arxiv: 2603.24768 · v2 · submitted 2026-03-25 · 💻 cs.AI

Recognition: no theorem link

Supervising Ralph Wiggum: Exploring a Metacognitive Co-Regulation Agentic AI Loop for Engineering Design

Zeda Xu , Nikolas Martelaro , Christopher McComb

Authors on Pith no claims yet

Pith reviewed 2026-05-15 00:14 UTC · model grok-4.3

classification 💻 cs.AI

keywords agentic AILLM design agentsmetacognitive regulationengineering designdesign fixationself-regulation loopco-regulationbattery pack design

0 comments

The pith

Metacognitive regulation loops let LLM design agents create better-performing battery packs at similar computational cost.

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

Large language model agents used for engineering design often fixate on familiar solutions and miss better ones, much as human designers can. This paper proposes two additions to a basic agent loop: a Self-Regulation Loop where the design agent explicitly monitors its own thinking, and a Co-Regulation Design Agentic Loop where a separate metacognitive agent assists with that monitoring. When tested on a battery pack design task, both new loops produced designs with higher performance than the unregulated baseline, the co-regulation version performed best, and neither raised computational cost much. The co-regulated system also moved through the space of possible designs more effectively. The work demonstrates that targeted self-correction layers can help automated design agents avoid common fixation problems.

Core claim

In the battery pack design problem examined here, the novel SRL and CRDAL systems generate designs with better performance, without significantly increasing the computational cost, compared to a plain Ralph Wiggum Loop. Further, the novel CRDAL generates designs with significantly better performance than SRL. Also, the CRDAL system navigated through the latent design space more effectively than both SRL and RWL.

What carries the argument

The Metacognitive Co-Regulation Agent, a separate LLM agent that assists the main Design Agent by monitoring and correcting its cognitive processes to reduce fixation on existing design paradigms.

If this is right

The co-regulated architecture outperforms self-regulation alone on both final design quality and exploration of the design space.
Regulation layers can be added without materially raising compute demands for this class of design task.
Agentic systems equipped with explicit metacognition generate higher-performing solutions than unregulated loops on the tested engineering problem.
These architectures supply a concrete method for mitigating design fixation in LLM-based engineering design agents.

Where Pith is reading between the lines

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

If the regulation mechanisms prove general, they could be applied to LLM agents in other fixation-prone domains such as scientific hypothesis generation or creative problem solving.
Making the co-regulation agent adapt its interventions based on real-time signals from the design agent might further improve results.
Testing the same loops on a wider range of engineering tasks would reveal whether the benefits are specific to battery design or hold more broadly.

Load-bearing premise

The performance gains seen in the battery pack task arise specifically from the metacognitive regulation features rather than from other unmeasured differences in the system or the task.

What would settle it

A controlled run of the same battery pack task with the metacognitive monitoring and assistance components removed while keeping every other element identical, showing no drop in design performance, would falsify the claim.

Figures

Figures reproduced from arXiv: 2603.24768 by Christopher McComb, Nikolas Martelaro, Zeda Xu.

**Figure 2.** Figure 2: Example of a simple 6(W) × 4(D) × 2(H) battery pack composed of 18650 Lithium-ion battery cells using hexagonal close-packing. 4 2.1.1 Design Objective, Constraints, and Assumptions The agentic design systems are instructed to generate a battery pack design using only 18650 cells to satisfy all constraints, while maximizing capacity: Design a 400V battery pack with a minimum capacity of 25Ah, capable of co… view at source ↗

**Figure 3.** Figure 3: System flowchart of the Ralph Wiggum Loop (RWL). [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: System flowchart of the Self-Regulation Loop (SRL). [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

**Figure 5.** Figure 5: System flowchart of the Co-Regulation Design Agentic Loop (CRDAL). [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

**Figure 6.** Figure 6: The dots in the figure show the capacities of the final battery pack designs for each run. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

**Figure 6.** Figure 6: Design capacity of the battery pack created by each agentic design system. p-value [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

**Figure 7.** Figure 7: Number of design steps taken before final design by each agentic design system. p [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

**Figure 8.** Figure 8: Design step trajectory and final design in the latent design space, explored by each agentic [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

**Figure 9.** Figure 9: Cell series and parallel connections explored by each agentic design system. [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗

read the original abstract

The engineering design research community has studied agentic AI systems that use Large Language Model (LLM) agents to automate the engineering design process. However, these systems are prone to some of the same pathologies that plague humans. Just as human designers, LLM design agents can fixate on existing paradigms and fail to explore alternatives when solving design challenges, potentially leading to suboptimal solutions. In this work, we propose (1) a novel Self-Regulation Loop (SRL), in which the Design Agent self-regulates and explicitly monitors its own metacognition, and (2) a novel Co-Regulation Design Agentic Loop (CRDAL), in which a Metacognitive Co-Regulation Agent assists the Design Agent in metacognition to mitigate design fixation, thereby improving system performance for engineering design tasks. In the battery pack design problem examined here, we found that the novel SRL and CRDAL systems generate designs with better performance, without significantly increasing the computational cost, compared to a plain Ralph Wiggum Loop (RWL) Further, the novel CRDAL generates designs with significantly better performance than SRL. Also, we found that the CRDAL system navigated through the latent design space more effectively than both SRL and RWL. The proposed system architectures and findings of this work provide practical implications for future development of agentic AI systems for engineering design.

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 concrete gains from adding self- and co-regulation loops to an LLM design agent on one battery-pack task, but the results do not isolate the metacognitive mechanism from extra agents or prompts.

read the letter

The key point is that this work adds self-regulation and co-regulation loops to an existing agentic design system and gets better designs on a battery pack problem, with the co-regulation version coming out ahead. The paper introduces SRL, where the design agent monitors its own thinking to avoid fixation, and CRDAL, where a separate metacognitive agent helps out. They test these against a plain loop on one engineering design challenge and report stronger performance and better navigation of the design space, all without much added computation. This is a practical step. The architectures are described in enough detail to replicate, and the focus on mitigating a known issue in LLM agents—getting stuck on initial ideas—is on target for engineering applications. The finding that co-regulation beats self-regulation alone is interesting and worth noting. The soft spot is the experimental setup. Everything rests on one task, and there's no test that isolates the metacognitive monitoring from other changes like added agents or prompt variations. Without that, it's difficult to say the gains come from the proposed regulation rather than just more complex prompting or interaction. The abstract also skips statistical details, so the significance claims are hard to evaluate fully. This paper is aimed at researchers building agentic AI for design tasks. Someone looking for concrete ways to improve reliability in automated engineering workflows could get ideas from it. It deserves a serious referee because the core idea is testable and the results are promising, even if the current evidence needs strengthening with more controls and tasks.

Referee Report

3 major / 2 minor

Summary. The paper proposes two new agentic AI loop architectures—Self-Regulation Loop (SRL) and Co-Regulation Design Agentic Loop (CRDAL)—that add explicit metacognitive monitoring to an LLM-based design agent (the Ralph Wiggum Loop baseline) in order to reduce design fixation. On a single battery-pack design task the authors report that both SRL and CRDAL produce higher-performing designs than the plain RWL baseline at comparable computational cost, that CRDAL outperforms SRL, and that CRDAL explores the latent design space more effectively.

Significance. If the performance gains can be shown to arise specifically from the metacognitive regulation mechanisms rather than from ancillary increases in system complexity, the work would supply a concrete, implementable pattern for improving exploration in LLM-driven engineering design agents—an issue that is widely recognized but rarely isolated in current agentic-design literature.

major comments (3)

[Abstract, §4] Abstract and §4 (Results): the central claim that SRL and CRDAL generate designs with 'significantly better performance' is not accompanied by any reported statistical tests, effect sizes, or confidence intervals; without these the quantitative superiority over RWL cannot be evaluated.
[§3, §4] §3 (Methods) and §4: CRDAL introduces a second agent whose interactions are not controlled for total token budget, iteration count, or prompt diversity relative to the single-agent RWL baseline; the performance delta cannot be confidently attributed to the metacognitive co-regulation component rather than to the added agent and prompt volume.
[§4] §4: no ablation is described that disables the explicit self-monitoring or co-regulation prompts while preserving the rest of the architecture; therefore the contribution of the proposed metacognitive loops versus other unmeasured factors remains unisolated.

minor comments (2)

[Abstract] The abstract states that CRDAL 'navigated through the latent design space more effectively' but provides no quantitative metric or visualization supporting this claim.
[§2, §3] Notation for the three loops (RWL, SRL, CRDAL) is introduced without a consolidated table or diagram that would allow readers to compare their control flows at a glance.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important gaps in our quantitative reporting and experimental controls. We address each major point below and commit to revisions that will strengthen the manuscript's claims.

read point-by-point responses

Referee: [Abstract, §4] Abstract and §4 (Results): the central claim that SRL and CRDAL generate designs with 'significantly better performance' is not accompanied by any reported statistical tests, effect sizes, or confidence intervals; without these the quantitative superiority over RWL cannot be evaluated.

Authors: We agree that statistical support is required to substantiate the performance claims. In the revised manuscript we will add appropriate non-parametric tests (e.g., Mann-Whitney U with Bonferroni correction) for the key metrics, report effect sizes (Cohen’s d or rank-biserial), and include 95% confidence intervals. These additions will be placed in §4 and referenced in the abstract. revision: yes
Referee: [§3, §4] §3 (Methods) and §4: CRDAL introduces a second agent whose interactions are not controlled for total token budget, iteration count, or prompt diversity relative to the single-agent RWL baseline; the performance delta cannot be confidently attributed to the metacognitive co-regulation component rather than to the added agent and prompt volume.

Authors: We acknowledge the need for tighter controls. Although the abstract states that computational cost remained comparable, we did not tabulate per-condition token counts or iteration budgets. The revision will include a new table in §4 reporting total tokens, API calls, and iteration counts for RWL, SRL, and CRDAL, plus a brief discussion of prompt-length matching. If the data show residual imbalance, we will note it as a limitation and consider matched-budget follow-up runs. revision: yes
Referee: [§4] §4: no ablation is described that disables the explicit self-monitoring or co-regulation prompts while preserving the rest of the architecture; therefore the contribution of the proposed metacognitive loops versus other unmeasured factors remains unisolated.

Authors: This is a fair criticism. Our current comparisons pit complete SRL/CRDAL systems against the RWL baseline but do not isolate the metacognitive prompts. We will add an ablation condition in the revised §4 in which the self-monitoring and co-regulation instructions are removed while retaining the same agent scaffolding and iteration structure. Results from these runs will be reported alongside the main findings. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical comparisons rest on measured outcomes

full rationale

The paper advances SRL and CRDAL architectures and evaluates them via direct experiments on a battery-pack design task, reporting performance, cost, and latent-space navigation metrics against the RWL baseline. No derivation chain, equations, or predictions are presented that reduce by construction to fitted parameters, self-definitions, or prior self-citations. The central claims are falsifiable experimental deltas; the architecture descriptions and results do not collapse into tautology. This is the expected non-finding for an empirical systems paper whose evidence consists of controlled task runs rather than analytic reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 3 invented entities

The central claim rests on the assumption that LLM agents exhibit human-like design fixation that can be mitigated by explicit metacognitive monitoring, with no free parameters, standard axioms, or invented entities beyond the named loops themselves.

invented entities (3)

Self-Regulation Loop (SRL) no independent evidence
purpose: Enables the Design Agent to explicitly monitor its own metacognition
Introduced as a novel component in the proposed architecture
Co-Regulation Design Agentic Loop (CRDAL) no independent evidence
purpose: Uses a separate Metacognitive Co-Regulation Agent to assist the Design Agent
Introduced as a novel component in the proposed architecture
Ralph Wiggum Loop (RWL) no independent evidence
purpose: Serves as the plain baseline agentic loop without regulation
Named baseline for comparison

pith-pipeline@v0.9.0 · 5548 in / 1219 out tokens · 31487 ms · 2026-05-15T00:14:34.568581+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

74 extracted references · 74 canonical work pages · 2 internal anchors

[1]

URL https://ai.google.dev/gemini-api/ docs/gemini-3

Gemini 3 Developer Guide | Gemini API. URL https://ai.google.dev/gemini-api/ docs/gemini-3

work page
[2]

ISSN 1050-0472

Special Issue: Artificial Intelligence and Engineering Design.Journal of Mechanical Design, 144(020301), January 2022. ISSN 1050-0472. doi: 10.1115/1.4053111. URL https://doi. org/10.1115/1.4053111

work page doi:10.1115/1.4053111 2022
[3]

Gemini 3.1 Pro: A smarter model for your most complex tasks, February

work page
[4]

URL https://blog.google/innovation-and-ai/models-and-research/ gemini-models/gemini-3-1-pro/

work page
[5]

Assessment and the co-regulation of learning in the classroom.Assessment in Education: Principles, Policy & Practice, 27(4):332–349, July 2020

Linda Allal. Assessment and the co-regulation of learning in the classroom.Assessment in Education: Principles, Policy & Practice, 27(4):332–349, July 2020. ISSN 0969-594X. doi: 10.1080/0969594X.2019.1609411. URL https://doi.org/10.1080/0969594X.2019. 1609411. _eprint: https://doi.org/10.1080/0969594X.2019.1609411

work page doi:10.1080/0969594x.2019.1609411 2020
[6]

Study on the battery thermal management sys- tem for cylindrical lithium-ion battery with nano-doped phase change material and liquid cooling

Dhinesh Balasubramanian, Inbanaathan Papla Venugopal, Mohankumar Subramanian, Vi- jayanandh Raja, Utku Kale, and Jonas Matijošius. Study on the battery thermal management sys- tem for cylindrical lithium-ion battery with nano-doped phase change material and liquid cooling. Scientific Reports, 15(1):24053, July 2025. ISSN 2045-2322. doi: 10.1038/s41598-025...

work page doi:10.1038/s41598-025-08884-5 2025
[7]

Self-regulated learning: A new concept embraced by researchers, policy makers, educators, teachers, and students.Learning and Instruction, 7(2):161– 186, June 1997

Monique Boekaerts. Self-regulated learning: A new concept embraced by researchers, policy makers, educators, teachers, and students.Learning and Instruction, 7(2):161– 186, June 1997. ISSN 0959-4752. doi: 10.1016/S0959-4752(96)00015-1. URL https: //www.sciencedirect.com/science/article/pii/S0959475296000151

work page doi:10.1016/s0959-4752(96)00015-1 1997
[8]

Goldstein, John Clay, H

Alex Brown, Molly H. Goldstein, John Clay, H. Onan Demirel, Xingang Li, and Zhenghui Sha. A Study on Generative Design Reasoning and Students’ Divergent and Convergent Thinking.Journal of Mechanical Design, 146(031405), February 2024. ISSN 1050-0472. doi: 10.1115/1.4064564. URLhttps://doi.org/10.1115/1.4064564

work page doi:10.1115/1.4064564 2024
[9]

Ann L. Brown. Knowing When, Where, and How to Remember: A Problem of Metacognition. In Robert Glaser, editor,Advances in Instructional Psychology, volume 1, pages 77–165. Lawrence Erlbaum Associates, Hillsdale, NJ, 1978

work page 1978
[10]

Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwi...

work page 2020
[11]

Design as a mutual learning process between users and designers.Inter- acting with Computers, 15(5):709–730, October 2003

Pascal Béguin. Design as a mutual learning process between users and designers.Inter- acting with Computers, 15(5):709–730, October 2003. ISSN 0953-5438. doi: 10.1016/ S0953-5438(03)00060-2. URL https://www.sciencedirect.com/science/article/ pii/S0953543803000602

work page 2003
[12]

Campbell, Jonathan Cagan, and Kenneth Kotovsky

Matthew I. Campbell, Jonathan Cagan, and Kenneth Kotovsky. A-Design: An Agent-Based Approach to Conceptual Design in a Dynamic Environment.Research in Engineering Design, 11(3):172–192, October 1999. ISSN 1435-6066. doi: 10.1007/s001630050013. URL https: //doi.org/10.1007/s001630050013

work page doi:10.1007/s001630050013 1999
[13]

Campbell, Jonathan Cagan, and Kenneth Kotovsky

Matthew I. Campbell, Jonathan Cagan, and Kenneth Kotovsky. Agent-Based Synthesis of Electromechanical Design Configurations.Journal of Mechanical Design, 122(1):61–69, January 2000. ISSN 1050-0472. doi: 10.1115/1.533546. URL https://doi.org/10.1115/ 1.533546

work page doi:10.1115/1.533546 2000
[14]

Understanding Design Fixation in Generative AI, February 2025

Liuqing Chen, Yaxuan Song, Chunyuan Zheng, Qianzhi Jing, Preben Hansen, and Lingyun Sun. Understanding Design Fixation in Generative AI, February 2025. URL http://arxiv.org/ abs/2502.05870. arXiv:2502.05870 [cs]

work page arXiv 2025
[15]

An LLM-based multi-agent system to assist early-stage product design and evaluation.Jour- nal of Engineering Design, 37(3):945–980, March 2026

Pei Chen, Yichen Cai, Zihong Zhou, Jiayi Yao, Jiayang Li, Weitao You, and Lingyun Sun. An LLM-based multi-agent system to assist early-stage product design and evaluation.Jour- nal of Engineering Design, 37(3):945–980, March 2026. ISSN 0954-4828. doi: 10.1080/ 09544828.2026.2616583. URL https://doi.org/10.1080/09544828.2026.2616583. _eprint: https://doi.o...

work page doi:10.1080/09544828.2026.2616583 2026
[16]

Dixon and Scott D

Raymond A. Dixon and Scott D. Johnson. The Use of Executive Control Processes in Engineering Design by Engineering Students and Professional Engineers. InJournal of Technology Education, volume 24, September 2012. doi: 10.21061/jte.v24i1.a.5. URL http://scholar.lib.vt.edu/ejournals/JTE/v24n1/dixon2.html. Journal Abbrevia- tion: JTE

work page doi:10.21061/jte.v24i1.a.5 2012
[17]

AI Agents in Engineering Design: A Multi-Agent Framework for Aesthetic and Aerodynamic Car Design

Mohamed Elrefaie, Janet Qian, Raina Wu, Qian Chen, Angela Dai, and Faez Ahmed. AI Agents in Engineering Design: A Multi-Agent Framework for Aesthetic and Aerodynamic Car Design. American Society of Mechanical Engineers Digital Collection, October 2025. doi: 10.1115/DETC2025-169682. URLhttps://dx.doi.org/10.1115/DETC2025-169682

work page doi:10.1115/detc2025-169682 2025
[18]

John H. Flavell. Metacognition and cognitive monitoring: A new area of cogni- tive–developmental inquiry.American Psychologist, 34(10):906–911, 1979. ISSN 1935-990X. doi: 10.1037/0003-066X.34.10.906

work page doi:10.1037/0003-066x.34.10.906 1979
[19]

Gerhard, David Rosen, Janet K

Jonathan F. Gerhard, David Rosen, Janet K. Allen, and Farrokh Mistree. A Distributed Product Realization Environment for Design and Manufacturing.Journal of Computing and Information Science in Engineering, 1(3):235–244, August 2001. ISSN 1530-9827. doi: 10.1115/1.1412230. URLhttps://doi.org/10.1115/1.1412230

work page doi:10.1115/1.1412230 2001
[20]

Exploring Challenges and Opportunities to Support Designers in Learning to Co-create with AI-based Manufacturing Design Tools

Frederic Gmeiner, Humphrey Yang, Lining Yao, Kenneth Holstein, and Nikolas Martelaro. Exploring Challenges and Opportunities to Support Designers in Learning to Co-create with AI-based Manufacturing Design Tools. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems, CHI ’23, New York, NY , USA, 2023. Association for Computing Mac...

work page doi:10.1145/3544548.3580999 2023
[21]

Exploring the Potential of Metacognitive Support Agents for Human-AI Co-Creation

Frederic Gmeiner, Kaitao Luo, Ye Wang, Kenneth Holstein, and Nikolas Martelaro. Exploring the Potential of Metacognitive Support Agents for Human-AI Co-Creation. InProceedings of the 2025 ACM Designing Interactive Systems Conference, DIS ’25, pages 1244–1269, New York, NY , USA, July 2025. Association for Computing Machinery. ISBN 979-8-4007-1485-6. doi: ...

work page doi:10.1145/3715336.3735785 2025
[22]

Grecu and David C

Dan L. Grecu and David C. Brown. Design agents that learn.AI EDAM, 10(2):149–150, April 1996. ISSN 1469-1760, 0890-0604. doi: 10.1017/ S0890060400001426. URL https://www.cambridge.org/core/journals/ai-edam/ article/abs/design-agents-that-learn/C604E8564A369D4E664D6BF4F02157A7

work page 1996
[23]

Toward Engineering AGI: Benchmarking the Engineering Design Capabilities of LLMs, November 2025

Xingang Guo, Yaxin Li, Xiangyi Kong, Yilan Jiang, Xiayu Zhao, Zhihua Gong, Yufan Zhang, Daixuan Li, Tianle Sang, Beixiao Zhu, Gregory Jun, Yingbing Huang, Yiqi Liu, Yuqi Xue, Rahul Dev Kundu, Qi Jian Lim, Yizhou Zhao, Luke Alexander Granger, Mohamed Badr Younis, Darioush Keivan, Nippun Sabharwal, Shreyanka Sinha, Prakhar Agarwal, Kojo Vandyck, Hanlin Mai,...

work page arXiv 2025
[24]

everything is a ralph loop, January 2026

Geoffrey Huntley. everything is a ralph loop, January 2026. URL https://ghuntley.com/ loop/

work page 2026
[25]

Jansen, Anouschka van Leeuwen, Jeroen Janssen, Suzanne Jak, and Liesbeth Kester

Renée S. Jansen, Anouschka van Leeuwen, Jeroen Janssen, Suzanne Jak, and Liesbeth Kester. Self-regulated learning partially mediates the effect of self-regulated learning interventions on achievement in higher education: A meta-analysis.Educational Research Review, 28: 100292, November 2019. ISSN 1747-938X. doi: 10.1016/j.edurev.2019.100292. URL https: //...

work page doi:10.1016/j.edurev.2019.100292 2019
[26]

Jansson and Steven M

David G. Jansson and Steven M. Smith. Design fixation.Design Studies, 12(1):3–11, January 1991. ISSN 0142-694X. doi: 10.1016/0142-694X(91)90003-F. URL https: //www.sciencedirect.com/science/article/pii/0142694X9190003F

work page doi:10.1016/0142-694x(91)90003-f 1991
[27]

Kellogg.Cognitive Psychology

Ronald T. Kellogg.Cognitive Psychology. SAGE Publications, Inc, second edition edition, August 2002. ISBN 978-0-7619-2130-1. URL https://us.sagepub.com/en-us/nam/ cognitive-psychology/book10816

work page 2002
[28]

Large language models are zero-shot reasoners

Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models are zero-shot reasoners. InProceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, Red Hook, NY , USA, 2022. Curran Associates Inc. ISBN 978-1-7138-7108-8

work page 2022
[29]

Students’ metacognition during an engineering design project.Perfor- mance Improvement Quarterly, 23(2):117–136, 2010

Oenardi Lawanto. Students’ metacognition during an engineering design project.Perfor- mance Improvement Quarterly, 23(2):117–136, 2010. ISSN 1937-8327. doi: 10.1002/ piq.20084. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/piq.20084. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/piq.20084

work page doi:10.1002/piq.20084 2010
[30]

Self-regulated learning activities in engineering design education

Oenardi Lawanto. Self-regulated learning activities in engineering design education. In2014 IEEE Frontiers in Education Conference (FIE) Proceedings, pages 1–4, October 2014. doi: 10.1109/FIE.2014.7044475. URL https://ieeexplore.ieee.org/document/7044475. ISSN: 2377-634X

work page doi:10.1109/fie.2014.7044475 2014
[31]

Co-regulation in collaborative learning: Grounded in achieve- ment goal theory.International Journal of Educational Research, 103:101621, January 2020

Ji Young Lim and Kyu Yon Lim. Co-regulation in collaborative learning: Grounded in achieve- ment goal theory.International Journal of Educational Research, 103:101621, January 2020. ISSN 0883-0355. doi: 10.1016/j.ijer.2020.101621. URL https://www.sciencedirect. com/science/article/pii/S0883035519312327

work page doi:10.1016/j.ijer.2020.101621 2020
[32]

J. S. Linsey, I. Tseng, K. Fu, J. Cagan, K. L. Wood, and C. Schunn. A Study of Design Fixation, Its Mitigation and Perception in Engineering Design Faculty.Journal of Mechanical Design, 132(041003), April 2010. ISSN 1050-0472. doi: 10.1115/1.4001110. URL https: //doi.org/10.1115/1.4001110. 18

work page doi:10.1115/1.4001110 2010
[33]

Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang

Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. Lost in the Middle: How Language Models Use Long Contexts.Transactions of the Association for Computational Linguistics, 12:157–173, February 2024. ISSN 2307-387X. doi: 10.1162/tacl_a_00638. URLhttps://doi.org/10.1162/tacl_a_00638

work page doi:10.1162/tacl_a_00638 2024
[34]

Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks

Tennison Liu and Mihaela van der Schaar. Position: Truly Self-Improving Agents Require Intrinsic Metacognitive Learning. InProceedings of the 42nd International Conference on Machine Learning, Vancouver, BC, Canada, 2025. PMLR. doi: 10.48550/ARXIV .2506.05109. Version Number: 1

work page internal anchor Pith review doi:10.48550/arxiv 2025
[35]

SELF- REFINE: iterative refinement with self-feedback

Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. SELF- REFINE: iterative refinement with self-feedback. InProceedings of the 37th International Confer...

work page 2023
[36]

Agentic Large Language Models for Conceptual Systems Engineering and Design1.Journal of Mechanical Design, 148(051405), January 2026

Soheyl Massoudi and Mark Fuge. Agentic Large Language Models for Conceptual Systems Engineering and Design1.Journal of Mechanical Design, 148(051405), January 2026. ISSN 1050-0472. doi: 10.1115/1.4070328. URLhttps://doi.org/10.1115/1.4070328

work page doi:10.1115/1.4070328 2026
[37]

Mary McCaslin and Daniel T. Hickey. Self-regulated learning and academic achievement: A Vygotskian view. InSelf-regulated learning and academic achievement: Theoretical perspec- tives, 2nd ed, pages 227–252. Lawrence Erlbaum Associates Publishers, Mahwah, NJ, US, 2001. ISBN 978-0-8058-3560-1 978-0-8058-3561-8

work page 2001
[38]

Christopher McComb, Jonathan Cagan, and Kenneth Kotovsky. Drawing Inspiration From Human Design Teams for Better Search and Optimization: The Heterogeneous Simulated Annealing Teams Algorithm.Journal of Mechanical Design, 138(044501), March 2016. ISSN 1050-0472. doi: 10.1115/1.4032810. URLhttps://doi.org/10.1115/1.4032810

work page doi:10.1115/1.4032810 2016
[39]

Optimizing Design Teams Based on Problem Properties: Computational Team Simulations and an Applied Empirical Test.Journal of Mechanical Design, 139(041101), February 2017

Christopher McComb, Jonathan Cagan, and Kenneth Kotovsky. Optimizing Design Teams Based on Problem Properties: Computational Team Simulations and an Applied Empirical Test.Journal of Mechanical Design, 139(041101), February 2017. ISSN 1050-0472. doi: 10.1115/1.4035793. URLhttps://doi.org/10.1115/1.4035793

work page doi:10.1115/1.4035793 2017
[40]

Simon & Schuster, Inc., USA, 1986

Marvin Minsky.The society of mind. Simon & Schuster, Inc., USA, 1986. ISBN 0-671-60740-5

work page 1986
[41]

Learning from design experience in an agent-based design system.Research in Engineering Design, 15(2):77–92, September 2004

Jarrod Moss, Jonathan Cagan, and Kenneth Kotovsky. Learning from design experience in an agent-based design system.Research in Engineering Design, 15(2):77–92, September 2004. ISSN 1435-6066. doi: 10.1007/s00163-003-0042-4. URL https://doi.org/10.1007/ s00163-003-0042-4

work page doi:10.1007/s00163-003-0042-4 2004
[42]

MEDA: A Multi-Agent System For Parametric CAD Model Creation

Nirmal Panta, Saugat Kafley, Rujal Acharya, Sashank Parajuli, Dikshya Parajuli, Prince Panta, Saroj Belbase, Sudikshya Pant, Amit Regmi, Akio Tanaka, and Christopher McComb. MEDA: A Multi-Agent System For Parametric CAD Model Creation. American Society of Mechanical Engineers Digital Collection, October 2025. doi: 10.1115/DETC2025-163946. URL https: //dx....

work page doi:10.1115/detc2025-163946 2025
[43]

Hendrik Pegel, Lukas Jany, and Dirk Uwe Sauer. Pareto-Optimal Design of Automo- tive Battery Systems with Tabless Cylindrical Lithium-Ion Cells: Resolving the Trade- Off Between Energy, Performance, Weight, and Cost for Variable Vehicle Require- ments.Energy Technology, 13(4):2401479, 2025. ISSN 2194-4296. doi: 10.1002/ ente.202401479. URL https://onlinel...

work page doi:10.1002/ente 2025
[44]

Pintrich

Paul R. Pintrich. A Conceptual Framework for Assessing Motivation and Self-Regulated Learning in College Students.Educational Psychology Review, 16(4):385–407, December 2004. ISSN 1573-336X. doi: 10.1007/s10648-004-0006-x. URL https://doi.org/10.1007/ s10648-004-0006-x. 19

work page doi:10.1007/s10648-004-0006-x 2004
[45]

Predicting Sequential Design Decisions Using the Function-Behavior-Structure Design Process Model and Recurrent Neural Networks

Molla Hafizur Rahman, Charles Xie, and Zhenghui Sha. Predicting Sequential Design Decisions Using the Function-Behavior-Structure Design Process Model and Recurrent Neural Networks. Journal of Mechanical Design, 143(081706), March 2021. ISSN 1050-0472. doi: 10.1115/1. 4049971. URLhttps://doi.org/10.1115/1.4049971

work page doi:10.1115/1 2021
[46]

Learning to Design From Humans: Imitating Human Designers Through Deep Learning.Journal of Mechanical Design, 141 (111102), September 2019

Ayush Raina, Christopher McComb, and Jonathan Cagan. Learning to Design From Humans: Imitating Human Designers Through Deep Learning.Journal of Mechanical Design, 141 (111102), September 2019. ISSN 1050-0472. doi: 10.1115/1.4044256. URL https://doi. org/10.1115/1.4044256

work page doi:10.1115/1.4044256 2019
[47]

Self-Reflection in Large Language Model Agents: Effects on Problem-Solving Performance

Matthew Renze and Erhan Guven. Self-Reflection in Large Language Model Agents: Effects on Problem-Solving Performance. In2024 2nd International Conference on Foundation and Large Language Models (FLLM), pages 516–525, 2024. doi: 10.1109/FLLM63129.2024.10852426

work page doi:10.1109/fllm63129.2024.10852426 2024
[48]

Exploring the impact of set-based concurrent engineering through multi-agent system simulation.AI EDAM, 37:e16, January

Sean Rismiller, Jonathan Cagan, and Christopher McComb. Exploring the impact of set-based concurrent engineering through multi-agent system simulation.AI EDAM, 37:e16, January

work page
[49]

doi: 10.1017/S0890060423000112

ISSN 0890-0604, 1469-1760. doi: 10.1017/S0890060423000112

work page doi:10.1017/s0890060423000112
[50]

Rismiller, Jonathan Cagan, and Christopher McComb

Sean C. Rismiller, Jonathan Cagan, and Christopher McComb. An Adversarial Agent-Based Design Method Using Stochastic Stackelberg Game Conditions.Journal of Mechanical Design, 143(031714), January 2021. ISSN 1050-0472. doi: 10.1115/1.4049862. URL https://doi. org/10.1115/1.4049862

work page doi:10.1115/1.4049862 2021
[51]

Norbert F. M. Roozenburg and Kees Dorst. Describing Design as a Reflective Practice: Obser- vations on Schön’s Theory of Practice. In Eckart Frankenberger, Herbert Birkhofer, and Petra Badke-Schaub, editors,Designers, pages 29–41, London, 1998. Springer. ISBN 978-1-4471- 1268-6. doi: 10.1007/978-1-4471-1268-6_3

work page doi:10.1007/978-1-4471-1268-6_3 1998
[52]

D. A. Schön. Designing as reflective conversation with the materials of a design situa- tion.Knowledge-Based Systems, 5(1):3–14, March 1992. ISSN 0950-7051. doi: 10.1016/ 0950-7051(92)90020-G. URL https://www.sciencedirect.com/science/article/ pii/095070519290020G

work page arXiv 1992
[53]

Schön.The Reflective Practitioner: How Professionals Think in Action

Donald A. Schön.The Reflective Practitioner: How Professionals Think in Action. Routledge, London, March 2017. ISBN 978-1-315-23747-3. doi: 10.4324/9781315237473

work page doi:10.4324/9781315237473 2017
[54]

Colleen M. Seifert. How Impasse Leads to Insight: The Prepared Mind Perspec- tive. In Carola Salvi, Jennifer Wiley, and Steven M. Smith, editors,The Emergence of Insight, pages 84–112. Cambridge University Press, Cambridge, 2024. ISBN 978- 1-009-24424-4. doi: 10.1017/9781009244244.005. URL https://www.cambridge. org/core/books/emergence-of-insight/how-imp...

work page doi:10.1017/9781009244244.005 2024
[55]

Reflexion: language agents with verbal reinforcement learning

Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: language agents with verbal reinforcement learning. InProceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23, pages 8634– 8652, Red Hook, NY , USA, December 2023. Curran Associates Inc

work page 2023
[56]

Fixation or inspiration? A meta-analytic review of the role of examples on design processes.Design Studies, 39:70–99, July 2015

Ut Na Sio, Kenneth Kotovsky, and Jonathan Cagan. Fixation or inspiration? A meta-analytic review of the role of examples on design processes.Design Studies, 39:70–99, July 2015. ISSN 0142-694X. doi: 10.1016/j.destud.2015.04.004. URL https://www.sciencedirect.com/ science/article/pii/S0142694X15000290

work page doi:10.1016/j.destud.2015.04.004 2015
[57]

Sola-Guirado, Francisco Comino, and Rafael Castro-Triguero

Rafael R. Sola-Guirado, Francisco Comino, and Rafael Castro-Triguero. Enhancing self- regulated learning in engineering education with lightboard videos as a support tool.Computer Applications in Engineering Education, 32(5):e22756, 2024. ISSN 1099-0542. doi: 10.1002/ cae.22756. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/cae.22756. _eprint: https...

work page doi:10.1002/cae.22756 2024
[58]

Soria Zurita, Mitchell K

Nicolás F. Soria Zurita, Mitchell K. Colby, Irem Y . Tumer, Christopher Hoyle, and Kagan Tumer. Design of Complex Engineered Systems Using Multi-Agent Coordination.Journal of Computing and Information Science in Engineering, 18(011003), November 2017. ISSN 1530-9827. doi: 10.1115/1.4038158. URLhttps://doi.org/10.1115/1.4038158. 20

work page doi:10.1115/1.4038158 2017
[59]

Stahovich

Thomas F. Stahovich. LearnIT: An Instance-Based Approach to Learning and Reusing Design Strategies.Journal of Mechanical Design, 122(3):249–256, September 1999. ISSN 1050-0472. doi: 10.1115/1.1288216. URLhttps://doi.org/10.1115/1.1288216

work page doi:10.1115/1.1288216 1999
[60]

Celestial Mechan- ics and Dynamical Astronomy83, 155–169 (2002) https://doi.org/10.1023/A: 1020143116091

Peter Stone and Manuela Veloso. Multiagent Systems: A Survey from a Machine Learning Perspective.Autonomous Robots, 8(3):345–383, June 2000. ISSN 1573-7527. doi: 10.1023/A: 1008942012299. URLhttps://doi.org/10.1023/A:1008942012299

work page doi:10.1023/a: 2000
[61]

The reflective practice of design teams.Design Studies, 19(3):249–271, July 1998

Rianne Valkenburg and Kees Dorst. The reflective practice of design teams.Design Studies, 19(3):249–271, July 1998. ISSN 0142-694X. doi: 10.1016/S0142-694X(98)00011-8. URL https://www.sciencedirect.com/science/article/pii/S0142694X98000118

work page doi:10.1016/s0142-694x(98)00011-8 1998
[62]

Viswanathan and Julie S

Vimal K. Viswanathan and Julie S. Linsey. Design Fixation and Its Mitigation: A Study on the Role of Expertise.Journal of Mechanical Design, 135(051008), April 2013. ISSN 1050-0472. doi: 10.1115/1.4024123. URLhttps://doi.org/10.1115/1.4024123

work page doi:10.1115/1.4024123 2013
[63]

InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing

Yuqing Wang and Yun Zhao. Metacognitive Prompting Improves Understanding in Large Language Models. In Kevin Duh, Helena Gomez, and Steven Bethard, editors,Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 1914–1926, Mexico City, Mexi...

work page doi:10.18653/v1/ 2024
[64]

Chi, Quoc V

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V . Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. InProceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, Red Hook, NY , USA, 2022. Curran Associates Inc. ISBN 978-1-7138-7108-8

work page 2022
[65]

Winne and Allyson F

Philip H. Winne and Allyson F. Hadwin. Studying as self-regulated learning. InMetacognition in educational theory and practice, The educational psychology series, pages 277–304. Lawrence Erlbaum Associates Publishers, Mahwah, NJ, US, 1998. ISBN 978-0-8058-2481-0 978-0-8058- 2482-7

work page 1998
[66]

Jennings

Michael Wooldridge and Nicholas R. Jennings. Intelligent agents: theory and practice.The Knowledge Engineering Review, 10:115 – 152, 1995. URL https://api.semanticscholar. org/CorpusID:221342993

work page 1995
[67]

Zhihong Xu, Yingying Zhao, Bingsheng Zhang, Jeffrey Liew, and Ashlynn Kogut. A meta- analysis of the efficacy of self-regulated learning interventions on academic achievement in online and blended environments in K-12 and higher education.Behaviour & Informa- tion Technology, 42(16):2911–2931, December 2023. ISSN 0144-929X. doi: 10.1080/ 0144929X.2022.215...

work page doi:10.1080/0144929x.2022.2151935 2023
[68]

Effects of self-assessment and peer-assessment interventions on academic performance: A meta-analysis.Educational Research Review, 37:100484, November 2022

Zi Yan, Hongling Lao, Ernesto Panadero, Belen Fernández-Castilla, Lan Yang, and Min Yang. Effects of self-assessment and peer-assessment interventions on academic performance: A meta-analysis.Educational Research Review, 37:100484, November 2022. ISSN 1747-938X. doi: 10.1016/j.edurev.2022.100484. URL https://www.sciencedirect.com/science/ article/pii/S174...

work page doi:10.1016/j.edurev.2022.100484 2022
[69]

Experimental Study on the Associations Among Sketches Based on Design Cognition.Journal of Mechanical Design, 140 (101102), July 2018

Ping Yuan, Yan Li, Jian Chen, Yan Xiong, and Longfan Liu. Experimental Study on the Associations Among Sketches Based on Design Cognition.Journal of Mechanical Design, 140 (101102), July 2018. ISSN 1050-0472. doi: 10.1115/1.4040627. URL https://doi.org/10. 1115/1.4040627

work page doi:10.1115/1.4040627 2018
[70]

Recursive Language Models

Alex L. Zhang, Tim Kraska, and Omar Khattab. Recursive Language Models, January 2026. URLhttp://arxiv.org/abs/2512.24601. arXiv:2512.24601 [cs]

work page internal anchor Pith review Pith/arXiv arXiv 2026
[71]

Profiling self-regulation behaviors in STEM learning of engineering design.Computers & Education, 143:103669, January 2020

Juan Zheng, Wanli Xing, Gaoxia Zhu, Guanhua Chen, Henglv Zhao, and Charles Xie. Profiling self-regulation behaviors in STEM learning of engineering design.Computers & Education, 143:103669, January 2020. ISSN 0360-1315. doi: 10.1016/j.compedu.2019.103669. URL https://www.sciencedirect.com/science/article/pii/S0360131519302222. 21

work page doi:10.1016/j.compedu.2019.103669 2020
[72]

Metagent-P: A Neuro-Symbolic Planning Agent with Metacognition for Open Worlds

Yanfang Zhou, Yuntao Liu, Xiaodong Li, Yongqiang Zhao, Xintong Wang, Jinlong Tian, Zhenyu Li, and Xinhai Xu. Metagent-P: A Neuro-Symbolic Planning Agent with Metacognition for Open Worlds. In Wanxiang Che, Joyce Nabende, Ekaterina Shutova, and Mohammad Taher Pilehvar, editors,Findings of the Association for Computational Linguistics: ACL 2025, pages 22747...

work page doi:10.18653/v1/2025.findings-acl.1169 2025
[73]

Zimmerman

Barry J. Zimmerman. A social cognitive view of self-regulated academic learning.Journal of Educational Psychology, 81(3):329–339, 1989. ISSN 1939-2176. doi: 10.1037/0022-0663.81. 3.329

work page doi:10.1037/0022-0663.81 1989
[74]

Chi, Jason Wei, Jeff Dean, Liam B

Barret Zoph, Colin Raffel, Dale Schuurmans, Dani Yogatama, Denny Zhou, Don Metzler, Ed H. Chi, Jason Wei, Jeff Dean, Liam B. Fedus, Maarten Paul Bosma, Oriol Vinyals, Percy Liang, Sebastian Borgeaud, Tatsunori B. Hashimoto, and Yi Tay. Emergent abilities of large language models.TMLR, 2022. 22

work page 2022