI for creative production as such, but the development of practical AI proficiency through sustained, self-directed projects; the creative domain was chosen as a vehicle for that learning, not as an end in itself. The illustrations are watercolour-style, generated through AI image generation and refined through a multi-stage production pipeline. The published e-books remain accessible via the FreemanJoyVentures author website and Instagram account. The Poltu Books remain early-stage work, in all their vulnerabilities: the value of this case lies not in the quality of the outputs themselves but in the visibility of the learning process they made possible.
The table’s analytical purpose is disaggregation: establishing a baseline of prior knowledge, isolating the technologies deployed, and only then asking what the practitioner actually learned. Reading across the layers reveals what was acquired through the process rather than what was already present at its outset. Select a dimension to see all three layers.
Collapsing what the technology contributed into what the practitioner learned, as contemporary discussion on AI and creativity frequently does, obscures the location of developed capability. Separating them allows a more explicit account of where new competence emerged: development was most pronounced where the practitioner had the least prior expertise (workflow, production, and visual design), took the form of articulation and refinement where prior dispositions were strongest (narrative), and occupied a middle ground of operational translation in publishing.
Scholarly commentary on human-AI engagement has focused on multiple themes, including cognitive offloading and ‘metacognitive laziness’; the measurement of human-AI productivity and output quality; and the identification of AI literacy as a set of prerequisite competencies. Another conversation is emerging, exploring capabilities developed through intentional interaction with generative AI. This paper contributes to that conversation through a structured practitioner reflection on the development of the Poltu Books, a series of ten children’s picture books. The reflection explores five dimensions (narrative and conceptual architecture; visual and structural design; workflow and production; publishing and positioning; holistic competence) across three analytical layers (prior knowledge; technologies; and developed practice). By disaggregating technological affordance from practitioner learning, the reflection delineates where new competence emerges in this case.
Grounded in established traditions of professional learning, including the Dreyfus model, Schön’s reflective practice, Polanyi’s account of tacit knowledge, and Kolb’s experiential learning cycle, the analysis identifies key dynamics: the compression of skill acquisition cycles, a recursive relationship between reflection-in-action and reflection-on-action, and the systematic surfacing of aspects of tacit knowledge through iterative demands for explicitness. Central to these dynamics is the reframing of the generative system as a diagnostic mirror that requires the practitioner to consistently justify and refine internal evaluative criteria. This process may provide a counter-dynamic to metacognitive laziness by requiring sustained diagnostic demand, ultimately locating authorship in the governing intelligence that orchestrates the multi-model workflow.
The Dreyfus model maps a five-stage progression from novice to expert, from rigid rule-following through conscious deliberation to fluid, intuitive performance grounded in accumulated experience. It supplies the arc against which the reflection locates its development: between advanced beginner and competent with publication of the Poltu Books.
Reflection-in-action occurs within the flow of practice, where the practitioner notices something unexpected, reframes the situation and adjusts in real time; reflection-on-action examines completed practice retrospectively. Both rest on knowing-in-action, the tacit knowledge that enables skilled performance.
Practitioners consistently know more than they can articulate through explicit rules or criteria. The relevance to AI-augmented practice is direct: generative AI cannot respond to unarticulated judgement, so the iterative workflow places a specific demand on the practitioner’s capacity to make tacit criteria explicit.
Learning proceeds through concrete experience, reflective observation, abstract conceptualisation and active experimentation, and each cycle can transform the practitioner’s understanding, not merely the output. AI-augmented workflows enact this cycle at considerable speed and frequency.
The Dreyfus model was developed observing practitioners whose learning unfolded over years. Generative AI accelerates the iterative cycle substantially, enabling hundreds of generation, evaluation and revision sequences within a single project, each a complete Kolb cycle. Whether this compression produces equivalent depth of acquisition or a qualitatively different form of it remains an open question.
In conventional practice the two reflective modes tend to operate in sequence. In AI-augmented workflows the speed of the production cycle allows reflection-in-action and reflection-on-action to alternate rapidly and repeatedly, each shaping the quality of the other, progressively externalising evaluative judgement.
Every iteration requires the practitioner to specify what they want, identify what is inadequate about an output, and articulate the criteria of evaluation. This demand for explicitness operates as a mechanism for converting aspects of tacit professional knowledge into explicit, communicable criteria: the practitioner who cannot specify sufficient criteria cannot direct the process.
Where generative AI responds poorly to unarticulated preferences, the practitioner is forced into heightened, rather than diminished, cognitive awareness. Every output that fails to meet the intended register functions as a diagnostic mirror, requiring the gap between result and goal to be explicitly defined. The reflective gains documented depended not on AI use alone but on a practice structure that sustained inquiry, iteration and evaluative scrutiny.
Authorship as the exercise of intent, constraint and selection, not necessarily the manual production of all text or image: the creative core resides in the governing intelligence that orchestrates the workflow, not in any single technology within it.
From Section 5 of the paper, which states the claim’s standing plainly: this is a practitioner’s observation, not a theoretical claim, and its generalisability remains to be tested. The reflection is single-practitioner and self-reported, with no external validation of the competencies claimed.
From the paper: this paper was written collaboratively using a suite of generative AI models including Claude Opus (Anthropic), NotebookLM, and Gemini (Google) as part of the same human-AI working relationship through which the Poltu Books themselves were produced. The practitioner was responsible for the reflection, the analytical framework, and all editorial judgements. The methodological consistency is deliberate: the paper constitutes an instance of the phenomenon it analyses. Acknowledging the constraints of higher education systems and traditional academic publishing timeframes in the face of unprecedented technological advances, the paper is released via the Social Science Research Network (SSRN) to enable immediate human readership and algorithmic discovery.
Almatrafi, O., Johri, A., & Lee, H. (2024). A systematic review of AI literacy conceptualization, constructs, and implementation and assessment efforts (2019--2023). Computers and Education Open, 6, 100173. https://doi.org/10.1016/j.caeo.2024.100173
Anderson, L. (2006). Analytic autoethnography. Journal of Contemporary Ethnography, 35(4), 373--395. https://doi.org/10.1177/0891241605280449
Artificiality Institute. (2026). The Artificiality Institute. https://www.artificialityinstitute.org/about
Benner, P. (1982). From novice to expert. AJN The American Journal of Nursing, 82(3), 402--407.
Chen, W. X., Srinivasan, S., & Zakerinia, S. (2025). Displacement or Complementarity? The Labor Market Impact of Generative AI. Working Paper 25-039. Harvard Business School. https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
Coeckelbergh, M. (2022). Self-improvement: Technologies of the soul in the age of artificial intelligence. Columbia University Press.
Creely, E., & Blannin, J. (2025). Creative partnerships with generative AI: Possibilities for education and beyond. Thinking Skills and Creativity, 56, 101727. https://doi.org/10.1016/j.tsc.2024.101727.
Dreyfus, H. L., & Dreyfus, S. E. (1986). Mind over machine: The power of human intuition and expertise in the era of the computer. Free Press.
Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489--530. https://doi.org/10.1111/bjet.13544.
Faraj, S., Perez-Torrents, J., Mantere, S., & Bhardwaj, A. (2025). A time for monsters: Organizational knowing after large language models. Strategic Organization. https://doi.org/10.1177/14761270251410675.
Fruits, E., & Stout, K. (2026). AI, Productivity, and Labor Markets: A Review of the Empirical Evidence. Available at SSRN 6323960.
Hartley, J. S., Jolevski, F., Melo, V., & Moore, B. (2026). The labor market effects of generative artificial intelligence (Working Paper). SSRN. https://ssrn.com/abstract=5136877.
Hou, Y. (2025). Thinking with ChatGPT: An autoethnographic inquiry into epistemic transformation in higher education. Artificial Intelligence in Educational Research, 2(1).
Illingworth, S. (2026). What is critical AI literacy? Slow AI [Substack]. https://theslowai.substack.com/p/what-is-critical-ai-literacy.
Institute for the Future of Work. (2025). Creative industries and GenAI: Good work impacts on a sector in transition. DOI: 10.5281/zenodo.15535400.
Kennedy, M. (2025). AI literacy review - March 11, 2025. AI Literacy Institute. https://ailiteracy.institute/ai-literacy-review-march-11-2025/
Kolb, D. A. (1984). Experiential learning: Experience as the source of learning and development. Prentice-Hall.
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X. H., Beresnitzky, A. V., ... & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task (Preprint). arXiv. https://doi.org/10.48550/arXiv.2506.08872.
Lee, M., Liang, P., & Yang, Q. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. Proceedings of the CHI Conference on Human Factors in Computing Systems. ACM. https://doi.org/10.1145/3706598.3713778.
Lin, Z., & Sohail, A. (2026). Recalibrating academic expertise in the age of generative AI. Patterns, 7(1), 101473. https://doi.org/10.1016/j.patter.2025.101473.
Maranger, A. P. (2025). Reimagining organizational learning in the AI era: A conceptual synthesis of Argyris and Schön, March, and Senge (Master’s thesis, Concordia University). https://spectrum.library.concordia.ca/id/eprint/995236/1/Maranger_MA_S2025.pdf
OECD. (2025). Empowering learners in the age of AI: An AI literacy framework for primary and secondary education. https://ailiteracyframework.org/#:~:text=The%20initiative%20contributes%20to%20the,along%20with%20AI%20literacy%20exemplars.
Panke, S. (2025). How can (A)I research this? An autoethnographic exploration of generative AI in research, teaching and instructional design. Journal of Teacher Education, 73(3), 230--244. https://doi.org/10.1177/00224871251325065.
Polanyi, M. (1966). The tacit dimension. Doubleday.
Reed, L. (1967). “I’ll be your mirror”. On The Velvet Underground & Nico (A. Warhol and T. Wilson, Producers) [Album]. The Velvet Underground. Verve Records.
Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676--688.
Rogers, T., & Carbonaro, M. (2025). From Understanding to Creating: Bridging AI Literacy and AI Fluency in K-12 Education. Journal of Teaching and Learning, 19(4), 20-38. https://doi.org/10.22329/jtl.v19i4.9420
Sarpong, P. (2025). From artificial to reflective intelligence: A synergistic design for cognitive and behavioral integration. CFPS Press.
Schön, D. A. (1983). The reflective practitioner: How professionals think in action. Basic Books.
UNESCO. (2024). AI competency framework for students.
Vallor, S. (2024). The AI mirror: How to reclaim our humanity in an age of machine thinking. Oxford University Press.
Wiles, F. (2025). Recursive cognition in practice: How AI dialogue generated and analyzed its own methodology. International Journal of Qualitative Methods, 24. https://doi.org/10.1177/16094069251381709.
Further papers will be added to this list as they are released. Each paper is published as its own page, with the full paper available for download as a PDF.