pith. sign in

arxiv: 2606.05532 · v1 · pith:ZRGIDOTHnew · submitted 2026-06-04 · 💻 cs.AI · cs.HC

Individual Gain, Collective Loss: Metacognitive Adaptation in AI-Assisted Creativity

Anna Mikeda This is my paper

Pith reviewed 2026-06-28 02:11 UTC · model grok-4.3

classification 💻 cs.AI cs.HC
keywords metacognitive adaptationAI-assisted creativitycollective diversityoriginality evaluationpartner modelingreflective integrationcreative convergencecognitive redistribution
0
0 comments X

The pith

Routine AI use redistributes metacognitive effort in creativity, amplifying partner modeling and surface control while under-supporting originality evaluation and reflective integration.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper identifies a paradox where AI improves individual creative outputs yet reduces collective diversity in ideas. It proposes selective metacognitive adaptation as the mechanism: users adapt by shifting effort toward capacities that work well with AI, such as modeling the AI as a partner and managing surface-level features, while capacities for judging originality and integrating reflections receive less support. This shift produces personal satisfaction because outputs feel efficient and polished, but it generates group-level convergence because fewer people engage deeply with what makes an idea truly new. The authors organize six metacognitive capacities by their timing before, during, and after creation, then trace how individually sensible choices scale to social costs.

Core claim

We propose selective metacognitive adaptation: routine AI use redistributes rather than uniformly diminishes metacognitive effort. Some capacities are amplified (partner modeling, surface control), while others are systematically under-supported (originality evaluation, reflective integration). This redistribution explains both individual satisfaction and collective convergence. We present a taxonomy of six metacognitive capacities organized by temporal phase, characterize their tendencies under routine AI use, and show how individually rational adaptation produces emergent social costs.

What carries the argument

selective metacognitive adaptation, the redistribution of effort across a taxonomy of six capacities grouped by temporal phase (pre-task, during-task, post-task)

If this is right

  • The taxonomy generates testable predictions about which creative domains will show faster convergence under widespread AI adoption.
  • Design interventions can target the under-supported capacities, such as prompting users to evaluate originality explicitly before finalizing outputs.
  • Individual satisfaction metrics will remain high even as aggregate novelty declines, because the amplified capacities deliver immediate perceived benefits.
  • Collective convergence will appear first in tasks where surface control and partner modeling yield the largest efficiency gains.
  • Practitioners can use the temporal-phase organization to insert lightweight supports at the moments when reflective integration normally occurs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same redistribution pattern could appear in AI-assisted domains outside creativity, such as code review or strategic planning, where modeling the tool crowds out deeper evaluation.
  • If left unaddressed, the mechanism might compound over generations of users, producing training data that itself reflects reduced originality and accelerates further convergence.
  • Empirical mapping of capacity use before and after AI introduction in existing creative teams would provide a direct test of the redistribution claim without requiring new tools.

Load-bearing premise

The identified redistribution of metacognitive effort will systematically under-support originality evaluation and reflective integration enough to produce collective convergence rather than other group-level patterns.

What would settle it

A controlled study that tracks the six capacities in matched groups of AI users and non-users across repeated creative tasks and finds no reduction in group-level idea diversity despite the expected shifts in effort would falsify the central claim.

read the original abstract

Recent studies reveal a paradox: AI enhances individual creative outputs while reducing collective diversity. Current explanations -- cognitive offloading and over-reliance -- identify symptoms but not mechanisms. We propose selective metacognitive adaptation: routine AI use redistributes rather than uniformly diminishes metacognitive effort. Some capacities are amplified (partner modeling, surface control), while others are systematically under-supported (originality evaluation, reflective integration). This redistribution explains both individual satisfaction and collective convergence. We present a taxonomy of six metacognitive capacities organized by temporal phase, characterize their tendencies under routine AI use, and show how individually rational adaptation produces emergent social costs. The framework generates specific predictions for researchers and design principles for practitioners seeking to preserve both individual creative satisfaction and collective creative diversity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that recent studies show a paradox in AI-assisted creativity: enhanced individual outputs but reduced collective diversity. Current explanations (cognitive offloading, over-reliance) are insufficient; instead, the authors propose 'selective metacognitive adaptation' in which routine AI use redistributes (rather than uniformly reduces) metacognitive effort across a taxonomy of six capacities organized by temporal phase. Partner modeling and surface control are amplified while originality evaluation and reflective integration are systematically under-supported, producing individual satisfaction alongside emergent collective convergence. The framework generates predictions for researchers and design principles for practitioners.

Significance. If the proposed taxonomy and causal pathway hold, the work would supply a mechanistic bridge between individual-level metacognitive shifts and collective creative outcomes, moving beyond symptom descriptions to generate testable predictions and actionable design guidance. The explicit generation of predictions and principles is a constructive feature for a conceptual framework in this domain.

major comments (3)
  1. [Abstract] Abstract: the load-bearing claim that originality evaluation and reflective integration are 'systematically under-supported' (while partner modeling and surface control are amplified) lacks any cited mechanism, prior literature, or derivation explaining why these two capacities (rather than planning, monitoring, or others) are preferentially affected by routine AI use.
  2. [Abstract] Abstract / framework description: the assertion that individually rational redistribution 'necessarily produces emergent social costs' via collective convergence is central yet unsupported by any model, simulation, or formal argument showing why convergence (rather than stable diversity or other group-level outcomes) follows from the taxonomy.
  3. [Taxonomy section] Taxonomy section: the characterization of each capacity's tendency under AI use rests on the ad-hoc axiom that 'routine AI use produces a selective redistribution of effort' without empirical grounding, formal derivation, or falsifiable criteria for when a capacity is 'under-supported.'
minor comments (2)
  1. Provide explicit operational definitions or boundary conditions for the six metacognitive capacities to reduce potential overlap between categories.
  2. Distinguish claims about 'routine' versus occasional or prompted AI use throughout the taxonomy and predictions.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for these constructive comments, which identify opportunities to better justify the core claims of our conceptual framework. We respond point-by-point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the load-bearing claim that originality evaluation and reflective integration are 'systematically under-supported' (while partner modeling and surface control are amplified) lacks any cited mechanism, prior literature, or derivation explaining why these two capacities (rather than planning, monitoring, or others) are preferentially affected by routine AI use.

    Authors: The selectivity follows from the functional properties of generative AI interfaces, which externalize candidate generation and thereby reduce demand on internal originality evaluation and reflective synthesis while increasing requirements for modeling the AI partner and exerting surface-level control. This draws on prior metacognition research distinguishing generative from evaluative processes. We will revise the abstract and taxonomy section to include an explicit derivation and supporting citations. revision: yes

  2. Referee: [Abstract] Abstract / framework description: the assertion that individually rational redistribution 'necessarily produces emergent social costs' via collective convergence is central yet unsupported by any model, simulation, or formal argument showing why convergence (rather than stable diversity or other group-level outcomes) follows from the taxonomy.

    Authors: Uniform under-support of originality evaluation and reflective integration across individuals removes key sources of idiosyncratic variation, yielding convergence as an emergent outcome rather than stable diversity. The manuscript outlines this pathway conceptually. We will strengthen the argument with a clearer step-by-step causal chain and an illustrative diagram in revision. revision: partial

  3. Referee: [Taxonomy section] Taxonomy section: the characterization of each capacity's tendency under AI use rests on the ad-hoc axiom that 'routine AI use produces a selective redistribution of effort' without empirical grounding, formal derivation, or falsifiable criteria for when a capacity is 'under-supported.'

    Authors: As a conceptual framework, the taxonomy organizes observed patterns in AI-assisted creativity and derives falsifiability from the explicit predictions it generates. We will add a derivation subsection linking AI affordances to each capacity's tendency and clarify the criteria for under-support. revision: yes

Circularity Check

0 steps flagged

No circularity: conceptual proposal with independent taxonomy

full rationale

The paper advances a selective metacognitive adaptation framework as an explanatory construct for observed AI-creativity paradoxes. It defines a taxonomy of six capacities organized by temporal phase and asserts differential amplification/under-support under routine AI use, but does so without equations, fitted parameters, self-citations that bear the central load, or any derivation that reduces the redistribution claim to its own inputs by construction. The argument is presented as a novel organizing lens rather than a quantity computed from prior results or data subsets within the work itself; therefore the derivation chain is self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The framework rests on domain assumptions about the structure of metacognition and introduces a new explanatory concept without independent empirical grounding supplied in the abstract.

axioms (2)
  • domain assumption Metacognitive capacities relevant to creativity can be usefully partitioned into six types organized by temporal phase of the creative process.
    The paper builds its taxonomy and redistribution claims on this partitioning.
  • ad hoc to paper Routine AI use produces a selective redistribution of effort across these capacities rather than a uniform reduction.
    This redistribution is the core load-bearing premise introduced to explain both individual and collective outcomes.
invented entities (1)
  • Selective metacognitive adaptation no independent evidence
    purpose: To explain the paradox of individual creative gains alongside collective loss of diversity under routine AI use.
    The concept is introduced in the abstract as the central explanatory mechanism.

pith-pipeline@v0.9.1-grok · 5642 in / 1521 out tokens · 32681 ms · 2026-06-28T02:11:50.735782+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

16 extracted references

  1. [1]

    and Shah, S

    Anderson, N. and Shah, S. and Kreminski, M. , title =. Proceedings of the 2024 ACM Conference on Creativity and Cognition , year =

    2024

  2. [2]

    Doshi, A. R. and Hauser, O. P. , title =. Science Advances , year =

  3. [3]

    and Tang, L

    Fan, Y. and Tang, L. and Le, H. and Shen, K. and Tan, S. and Zhao, Y. and Shen, Y. and Li, X. and Gasevic, D. , title =. British Journal of Educational Technology , year =

  4. [4]

    Flavell, J. H. , title =. American Psychologist , year =

  5. [5]

    and Demazure, T

    Grange, C. and Demazure, T. and Ringeval, M. and Bourdeau, S. and Martineau, C. , title =. Information Systems Journal , year =

  6. [6]

    AI-Augmented Approaches to Creative Problem-Solving: A Metacognitive Perspective , journal =

    Ho. AI-Augmented Approaches to Creative Problem-Solving: A Metacognitive Perspective , journal =. 2025 , volume =

    2025

  7. [7]

    and So, H.-J

    Kim, J. and So, H.-J. and Park, K. , title =. British Journal of Educational Technology , year =

  8. [8]

    and Tarpin-Bernard, F

    Kosmyna, N. and Tarpin-Bernard, F. and Rivet, B. , title =. 2025 , howpublished =

    2025

  9. [9]

    and Cropley, D

    Medeiros, K. and Cropley, D. H. and Marrone, R. L. and Reiter-Palmon, R. , title =. The Journal of Creative Behavior , year =

  10. [10]

    and Gillett, A

    Menary, R. and Gillett, A. , title =. Topics in Cognitive Science , year =

  11. [11]

    and Green, J

    Moon, A. and Green, J. and Kushlev, K. , title =. Computers in Human Behavior: Artificial Humans , year =

  12. [12]

    and Weidmann, B

    Riedl, C. and Weidmann, B. , title =. 2025 , note =

    2025

  13. [13]

    Risko, E. F. and Gilbert, S. J. , title =. Trends in Cognitive Sciences , year =

  14. [14]

    and Kewenig, V

    Tankelevitch, L. and Kewenig, V. and Simkute, A. and Scott, A. E. and Sarkar, A. and Sellen, A. and Rintel, S. , title =. Proceedings of the CHI Conference on Human Factors in Computing Systems , year =

  15. [15]

    and Ouyang, F

    Xu, Y. and Ouyang, F. and Huan, Y. , title =. Learning and Individual Differences , year =

  16. [16]

    and Shao, Y

    Zhang, C. and Shao, Y. and Yuan, Y. and Shen, W. , title =. PsyCh Journal , year =