Students Know AI Should Not Replace Thinking, but How Do They Regulate It? The TACO Framework for Human-AI Cognitive Partnership

Cecilia Ka Yuk Chan

arxiv: 2604.18737 · v1 · submitted 2026-04-20 · 💻 cs.CY · cs.HC

Students Know AI Should Not Replace Thinking, but How Do They Regulate It? The TACO Framework for Human-AI Cognitive Partnership

Cecilia Ka Yuk Chan This is my paper

Pith reviewed 2026-05-10 03:07 UTC · model grok-4.3

classification 💻 cs.CY cs.HC

keywords AI in educationcognitive regulationstudent-AI partnershipgenerative AI useawareness versus behaviorTACO frameworksecondary studentsHong Kong education

0 comments

The pith

Students understand AI should not replace thinking but do not reliably regulate its use in practice.

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

The paper examines the disconnect between Hong Kong secondary students' stated belief that AI must remain a support tool and their actual behaviors when using generative AI for schoolwork. Survey and interview data show that conceptual awareness often fails to produce consistent regulatory actions, ethical agreement does not produce strategic habits, and endorsement of limits does not ensure students stay in control of the thinking process. The study therefore shifts focus from repeating the ethical boundary to supplying a concrete process for managing AI interactions during learning.

Core claim

While students already know that AI should not replace thinking, this knowledge does not consistently produce structured regulation in real use. The study introduces the TACO framework (Think-Ask-Check-Own) as a learner-grounded process model that turns the abstract boundary into a repeatable sequence: students first think independently, then ask targeted questions of the AI, check outputs against their own reasoning, and claim ownership of the final product.

What carries the argument

The TACO framework (Think-Ask-Check-Own), a four-step process model that operationalizes the boundary between cognitive support and cognitive substitution by guiding learners through independent thinking, targeted AI queries, verification, and ownership.

If this is right

Educational efforts should move from teaching the ethical limit to teaching explicit regulation steps.
AI remains a dynamic partner only when learners apply operational checks during use.
Conceptual understanding alone is insufficient to prevent substitution of thinking.
Process models like TACO can turn general awareness into repeatable learner behavior.

Where Pith is reading between the lines

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

Classroom adoption of TACO steps could reduce passive copying from AI outputs and increase student accountability for final work.
The same awareness-regulation gap may appear in university or adult learning settings where AI tools are also common.
Longitudinal tracking of TACO use could reveal whether repeated practice strengthens independent thinking over time.

Load-bearing premise

The gap between students' awareness and their actual regulation can be closed by presenting the TACO framework without testing whether the framework itself changes behavior.

What would settle it

A before-and-after study that measures how often students outsource thinking tasks when using AI, then trains them on TACO steps and re-measures the same behaviors.

read the original abstract

As generative artificial intelligence becomes increasingly embedded in educational practice, a central concern is whether students use AI as cognitive support or as a substitute for thinking. Prior research shows that learners recognise this boundary conceptually and acknowledge that "AI should not replace thinking." However, whether such awareness translates into structured regulation during actual AI use remains unclear. Drawing on data from Hong Kong secondary students, this study examines how learners perceive their management of the boundary between assistance and outsourcing in practice. Findings show that awareness did not consistently translate into regulation; ethical belief did not necessarily lead to strategic execution; and conceptual endorsement did not guarantee operational behaviour. These findings suggest that the challenge is not teaching students that AI should not replace thinking, as they already know this, but providing them with structured mechanisms to regulate how AI is used within learning processes. In response, the study introduces the TACO framework (Think-Ask-Check-Own), a process-oriented model designed to operationalise the boundary between cognitive support and cognitive substitution. By shifting attention from ethical awareness to cognitive regulation, the study contributes a learner-grounded approach to sustaining AI as a dynamic cognitive partner in education.

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.

Desk Editor's Note private letter to a colleague

The paper shows students know AI should not replace thinking but often fail to regulate it in practice, then offers the untested TACO framework as the solution.

read the letter

The paper's main point is straightforward: secondary students in Hong Kong already understand that AI should support rather than replace their thinking, yet this understanding does not reliably turn into actual regulation during use. The authors respond by creating the TACO framework as a four-step process to guide that regulation. What the work does well is to collect and present qualitative evidence of this mismatch. It shows specific ways awareness fails to translate into practice, which is a useful distinction from just measuring ethical attitudes. The framework itself is simple and tied directly to the observed behaviors, which gives it some grounding. The limitations are clear and fairly significant for the claims being made. There are no details provided on the sample, data collection methods, or analysis process, so the strength of the evidence for the gap cannot be evaluated properly. Even more central, the TACO model is offered as the fix without any description of testing it with students or measuring whether it leads to better outcomes. This leaves the prescriptive part of the paper on shaky ground. Readers who work on AI ethics in education or who design student tools would get the most from this. It is not a finished intervention study but could spark ideas for one. The paper shows clear thinking about the problem and engages with the literature on student AI use. It deserves to go through peer review so that the methods and the framework can be examined and strengthened. I'd recommend sending it for review.

Referee Report

2 major / 2 minor

Summary. The paper claims that Hong Kong secondary students conceptually recognize AI should not replace thinking, yet this awareness does not reliably produce operational regulation of AI use in practice. Drawing on qualitative data, it identifies a gap between ethical beliefs, conceptual endorsement, and actual behavior, then introduces the TACO (Think-Ask-Check-Own) framework as a structured, process-oriented model to operationalize cognitive partnership and shift focus from awareness-raising to regulation mechanisms.

Significance. If the observed awareness-regulation gap holds and the TACO framework can be shown to improve regulation, the work would usefully redirect educational efforts toward practical scaffolds rather than repeated ethical messaging. The learner-grounded derivation of TACO from student observations is a positive feature, as is the explicit distinction between knowing the boundary and enacting it. However, the absence of any test of TACO's usability or effectiveness substantially limits the prescriptive contribution at present.

major comments (2)

Methods / Data section: The manuscript reports qualitative findings on the awareness-regulation gap but provides no information on sample size, participant selection, interview protocol, coding scheme, or inter-rater reliability. These omissions are load-bearing because the central empirical claim (awareness does not consistently translate into regulation) cannot be evaluated without them.
TACO Framework section (and Discussion): The paper presents TACO as the needed structured mechanism to address the identified gap, yet no pilot, intervention, pre/post measure, or even qualitative check is described that would show students using the Think-Ask-Check-Own steps actually regulate AI use more effectively. This renders the prescriptive claim that TACO supplies the missing operational bridge untested and therefore unsupported by the reported evidence.

minor comments (2)

Abstract and introduction: The phrasing 'students know AI should not replace thinking' is repeated without clarifying whether this is a direct quote from participants or a paraphrase; a brief illustrative excerpt would improve precision.
Figure or table (if present): Any visual representation of the TACO cycle would benefit from explicit mapping to the observed student behaviors described in the findings.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

Referee: Methods / Data section: The manuscript reports qualitative findings on the awareness-regulation gap but provides no information on sample size, participant selection, interview protocol, coding scheme, or inter-rater reliability. These omissions are load-bearing because the central empirical claim (awareness does not consistently translate into regulation) cannot be evaluated without them.

Authors: We agree that the absence of methodological details prevents proper evaluation of the findings. This was an oversight in the current draft. In the revised manuscript, we will expand the Methods section to report the sample size, participant selection criteria and recruitment process, the interview protocol used, the coding scheme for thematic analysis, and any steps taken to ensure reliability such as inter-rater checks. This will allow the central claim regarding the awareness-regulation gap to be properly assessed. revision: yes
Referee: TACO Framework section (and Discussion): The paper presents TACO as the needed structured mechanism to address the identified gap, yet no pilot, intervention, pre/post measure, or even qualitative check is described that would show students using the Think-Ask-Check-Own steps actually regulate AI use more effectively. This renders the prescriptive claim that TACO supplies the missing operational bridge untested and therefore unsupported by the reported evidence.

Authors: We appreciate this point and agree that the current presentation may overstate the immediate applicability of TACO. The framework was developed inductively from the student data to address the observed gap between awareness and regulation. However, we do not claim in the manuscript that TACO has been empirically tested for effectiveness. In the revision, we will clarify in the TACO section and Discussion that it is a proposed process-oriented model intended to guide future intervention studies, rather than a validated tool. We will also discuss potential next steps for testing its usability and impact. revision: partial

Circularity Check

0 steps flagged

No significant circularity: TACO is a conceptual proposal grounded in new observational data.

full rationale

The paper presents fresh survey/observational findings from Hong Kong secondary students documenting a gap between ethical awareness that AI should not replace thinking and actual regulatory behavior during AI use. It then introduces the TACO (Think-Ask-Check-Own) framework as a suggested process model to bridge that gap. No equations, parameter fits, self-definitional loops, or load-bearing self-citations appear; the framework is offered as an untested but logically responsive construct rather than a derived prediction or renamed prior result. The central claims rest on independent empirical observations rather than reducing to the paper's own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the assumption that a four-step process can bridge the observed awareness-regulation gap; no free parameters, formal axioms, or new invented entities beyond the named framework are introduced.

axioms (1)

domain assumption Students' self-reported and observed behavior accurately reflects their cognitive regulation strategies with AI.
Invoked when interpreting interview and observation data as evidence of regulation failure.

invented entities (1)

TACO framework no independent evidence
purpose: Operational model to regulate AI use in learning
Newly introduced process (Think-Ask-Check-Own) to translate awareness into behavior.

pith-pipeline@v0.9.0 · 5507 in / 1167 out tokens · 34660 ms · 2026-05-10T03:07:22.364764+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages · 1 internal anchor

[1]

Personality and social psychology review, 219–35 (2009) https://doi.org/ 10.1177/1088868309341564

https://doi.org/10.1177/1088868309341564 Alwaqdani, M. (2025). Investigating teachers’ perceptions of artificial intelligence tools in education: potential and difficulties. Educ Inf Technol 30, 2737–2755 https://doi.org/10.1007/s10639-024-12903-9 Chan, C. K. Y . (2025). Students’ perceptions of “AI-giarism”: Investigating changes in understandings of aca...

work page doi:10.1177/1088868309341564 2025
[2]

When does AI support thinking, and when does it replace it? Learners' conceptualisations of AI as a dynamic cognitive partner: A typology

https://doi.org/10.1007/s10639-024-13151-7 Chan, C. K. Y . (2026). Who is doing the thinking? AI as a dynamic cognitive partner: A learner- informed framework (Version 1). arXiv. https://doi.org/10.48550/arXiv.2602.15638 Chan, C. K. Y ., & Colloton, T. (2024). Generative AI in Higher Education: The ChatGPT Effect. Routledge. https://doi.org/10.4324/978100...

work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/s10639-024-13151-7 2026

[1] [1]

Personality and social psychology review, 219–35 (2009) https://doi.org/ 10.1177/1088868309341564

https://doi.org/10.1177/1088868309341564 Alwaqdani, M. (2025). Investigating teachers’ perceptions of artificial intelligence tools in education: potential and difficulties. Educ Inf Technol 30, 2737–2755 https://doi.org/10.1007/s10639-024-12903-9 Chan, C. K. Y . (2025). Students’ perceptions of “AI-giarism”: Investigating changes in understandings of aca...

work page doi:10.1177/1088868309341564 2025

[2] [2]

When does AI support thinking, and when does it replace it? Learners' conceptualisations of AI as a dynamic cognitive partner: A typology

https://doi.org/10.1007/s10639-024-13151-7 Chan, C. K. Y . (2026). Who is doing the thinking? AI as a dynamic cognitive partner: A learner- informed framework (Version 1). arXiv. https://doi.org/10.48550/arXiv.2602.15638 Chan, C. K. Y ., & Colloton, T. (2024). Generative AI in Higher Education: The ChatGPT Effect. Routledge. https://doi.org/10.4324/978100...

work page internal anchor Pith review Pith/arXiv arXiv doi:10.1007/s10639-024-13151-7 2026