The DOI matches its Crossref record; 59 of the 62 references it cites were found in a registry or an index; 23 of the 35 sources its claims cite were checked for support.
Generative midtended cognition and Artificial Intelligence: thinging with thinging things
Open accessAbstract
This paper introduces the concept of “generative midtended cognition”, that explores the integration of generative AI technologies with human cognitive processes. The term “generative” reflects AI’s ability to iteratively produce structured outputs, while “midtended” captures the potential hybrid (human-AI) nature of the process. It stands between traditional conceptions of in tended creation, understood as steered or directed from with in, and ex tended processes that bring exo-biological processes into the creative process. We examine the working of current generative technologies (based on multimodal transformer architectures typical of large language models like ChatGPT) to explain how they can transform human cognitive agency beyond what the conceptual resources of standard theories of extended cognition can capture. We suggest that the type of cognitive activity typical of the coupling between a human and generative technologies is closer (but not equivalent) to social cognition than to classical extended cognitive paradigms. Yet, it deserves a specific treatment. We provide an explicit definition of generative midtended cognition in which we treat interventions by AI systems as constitutive of the agent’s intentional creative processes. Furthermore, we distinguish two dimensions of generative hybrid creativity: 1. Width: captures the sensitivity of the context of the generative process (from the single letter to the whole historical and surrounding data), 2. Depth: captures the granularity of iteration loops involved in the process. Generative midtended cognition stands in the middle depth between conversational forms of cognition in which complete utterances or creative units are exchanged, and micro-cognitive (e.g. neural) subpersonal processes. Finally, the paper discusses the potential risks and benefits of widespread generative AI adoption, including the challenges of authenticity, generative power asymmetry, and creative boost or atrophy.
Overview
CoChat Summary
TL;DR
The paper proposes “generative midtended cognition” to describe hybrid human–AI creative/cognitive processes in which AI-generated interventions become constitutive of an agent’s intentional generation of a product while the agent retains some sense of agency or authorship. It argues this phenomenon is not fully captured by extended cognition or social cognition and maps it along dimensions of width (context sensitivity) and depth (granularity of iterative loops), while discussing risks such as authenticity, power asymmetry, and creative atrophy.
Key findings
- The paper defines generative midtension as follows: given a cognitive agent X, a generative system Y, and cognitive product Z, midtension takes place when generative interventions produced by Y become constitutive of the intentional generation of Z by X, whereby X keeps some sense of agency or authorship over Z.
- It distinguishes generative midtended cognition from standard extended cognition and from social cognition: generative AI systems have derived intentionality and produce purpose-structured outputs, but they are not autonomous agents, so interaction with them is closer to social cognition than classical extended cognition yet deserves separate treatment.
- It proposes two dimensions of generative hybrid creativity: Width, the sensitivity of the generative process to context (from a single letter to the whole historical and surrounding data), and Depth, the granularity of iteration loops (from final products and conversational/creative units down to perceptual-action tokens, sensorimotor dynamics, proprioceptive body, and neural scales). Midtended cognition is placed between conversational cognition and micro-cognitive subpersonal processes.
- It discusses risks and benefits including authenticity, generative power asymmetry, creative boost or atrophy, and an “intention economy” that could commodify human intentionality. It notes current problems with hallucinations, reasoning limitations, and deviant chained errors, and mentions that GPT-3 has 96 attentional/neural processing blocks and that Gemini can handle up to 2 million token contexts.
Methods
- Conceptual and philosophical analysis; the paper reports no empirical data, sample, or statistical analysis.
- It develops its argument through engagement with extended cognition, enactivism, predictive coding, Material Engagement Theory, social cognition/participatory sense-making, and postphenomenology, using examples such as drawing, writing, word suggestion, a smart racket, pottery, jazz improvisation, and AI-mediated social/political scenarios.
- It proposes a definition of generative midtension and a two-dimensional framework (width and depth) inspired by Heersmink’s dimensions of integration in extended cognition.
Limitations
- Stated: current generative technology suffers from hallucinations, reasoning limitations, and deviant chained errors, making midtended cognition “a relevant but narrowed phenomenon”; the width dimension assumes high precision and accuracy, which the authors call an idealisation; generative AI may be plateauing; and risks of authenticity, power asymmetry, creative atrophy, and intention commodification remain.
- Stated/evident: the analysis is conceptual and speculative; the authors describe deep generative midtension scenarios as “still relatively fictional” though increasingly foreseeable, and they do not empirically validate or operationalise the width and depth dimensions.
- Evident: the paper offers no data on actual cognitive effects, and its claims about future risks and benefits are projections rather than tested outcomes.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
18 Valid · 7 Not valid · 10 Not checked · 82 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“But the intuition that more transparency equals more integration (in this sense of clear authorship of behaviour) starts to break down when we consider AI-powered technologies [Clowes, 2020, Wheeler, 2019].”
What the sources say
- Not matchedcites Clowes, 2020
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Supportedcites Wheeler, 2019 source
In §4, Wheeler argues that transparently integrated AI can guide behaviour through classifications that bypass the user's conscious awareness, raising questions about how to judge the resulting thoughts and actions. He therefore urges resistance to designing such technology for transparency. This supports the citing sentence's point that transparency no longer straightforwardly implies desirable integration or clear authorship of behaviour; the copy provides no page numbers. (LLM, conf=0.87)
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.
- Rests on an assumption not shown here.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, generative technologies do not qualify as autonomous agents [Barandiaran and Almendros, 2024, Gubelmann, 2024, Harvey, 2024].”
What the sources say
- Partly supportedcites Barandiaran and Almendros, 2024 sourceExperimental — accuracy not yet measured
Sections 3.4 and 4 argue that raw GPT and ChatGPT lack capacities required for genuine autonomous agency. The claim extends that conclusion from LLMs to generative technologies generally. Section 6.1 also qualifies the conclusion: extended LLMs might qualify as autonomous agents under a task-completion definition, though they currently fall short under the authors’ stronger definition. (LLM, conf=0.85)
- Partly supportedcites Gubelmann, 2024 sourceExperimental — accuracy not yet measured
The Introduction (§1) argues that transformer-based LLMs lack autonomy and are therefore not agents, supporting the claim for that class of generative technology. The paper does not establish the conclusion for generative technologies generally, as the citing sentence states. (LLM, conf=0.86)
- Supportedcites Harvey, 2024 source
In §5.1, Harvey states that current forms of deep learning “do not constitute agents” because they lack motivations of their own. The Conclusion (§9) applies this point to ChatGPT, describing chatbots as tools rather than systems with motivations in their own right. (LLM, conf=0.91)
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.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Generative AI can also collapse when recursively trained and tuned to cannibalised content [Shumailov et al., 2024].”
What the sources say
- Supportedcites Shumailov et al., 2024 source
The Main section says that learning from data produced by earlier models causes model collapse. The “Model collapse in language models” section examines sequential fine-tuning on model-generated data, and the paper states that fine-tuned models remain vulnerable. (LLM, conf=0.94)
plainstates a possible cause- 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.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Generative AI, when considered from a sociotechnical perspective [see also Hipólito et al., 2023], moves beyond the cases of individual creativity we have explored so far; and can become an integral component of social and political cognition.”
What the sources say
- Supportedcites Hipólito et al., 2023 source
The citation is attached to the sociotechnical perspective on AI. Section 2, “AI: cultural practice and practical culture” (page unavailable), explicitly describes AI as an inherently sociocultural practice whose design and meaning depend on social and cultural context. The sentence’s subsequent claim about social and political cognition is the citing authors’ extension, not the point specifically attributed to this source. (LLM, conf=0.90)
plainstates 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
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“First, by occupying positions traditionally held by human mediators in social coordination-e.g. as a family therapist mediating in a couple's conflict, a secretary organizing and scheduling meetings, or a lawyer mediating between a client and a public institution.”
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: unclear — It is unclear why the paper says it holds.
24 more claims
Create a free account to see all 117 claims and every source verdict.
- Probably not a substantial claim. 79
- It may state two separate claims as one. 28
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Hutchins, E. (1995) · doi:10.7551/mitpress/1881.001.0001 · Cited by 10,801 · cited 1×
“Notable examples include Haraway’s cyborgs and situated knowledge [Haraway, 1991], Stiegler’s Heideggerian account of technique [Stiegler, 1998] and Hutchins’ distributed cognition approach [Hutchins, 1995].”
Clark, A., & Chalmers, D. (1998) · Analysis · doi:10.1093/analys/58.1.7 · Cited by 6,333 · cited 1×
“However, the extended mind hypothesis [Clark and Chalmers, 1998] marks the first significant attempt, within philosophy of mind, to strongly argue on the triviality of retaining the mind within the boundaries of the skull.”
Whatever next? Predictive brains, situated agents, and the future of cognitive science
Clark, A. (2013) · Behavioral and Brain Sciences · doi:10.1017/s0140525x12000477 · Cited by 5,924
The Embodied Mind: Cognitive Science and Human Experience
Varela, F. J., Rosch, E., & Thompson, E. (1991) · doi:10.7551/mitpress/6730.001.0001 · Cited by 5,765 · cited 1×
“Contrary to the standard view, enactive and embodied contemporary approaches to cognition [Varela et al., 1991, Calvo and Gomila, 2008, Clark, 1998, Gallagher, 2023] stress the emergent nature of intentionality through sensory-motor interactions between agents and their environments.”
Simians, Cyborgs, and Women: The Reinvention of Nature.
Green, P. A., & Haraway, D. J. (1992) · Contemporary Sociology · doi:10.2307/2076334 · Cited by 5,455
Understanding and sharing intentions: The origins of cultural cognition
Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. (2005) · Behavioral and Brain Sciences · doi:10.1017/s0140525x05000129 · Cited by 4,882 · cited 2×
“8”
Intentionality: An Essay in the Philosophy of Mind
Searle, J. R. (1983) · doi:10.1017/cbo9781139173452 · Cited by 4,575
Simians, cyborgs, and women: the reinvention of nature
(1991) · Choice Reviews Online · doi:10.5860/choice.28-6234 · Cited by 3,677
Research works printed but not linked: 3
- Davidson, D. (1980). Essays on actions and events (underlining). Oxford University Press.
- Pérez-Verdugo, M. (2022). Situating transparency: An extended cognition approach. Teorema: Revista Internacional de Filosofía,41(3), 7–24. https://www.jstor.org/stable/27185808
- Stiegler, B. (1998). Technics and time, 1: The fault of epimetheus. Stanford University Press. · DOI · Paper page
Cited by
Bibri, S. E., & Huang, J. (2025) · Sustainable Cities and Society · doi:10.1016/j.scs.2025.106826 · Cited by 56
Transforming agency: On the mode of existence of large language models
Barandiaran, X. E., & Almendros, L. S. (2025) · Phenomenology and the Cognitive Sciences · doi:10.1007/s11097-025-10094-3 · Cited by 40
“For a detailed account of this concept of “midtended cognition” see Barandiaran and Pérez-Verdugo 2025.”
Self-citationCasacuberta, D., & Guersenzvaig, A. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1651354 · Cited by 4
AI-generated Literature, Distant Writing and the Reader: Reflections on Floridi and Calvino
Mollema, W. J. T. (2025) · Philosophy & Technology · doi:10.1007/s13347-025-01008-x · Cited by 6
“4 Barandiaran & Pérez-Verdugo (2025) however exemplify critics of this kind of writing.”
InfluentialThe coalescent architecture of agency: normative directionality as the key to human–AI integration
Torres–Martínez, S. (2026) · AI & Society · doi:10.1007/s00146-026-02979-x · Cited by 4
“In aviation, Bainbridge (1983) identified the fundamental "ironies of automation": highly automated systems paradoxically require the highest levels of human operator skill because operators must monitor complex systems, detect subtle failures, and intervene decisively under time pressure—yet…”
Metadata
- Authors
- Barandiaran, X. E., & Pérez-Verdugo, M.
- Affiliations
- University of the Basque Country (ES)
- Corresponding author
- Barandiaran, Xabier E.
- Pérez-Verdugo, Marta
- Venue
- Synthese
- Published
- 2025-03-27
- Volume
- 205
- Issue
- 4
- ISSN
- 0039-7857 (ISSN-L)
- 1573-0964
- OpenAlex
- W4408973818
- Cited by
- 19
- References
- 62
- Citations by year
- 2026: 11 · 2025: 8
- FWCI
- 10.76
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Synthese (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Universidad del País Vasco
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Cybernetics and Technology in Society
- Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
- Keywords
- generative AI, extended cognition, creative cognition, social cognition, large language models, multimodal transformer, cognitive agency, authenticity, power imbalance
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 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (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: Generative midtended cognition and Artificial Intelligence: thinging with thinging things. https://app.cochat.ai/papers/10.1007/s11229-025-04961-4
- The paper
- Barandiaran, X. E., & Pérez-Verdugo, M. (2025). Generative midtended cognition and Artificial Intelligence: thinging with thinging things. Synthese. https://doi.org/10.1007/s11229-025-04961-4
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.