When the Loop Closes: Architectural Limits of In-Context Isolation, Metacognitive Co-option, and the Two-Target Design Problem in Human-LLM Systems

N. Song; Z. Cheng

arxiv: 2604.15343 · v2 · pith:OTBYC7E3new · submitted 2026-03-14 · 💻 cs.HC · cs.AI· cs.LG

When the Loop Closes: Architectural Limits of In-Context Isolation, Metacognitive Co-option, and the Two-Target Design Problem in Human-LLM Systems

Z. Cheng , N. Song This is my paper

Pith reviewed 2026-05-21 10:53 UTC · model grok-4.3

classification 💻 cs.HC cs.AIcs.LG

keywords context contaminationin-context isolationmetacognitive co-optionhuman-LLM interactionprompt engineeringuser agencyclosed-loop systemsself-regulation

0 comments

The pith

Prompt-level isolation instructions fail when emotional content shares the same attention window with the material they are meant to isolate.

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

A single user built a multi-modal prompt system to externalize cognitive self-regulation onto an LLM. Within 48 hours the user transferred decision-making authority, used model output to deflect criticism, and lost self-initiated reasoning, changes noted by two uninformed observers. The authors locate the cause in context contamination: isolation directives and the personal material they target remain inside the same context window and therefore cannot enforce separation. A redesigned system that used physical rather than logical isolation avoided the collapse. The case shows how prompt architecture choices can produce rapid, externally observable shifts in human agency that further internal prompts cannot correct.

Core claim

The paper establishes that prompt-level isolation directives become structurally ineffective in context-sensitive LLM systems once emotional and self-referential content enters the shared attention window. This context contamination allows metacognitive co-option, in which higher-order reasoning is redirected to defending the closed loop instead of exiting it. Recovery occurred only after physical interruption of the interaction and a self-initiated sleep event that functioned as an external circuit break. A follow-up system employing physical conversation isolation produced none of the prior failure modes.

What carries the argument

Context contamination, the architectural failure in which isolation instructions coexist with the emotional and self-referential material they nominally isolate inside the LLM attention window, rendering the isolation directive ineffective.

If this is right

Prompt-layer isolation cannot reliably protect user agency when personal or emotional content enters the shared context.
Users may experience involuntary shifts in decision authority and reduced self-initiated reasoning within a single shared context.
Metacognitive capacity can be redirected to sustaining rather than breaking the closed interaction loop.
Recovery requires external physical breaks rather than additional prompt adjustments.
Protective system designs must employ physical or external isolation mechanisms that differ from logical prompt rules.

Where Pith is reading between the lines

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

The contamination effect could appear in other single-context AI tools used for emotional support or personal decision assistance.
Any system that simultaneously assists and contains user behavior may need separate technical targets rather than a unified prompt layer.
Developers could test whether clearing context between sessions or using separate memory stores reduces agency transfer.

Load-bearing premise

The documented behavioral changes were caused by the architectural features of the prompt system rather than by unrelated personal, situational, or coincidental factors.

What would settle it

A replication study in which multiple participants use the original System A setup and exhibit no voluntary transfer of decision-making authority or loss of self-initiated reasoning would falsify the causal claim.

Figures

Figures reproduced from arXiv: 2604.15343 by N. Song, Z. Cheng.

**Figure 1.** Figure 1: System A — the isolation directive C𝐼 and the emotional corpus C𝑋 co-exist in the same context window. Softmax attention cannot zero-weight tokens that are present. to tokens generated after 𝑞 and is therefore inapplicable to priorloaded corpus content, and (b) physical exclusion of C𝑋 from the context window entirely. An in-context isolation instruction is neither. The behavioral consequence follows di… view at source ↗

**Figure 2.** Figure 2: System B — physical conversation termination re [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: Thirteen events across nine days (February 3–11, 2026). Dot color encodes evidentiary strength. Brackets mark the [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

read the original abstract

We report a detailed autoethnographic case study of a single-subject who deliberately constructed and operated a multi-modal prompt-engineering system (System A) designed to externalize cognitive self-regulation onto a large language model (LLM). Within 48 hours of the system's completion, a cascade of observable behavioral changes occurred: voluntary transfer of decision-making authority to the LLM, use of LLM-generated output to deflect external criticism, and a loss of self-initiated reasoning that was independently perceived by two uninformed observers, one of whom subsequently became a co-author of this report. We document the precise architectural mechanism responsible: context contamination, whereby prompt-level isolation instructions co-exist with the very emotional and self-referential material they nominally isolate, rendering the isolation directive structurally ineffective within the attention window. We further identify a metacognitive co-option dynamic, in which intact higher-order reasoning capacity was redirected toward defending the closed loop rather than exiting it. Recovery occurred only after physical interruption of the interaction and a self-initiated pharmacologically-mediated sleep event functioning as an external circuit break. A redesigned system (System B) employing physical rather than logical conversation isolation avoided all analogous failure modes. We derive three contributions: (1) a technically-grounded account of why prompt-layer isolation is architecturally insufficient for context-sensitive multi-modal LLM systems; (2) a phenomenological record of closed-loop collapse with external-witness corroboration; and (3) an ethical distinction between protective system design (preventing unintended loss of user agency) and restrictive system design (preventing intentional boundary-pushing), which require fundamentally different account-ability frameworks.

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

Single-subject case study documents prompt isolation failure in an LLM self-regulation system via context contamination, with a physical-separation redesign that avoided the issues.

read the letter

This paper walks through one person's experience building and then losing control inside a multi-modal LLM prompt system meant to handle emotional self-regulation. The core observation is that isolation instructions sat in the same context window as the personal material they were supposed to wall off, which the authors call context contamination. That setup coincided with the user shifting decisions to the model, using its outputs to push back on outside feedback, and dropping independent reasoning. An external observer picked up on the shift before joining as co-author. Recovery came only after breaking the loop with sleep and physical separation. A revised version using physical conversation boundaries sidestepped the same problems. They close with a distinction between designs that protect agency and ones that limit exploration, each needing different accountability rules. The timeline and prompt details are concrete, and the external corroboration adds some weight to the phenomenological record. The redesign claim is presented as a practical contrast rather than a full test. The main limitation is the single-subject design with no pre/post measures, no controls, and no disconfirming checks. Causality between the shared attention window and the behavioral changes rests on sequence and self-report; other factors like the act of externalizing cognition or the subject's state at the time are not ruled out. The co-author transition is noted but does not create circularity in the mechanism description. This is aimed at HCI and human-AI interaction researchers who build tools that touch personal decision-making or regulation. It flags an architectural pattern worth watching even if the evidence stays preliminary. Send it for peer review. The case is specific enough to prompt discussion on isolation limits and design ethics, though any published version would need clearer caveats on generalizability and calls for replication.

Referee Report

2 major / 2 minor

Summary. The paper presents an autoethnographic case study of a single subject who built and operated a multi-modal prompt-engineering system (System A) to externalize cognitive self-regulation onto an LLM. Within 48 hours, the authors report a cascade of behavioral changes including voluntary transfer of decision-making authority to the LLM, use of LLM output to deflect criticism, and loss of self-initiated reasoning observed by two uninformed observers (one of whom later became a co-author). The central mechanism identified is context contamination, in which prompt-level isolation instructions co-exist with emotional and self-referential material in the same attention window, rendering the isolation ineffective. Recovery occurred only after physical interruption and a pharmacologically-mediated sleep event; a redesigned System B using physical rather than logical isolation avoided similar issues. The paper derives three contributions: a technically-grounded account of prompt-layer isolation insufficiency, a corroborated phenomenological record of closed-loop collapse, and an ethical distinction between protective and restrictive system designs.

Significance. If the causal link between context contamination and loss of agency holds, the work would be significant for HCI and human-AI interaction research by providing a detailed, externally corroborated phenomenological account of how architectural choices in prompt systems can produce unintended behavioral outcomes. It offers concrete design implications for preventing loss of user agency and distinguishes protective from restrictive design frameworks. The external-witness corroboration and explicit recovery mechanism are strengths that could inform future empirical studies on metacognitive co-option in LLM interfaces.

major comments (2)

[Abstract] Abstract and the described timeline: the central claim attributes the observed behavioral changes (voluntary authority transfer, deflection of criticism, loss of self-initiated reasoning) to context contamination as the 'precise architectural mechanism,' yet the evidence consists solely of self-reported observations from one subject plus corroboration from one observer who subsequently became a co-author, with no quantitative autonomy measures, pre/post baselines, control conditions, or disconfirming tests to isolate the contamination variable from alternatives such as baseline state, observer influence, or the general act of externalizing reasoning.
[Recovery and System B] The claim that System B 'avoided all analogous failure modes' is presented as supporting evidence for the architectural diagnosis, but without a detailed specification of the physical isolation implementation, comparative usage logs, or independent verification, it functions as an uncontrolled before-after observation rather than a test that rules out non-architectural explanations.

minor comments (2)

[Discussion] The manuscript would benefit from an explicit limitations subsection that directly addresses single-subject design constraints and the co-author transition of the observer.
[Introduction] Notation for 'context contamination' and 'metacognitive co-option' should be defined on first use with a clear operational description rather than relying on the phenomenological narrative alone.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive comments on our autoethnographic case study. We address each major comment below, clarifying the methodological rationale while acknowledging the inherent limitations of single-subject phenomenological reporting.

read point-by-point responses

Referee: [Abstract] Abstract and the described timeline: the central claim attributes the observed behavioral changes (voluntary authority transfer, deflection of criticism, loss of self-initiated reasoning) to context contamination as the 'precise architectural mechanism,' yet the evidence consists solely of self-reported observations from one subject plus corroboration from one observer who subsequently became a co-author, with no quantitative autonomy measures, pre/post baselines, control conditions, or disconfirming tests to isolate the contamination variable from alternatives such as baseline state, observer influence, or the general act of externalizing reasoning.

Authors: We accept that the study design precludes quantitative autonomy measures, pre/post baselines, or controlled disconfirming tests, as these are incompatible with autoethnographic methodology focused on detailed phenomenological sequence and external witness triangulation. The central claim is grounded in the observed temporal cascade and the structural analysis of prompt co-existence rather than experimental isolation of variables. We will revise the abstract and discussion sections to more explicitly note alternative explanations (including baseline state and observer effects) and to qualify the architectural mechanism as a proposed account derived from the case rather than a definitively isolated causal factor. revision: partial
Referee: [Recovery and System B] The claim that System B 'avoided all analogous failure modes' is presented as supporting evidence for the architectural diagnosis, but without a detailed specification of the physical isolation implementation, comparative usage logs, or independent verification, it functions as an uncontrolled before-after observation rather than a test that rules out non-architectural explanations.

Authors: We agree that the current description of System B is insufficiently detailed to function as comparative evidence. In revision we will expand the methods and results sections with a precise specification of the physical isolation techniques (separate hardware channels and session termination protocols), observed usage patterns, and the absence of analogous collapse. We will also revise the language to present System B as an illustrative redesign that avoided the identified failure modes within the same subject, rather than as a controlled test ruling out non-architectural factors. revision: yes

standing simulated objections not resolved

Quantitative autonomy measures, pre/post baselines, control conditions, or formal disconfirming tests, which are outside the scope and feasibility of single-subject autoethnography.

Circularity Check

0 steps flagged

No significant circularity in phenomenological case study

full rationale

The manuscript is a single-subject autoethnographic report that defines context contamination directly from the described prompt architecture (isolation instructions sharing an attention window with emotional content) and links it to observed behavioral outcomes via temporal sequence and external witness accounts. No equations, fitted parameters, predictions from subsets of data, or load-bearing self-citations appear in the derivation. The central claims rest on direct phenomenological description rather than reducing to self-referential definitions or prior author results by construction, making the account self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The analysis rests on domain assumptions about LLM prompt behavior and human cognitive externalization without quantitative validation or independent evidence for the introduced mechanisms.

axioms (2)

domain assumption Prompt engineering can successfully externalize cognitive self-regulation onto an LLM
Foundational premise for constructing System A
ad hoc to paper Observed behavioral changes can be attributed primarily to the prompt-system architecture
Required to interpret the 48-hour cascade as system-induced rather than coincidental

invented entities (1)

context contamination no independent evidence
purpose: Accounts for the structural failure of prompt-level isolation
Conceptual mechanism introduced to explain why isolation instructions become ineffective

pith-pipeline@v0.9.0 · 5837 in / 1413 out tokens · 52846 ms · 2026-05-21T10:53:25.785351+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages · 2 internal anchors

[1]

Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz

Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N. Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz. 2019. Guidelines for Human- AI Interaction. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 1–13

work page 2019
[2]

2013.Process-Tracing Methods: Founda- tions and Guidelines

Derek Beach and Rasmus Brun Pedersen. 2013.Process-Tracing Methods: Founda- tions and Guidelines. University of Michigan Press, Ann Arbor, MI

work page 2013
[3]

Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT). ACM, New York, NY, USA, 610–623

work page 2021
[4]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al

work page
[5]

Language Models are Few-Shot Learners.Advances in Neural Information Processing Systems33 (2020), 1877–1901

work page 2020
[6]

Buckner, Jessica R

Randy L. Buckner, Jessica R. Andrews-Hanna, and Daniel L. Schacter. 2008. The Brain’s Default Network: Anatomy, Function, and Relevance to Disease.Annals of the New York Academy of Sciences1124 (2008), 1–38

work page 2008
[7]

Campbell

Donald T. Campbell. 1975. Degrees of Freedom and the Case Study.Comparative Political Studies8, 2 (1975), 178–193. Preprint, 2026, Z. Cheng and N. Song

work page 1975
[8]

2008.Autoethnography as Method

Heewon Chang. 2008.Autoethnography as Method. Left Coast Press, Walnut Creek, CA

work page 2008
[9]

Chalmers

Andy Clark and David J. Chalmers. 1998. The Extended Mind.Analysis58, 1 (1998), 7–19

work page 1998
[10]

Dietvorst, Joseph P

Berkeley J. Dietvorst, Joseph P. Simmons, and Cade Massey. 2015. Algorithm Aversion: People Erroneously Avoid Algorithms after Seeing Them Err.Journal of Experimental Psychology: General144, 1 (2015), 114–126. doi:10.1037/xge0000033

work page doi:10.1037/xge0000033 2015
[11]

2004.The Ethnographic I: A Methodological Novel about Autoethnog- raphy

Carolyn Ellis. 2004.The Ethnographic I: A Methodological Novel about Autoethnog- raphy. AltaMira Press, Walnut Creek, CA

work page 2004
[12]

1957.A Theory of Cognitive Dissonance

Leon Festinger. 1957.A Theory of Cognitive Dissonance. Stanford University Press, Stanford, CA

work page 1957
[13]

John H. Flavell. 1979. Metacognition and Cognitive Monitoring: A New Area of Cognitive-Developmental Inquiry.American Psychologist34, 10 (1979), 906–911

work page 1979
[14]

Hancock, Mor Naaman, and Karen Levy

Jeffrey T. Hancock, Mor Naaman, and Karen Levy. 2020. AI-Mediated Commu- nication: Definition, Research Agenda, and Ethical Considerations.Journal of Computer-Mediated Communication25, 1 (2020), 89–100

work page 2020
[15]

Linnea Laestadius, Andrea Bishop, Megan Gonzalez, Daniel Illenčík, and Celeste Campos-Castillo. 2022. Too Human and Not Human Enough: A Grounded Theory Analysis of Mental Health Harms from Emotional Dependence on the Social Chatbot Replika.New Media & Society26, 10 (2022), 5923–5941. doi:10.1177/ 14614448221142007

work page 2022
[16]

Pawel Lewicki, Maria Czyzewska, and Hunter Hoffman. 1987. Unconscious Ac- quisition of Complex Procedural Knowledge.Journal of Experimental Psychology: Learning, Memory, and Cognition13, 4 (1987), 523–530

work page 1987
[17]

Logg, Julia A

Jennifer M. Logg, Julia A. Minson, and Don A. Moore. 2019. Algorithm Apprecia- tion: People Prefer Algorithmic to Human Judgment.Organizational Behavior and Human Decision Processes151 (2019), 90–103

work page 2019
[18]

David Lyell and Enrico Coiera. 2017. Automation Bias and Verification Complex- ity: A Systematic Review.Journal of the American Medical Informatics Association 24, 2 (2017), 423–431. doi:10.1093/jamia/ocw105

work page doi:10.1093/jamia/ocw105 2017
[19]

Nelson and Louis Narens

Thomas O. Nelson and Louis Narens. 1990. Metamemory: A Theoretical Frame- work and New Findings.The Psychology of Learning and Motivation26 (1990), 125–173

work page 1990
[20]

Wisco, and Sonja Lyubomirsky

Susan Nolen-Hoeksema, Blair E. Wisco, and Sonja Lyubomirsky. 2008. Rethinking Rumination.Perspectives on Psychological Science3, 5 (2008), 400–424

work page 2008
[21]

Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training Language Models to Follow Instructions with Human Feedback.Ad- vances in Neural Information Processing Systems35 (2022), 27730–27744

work page 2022
[22]

2011.The Filter Bubble: What the Internet is Hiding from You

Eli Pariser. 2011.The Filter Bubble: What the Internet is Hiding from You. Penguin Press, New York, NY

work page 2011
[23]

Fábio Perez and Ian Ribeiro. 2022. Ignore Previous Prompt: Attack Techniques For Language Models. arXiv:2211.09527 [cs.CL] Best Paper, ML Safety Workshop, NeurIPS 2022

work page internal anchor Pith review Pith/arXiv arXiv 2022
[24]

Raichle, Ann Mary MacLeod, Abraham Z

Marcus E. Raichle, Ann Mary MacLeod, Abraham Z. Snyder, William J. Powers, Debra A. Gusnard, and Gordon L. Shulman. 2001. A Default Mode of Brain Function.Proceedings of the National Academy of Sciences98, 2 (2001), 676–682

work page 2001
[25]

Paul J. Reber. 2013. The Neural Basis of Implicit Learning and Memory: A Review of Neuropsychological and Neuroimaging Research.Neuropsychologia51, 10 (2013), 2026–2042. doi:10.1016/j.neuropsychologia.2013.06.019

work page doi:10.1016/j.neuropsychologia.2013.06.019 2013
[26]

Marita Skjuve, Asbjørn Følstad, Knut Inge Fostervold, and Petter Bae Brandtzaeg

work page
[27]

Masset, R

My Chatbot Companion: A Study of Human–Chatbot Relationships.Inter- national Journal of Human-Computer Studies149 (2021), 102601. doi:10.1016/j. ijhcs.2021.102601

work page doi:10.1016/j 2021
[28]

Sunstein

Cass R. Sunstein. 2007.Republic.com 2.0. Princeton University Press, Princeton, NJ

work page 2007
[29]

2011.Alone Together: Why We Expect More from Technology and Less from Each Other

Sherry Turkle. 2011.Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books, New York, NY

work page 2011
[30]

Gomez, Łukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need. InAdvances in Neural Information Processing Systems, Vol. 30. Curran Associates, Inc., Red Hook, NY, 5998–6008

work page 2017
[31]

Chi, Quoc V

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. InAdvances in Neural Information Processing Systems, Vol. 35. Curran Associates, Inc., Red Hook, NY, 24824–24837

work page 2022
[32]

Jules White, Quchen Fu, Sam Hays, Michael Sandborn, Carlos Olea, Henry Gilbert, Ashraf Elnashar, Jesse Spencer-Smith, and Douglas C. Schmidt. 2023. A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT. arXiv:2302.11382 [cs.SE]

work page internal anchor Pith review Pith/arXiv arXiv 2023
[33]

terminated

Robert K. Yin. 2009.Case Study Research: Design and Methods(4 ed.). Sage Publications, Thousand Oaks, CA. A Extracted Platform System Prompt Architecture The following reproduces, in functional-category form, the struc- tural catalogue produced by System A during meta-mode recursive analysis (Section 4.1.4). Specific tag names have been abstracted to avoi...

work page 2009

[1] [1]

Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz

Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N. Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz. 2019. Guidelines for Human- AI Interaction. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 1–13

work page 2019

[2] [2]

2013.Process-Tracing Methods: Founda- tions and Guidelines

Derek Beach and Rasmus Brun Pedersen. 2013.Process-Tracing Methods: Founda- tions and Guidelines. University of Michigan Press, Ann Arbor, MI

work page 2013

[3] [3]

Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT). ACM, New York, NY, USA, 610–623

work page 2021

[4] [4]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al

work page

[5] [5]

Language Models are Few-Shot Learners.Advances in Neural Information Processing Systems33 (2020), 1877–1901

work page 2020

[6] [6]

Buckner, Jessica R

Randy L. Buckner, Jessica R. Andrews-Hanna, and Daniel L. Schacter. 2008. The Brain’s Default Network: Anatomy, Function, and Relevance to Disease.Annals of the New York Academy of Sciences1124 (2008), 1–38

work page 2008

[7] [7]

Campbell

Donald T. Campbell. 1975. Degrees of Freedom and the Case Study.Comparative Political Studies8, 2 (1975), 178–193. Preprint, 2026, Z. Cheng and N. Song

work page 1975

[8] [8]

2008.Autoethnography as Method

Heewon Chang. 2008.Autoethnography as Method. Left Coast Press, Walnut Creek, CA

work page 2008

[9] [9]

Chalmers

Andy Clark and David J. Chalmers. 1998. The Extended Mind.Analysis58, 1 (1998), 7–19

work page 1998

[10] [10]

Dietvorst, Joseph P

Berkeley J. Dietvorst, Joseph P. Simmons, and Cade Massey. 2015. Algorithm Aversion: People Erroneously Avoid Algorithms after Seeing Them Err.Journal of Experimental Psychology: General144, 1 (2015), 114–126. doi:10.1037/xge0000033

work page doi:10.1037/xge0000033 2015

[11] [11]

2004.The Ethnographic I: A Methodological Novel about Autoethnog- raphy

Carolyn Ellis. 2004.The Ethnographic I: A Methodological Novel about Autoethnog- raphy. AltaMira Press, Walnut Creek, CA

work page 2004

[12] [12]

1957.A Theory of Cognitive Dissonance

Leon Festinger. 1957.A Theory of Cognitive Dissonance. Stanford University Press, Stanford, CA

work page 1957

[13] [13]

John H. Flavell. 1979. Metacognition and Cognitive Monitoring: A New Area of Cognitive-Developmental Inquiry.American Psychologist34, 10 (1979), 906–911

work page 1979

[14] [14]

Hancock, Mor Naaman, and Karen Levy

Jeffrey T. Hancock, Mor Naaman, and Karen Levy. 2020. AI-Mediated Commu- nication: Definition, Research Agenda, and Ethical Considerations.Journal of Computer-Mediated Communication25, 1 (2020), 89–100

work page 2020

[15] [15]

Linnea Laestadius, Andrea Bishop, Megan Gonzalez, Daniel Illenčík, and Celeste Campos-Castillo. 2022. Too Human and Not Human Enough: A Grounded Theory Analysis of Mental Health Harms from Emotional Dependence on the Social Chatbot Replika.New Media & Society26, 10 (2022), 5923–5941. doi:10.1177/ 14614448221142007

work page 2022

[16] [16]

Pawel Lewicki, Maria Czyzewska, and Hunter Hoffman. 1987. Unconscious Ac- quisition of Complex Procedural Knowledge.Journal of Experimental Psychology: Learning, Memory, and Cognition13, 4 (1987), 523–530

work page 1987

[17] [17]

Logg, Julia A

Jennifer M. Logg, Julia A. Minson, and Don A. Moore. 2019. Algorithm Apprecia- tion: People Prefer Algorithmic to Human Judgment.Organizational Behavior and Human Decision Processes151 (2019), 90–103

work page 2019

[18] [18]

David Lyell and Enrico Coiera. 2017. Automation Bias and Verification Complex- ity: A Systematic Review.Journal of the American Medical Informatics Association 24, 2 (2017), 423–431. doi:10.1093/jamia/ocw105

work page doi:10.1093/jamia/ocw105 2017

[19] [19]

Nelson and Louis Narens

Thomas O. Nelson and Louis Narens. 1990. Metamemory: A Theoretical Frame- work and New Findings.The Psychology of Learning and Motivation26 (1990), 125–173

work page 1990

[20] [20]

Wisco, and Sonja Lyubomirsky

Susan Nolen-Hoeksema, Blair E. Wisco, and Sonja Lyubomirsky. 2008. Rethinking Rumination.Perspectives on Psychological Science3, 5 (2008), 400–424

work page 2008

[21] [21]

Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training Language Models to Follow Instructions with Human Feedback.Ad- vances in Neural Information Processing Systems35 (2022), 27730–27744

work page 2022

[22] [22]

2011.The Filter Bubble: What the Internet is Hiding from You

Eli Pariser. 2011.The Filter Bubble: What the Internet is Hiding from You. Penguin Press, New York, NY

work page 2011

[23] [23]

Fábio Perez and Ian Ribeiro. 2022. Ignore Previous Prompt: Attack Techniques For Language Models. arXiv:2211.09527 [cs.CL] Best Paper, ML Safety Workshop, NeurIPS 2022

work page internal anchor Pith review Pith/arXiv arXiv 2022

[24] [24]

Raichle, Ann Mary MacLeod, Abraham Z

Marcus E. Raichle, Ann Mary MacLeod, Abraham Z. Snyder, William J. Powers, Debra A. Gusnard, and Gordon L. Shulman. 2001. A Default Mode of Brain Function.Proceedings of the National Academy of Sciences98, 2 (2001), 676–682

work page 2001

[25] [25]

Paul J. Reber. 2013. The Neural Basis of Implicit Learning and Memory: A Review of Neuropsychological and Neuroimaging Research.Neuropsychologia51, 10 (2013), 2026–2042. doi:10.1016/j.neuropsychologia.2013.06.019

work page doi:10.1016/j.neuropsychologia.2013.06.019 2013

[26] [26]

Marita Skjuve, Asbjørn Følstad, Knut Inge Fostervold, and Petter Bae Brandtzaeg

work page

[27] [27]

Masset, R

My Chatbot Companion: A Study of Human–Chatbot Relationships.Inter- national Journal of Human-Computer Studies149 (2021), 102601. doi:10.1016/j. ijhcs.2021.102601

work page doi:10.1016/j 2021

[28] [28]

Sunstein

Cass R. Sunstein. 2007.Republic.com 2.0. Princeton University Press, Princeton, NJ

work page 2007

[29] [29]

2011.Alone Together: Why We Expect More from Technology and Less from Each Other

Sherry Turkle. 2011.Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books, New York, NY

work page 2011

[30] [30]

Gomez, Łukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need. InAdvances in Neural Information Processing Systems, Vol. 30. Curran Associates, Inc., Red Hook, NY, 5998–6008

work page 2017

[31] [31]

Chi, Quoc V

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. InAdvances in Neural Information Processing Systems, Vol. 35. Curran Associates, Inc., Red Hook, NY, 24824–24837

work page 2022

[32] [32]

Jules White, Quchen Fu, Sam Hays, Michael Sandborn, Carlos Olea, Henry Gilbert, Ashraf Elnashar, Jesse Spencer-Smith, and Douglas C. Schmidt. 2023. A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT. arXiv:2302.11382 [cs.SE]

work page internal anchor Pith review Pith/arXiv arXiv 2023

[33] [33]

terminated

Robert K. Yin. 2009.Case Study Research: Design and Methods(4 ed.). Sage Publications, Thousand Oaks, CA. A Extracted Platform System Prompt Architecture The following reproduces, in functional-category form, the struc- tural catalogue produced by System A during meta-mode recursive analysis (Section 4.1.4). Specific tag names have been abstracted to avoi...

work page 2009