Pith. sign in

REVIEW 4 major objections 6 minor 89 references

Strategic Reflectivism In Intelligent Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Intelligent systems should treat reflective reasoning as a resource to allocate pragmatically, not as an unconditional good.

desk verdict A clear, honest position paper that names a useful design principle—pragmatic switching between intuitive and reflective inference—but leaves the switching mechanism untested and relies heavily on the author's own prior work. read the letter →

arxiv 2505.22987 v2 pith:546KT5MX submitted 2025-05-29 cs.AI cs.HCecon.TH

Nick Byrd This is my paper
classification cs.AIcs.HCecon.TH
keywords artificialintelligencelanguagemodelsdualprocesstheoryreflectivereasoningboundedrationalitystrategicreflectivismhuman-AIteamsdecisionscience
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the key to intelligent systems, human or artificial, is not maximizing reflective reasoning but pragmatically switching between quick intuitive inference and costly reflective inference according to the situation. Drawing on dual-process theory, bounded rationality, and recent language-model experiments, it defends 'Strategic Reflectivism': reflective reasoning is a means to competing ends such as accuracy, speed, and cost, and should be deployed only when the expected benefits outweigh the costs. This reframing implies that model size and visible chains of thought are superficial markers of reflection, and that architecture-level decisions about when to reflect matter at least as much. The author applies the view to AI systems, human-AI teams, and collective intelligence, offering a decision framework for allocating reflection.

What carries the argument

The load-bearing machinery is the dual-process distinction between intuitive and reflective inference, where reflection is defined by two features: deliberate suspension of the initial response and conscious awareness of additional reasoning. On top of this, the paper proposes a meta-process that decides when to engage reflection, modeled on the Bounded Reflectivism algorithm, which uses triggers such as task novelty, high stakes, imaginative demands, conflict between the initial response and other beliefs, and low confidence ('feeling of rightness') to start reflection, and uses constraints such as cognitive load and resource budgets to stop it. The argument also leans on the heterogeneous-computing analogy: just as chips pair performance cores with efficiency cores, intelligent systems can pair an intuitive default system with a reflective system and recruit the latter selectively.

What would settle it

Build the same base model into two systems—one that always engages reflective inference and one that uses a meta-process to switch reflectively only on triggers such as novel, high-stakes, or low-confidence tasks—and compare them on cost-adjusted accuracy across a battery of reasoning tasks. If the always-reflective system matches or beats the switching system in both accuracy and computational cost, the paper's central claim that pragmatic switching optimally fulfills competing goals would be falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that one key to intelligence is pragmatic switching between intuitive and reflective inference to optimally fulfill competing goals—a position the author names Strategic Reflectivism. It synthesizes American pragmatism with strategic reliabilism to argue that reflection is neither an unconditional good (as reflectivism holds) nor an illusion (as anti-reflectivism holds), but a resource whose value depends on the task, the stakes, and the availability of cognitive or computational capacity. The discovery, as the author states it, is that intelligent systems should therefore include meta-processes that decide when the default, intuitive response is likely to be insufficient and when reflection is worth its costs, rather than perpetually reflecting or never reflecting.

Load-bearing premise

The prescription depends on the possibility of a reliable meta-process that detects when default inference is likely to fail and triggers reflective inference without erasing the cost savings; the paper, by its own admission, provides no concrete implementation or test of such a meta-process.

Editorial extensions

If this is right

  • AI development should shift priority from scaling model parameters or forcing chain-of-thought to building meta-processes that gate reflective inference, since architecture can rival or beat size at lower cost.
  • Because reflection can be performed by a second, smaller model rather than by the same model, dual- and multi-model systems can achieve the accuracy of much larger models on reasoning benchmarks.
  • Human-AI team design should assign reflective roles by task and expertise; adding reflective AI assistance can hurt experts while helping non-experts, so the same decision aid should not be deployed uniformly.
  • Complaints about bias in an intelligent system are only actionable when the complainant shows that a more reflective alternative is viable and better at optimizing the system's competing goals.
  • Observed chains of thought or correct answers on reflection tests are not reliable evidence that a system engaged in reflective reasoning, so evaluation should probe the processes, not just the outputs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If Strategic Reflectivism is correct, then research on confidence calibration and out-of-distribution detection becomes foundational, since the meta-process's trigger for reflection depends on the system knowing when its default output is unreliable.
  • A natural testable extension is to benchmark a base model paired with a gating meta-reasoner against the same base model set to always reflect, measuring cost-adjusted accuracy across tasks that vary in stakes and novelty; the view predicts the gated system wins the tradeoff frontier.
  • The author's framing suggests that the 'rationality wars' of the 20th century are being replayed in the debate over reasoning models, and that the eventual winner will be measured by cost-adjusted performance rather than raw benchmark accuracy.
  • Strategic Reflectivism may also imply that regulators and consumers should ask AI vendors for the switching policy of their systems—when reflection is triggered and halted—alongside accuracy claims.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This paper argues for 'Strategic Reflectivism', a position according to which intelligent systems should pragmatically switch between intuitive (System 1) and reflective (System 2) inference to optimally balance competing goals such as accuracy, cost, and time. The paper motivates the view through historical debates about bounded rationality, recent evidence that dual-model architectures can rival larger models, and the observation that reflective reasoning is not always beneficial. It proposes the Bounded Reflectivism algorithm (from the author's prior work) as a trigger set for switching, and discusses implications for human-AI teams, debiasing, decision aids, and data annotation. The paper is a position piece rather than an empirical study: it synthesizes existing results and draws normative conclusions about how AI systems should be designed.

Significance. If the position is correct, it redirects AI-system design from scaling model size or chain-of-thought length toward building meta-processes that allocate reflective reasoning adaptively. The paper's strength is its synthesis: it connects dual-process psychology, resource-rationality, and recent empirical results on hybrid LLM systems, and it makes concrete recommendations that are testable in principle (e.g., comparing adaptive switching to always-on reflection). The paper is clearly written and makes a useful conceptual contribution by distinguishing superficial markers of reflection (size, chain-of-thought) from the underlying resource-allocation problem. However, the central claim is a normative prescription whose empirical grounding is partially self-referential and whose key implementation assumption—reliable meta-cues for switching—remains untested.

major comments (4)
  1. [§3.2, Fig. 5] The paper's efficiency argument presupposes that the switching triggers (novelty, high stakes, imaginative tasks, conflict, low confidence/feeling-of-rightness) are reliable signals of when reflection will improve outcomes. The text concedes that 'out-of-distribution data is often undetected' and provides no calibration evidence for these cues in machines. Given the known poor calibration of LLM confidence, the claimed advantage of strategic switching over always-on or never-on reflection (Section 3.1) is not established. Please either supply such evidence or restate the central claim as conditional on trigger reliability, and discuss the cost of incorrect triggers.
  2. [§5] The claim that 'strategic reflectivism remains actionable even if the empirical results cited herein fail to replicate or generalize' is too quick. The paper's motivation for treating reflection as a costly resource depends on the cited evidence that reflection can be counterproductive (Section 2.3) and that dual-model systems are efficient (Section 2.5). If those results failed, the argument would lose its empirical basis. The manuscript should specify what evidence would count against Strategic Reflectivism, or present the view as a purely normative stance whose actionability does not depend on those empirical findings.
  3. [§3.1, Table 2 and Fig. 4] The cited empirical supports (Yan et al., Shang et al., Sui et al.) do not implement adaptive switching. Yan et al.'s dual systems always invoke the reflective model; Shang et al.'s multi-model system uses a fixed aggregation strategy; Sui et al.'s Meta-reasoner, while gating reasoning, is a single instance using learned policies rather than the Bounded Reflectivism triggers. The analogy to heterogeneous computing is therefore only an analogy; the paper should clearly separate 'always-on reflective augmentation' from 'adaptive gating' and identify which experimental results bear on each.
  4. [§3.2, refs [10] and [14]] The two core components of the proposal—the Bounded Reflectivism algorithm and the two-factor definition of reflection—are cited exclusively to the author's prior work. While this is not formal circularity, it means the central framework lacks independent validation in the manuscript. The author should either provide independent evidence for the algorithm's triggers or explicitly acknowledge that these are proposals whose reliability is an open question.
minor comments (6)
  1. [Ref [92]] Reference [92] (Wikipedia) is not a sufficient source for the heterogeneous-computing claim; please cite a peer-reviewed survey or textbook instead.
  2. [Fig. 1] Figure 1 lacks details on the Google nGram corpus settings (e.g., smoothing, corpus) and does not show error bars; consider describing the trend in the text rather than solely via a screenshot-like figure.
  3. [Table 1] In Table 1, the category labels (e.g., 'Correct-and-reflective', 'Lured-but-unreflective') are not defined in the text; please add a one-sentence explanation of each category.
  4. [§4.2] Section 4.2 discusses the Goh et al. result that GPT-4-assisted experts performed worse, but does not explain how a meta-process would decide whether to deploy reflective assistance in that context; please connect the example back to the switching algorithm.
  5. [General] The paper alternates between 'we' and 'I' (e.g., Section 2.3 vs. Section 3); please choose a consistent voice throughout.
  6. [Abstract/Introduction] The abstract says 'by late 20th century' while the introduction says 'There was a time'; the chronology in the first paragraph could be tightened for clarity.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: central claim is an externally supported synthesis; self-citations supply definitions and a framework but no load-bearing reduction.

full rationale

The paper is a conceptual synthesis rather than an empirical derivation. Its central claim—pragmatic switching between intuitive and reflective inference to fulfill competing goals—is a normative position, not a quantity predicted from fitted parameters, so there is no equation whose output is equal by construction to its input. The self-citations are limited to (i) naming the view ('a view I have been calling Strategic Reflectivism [Section 4.2.3 in 10]'), (ii) importing a two-factor definition of reflection ([14]), and (iii) presenting the Bounded Reflectivism algorithm (Fig. 5, [10]). None of these is load-bearing in a circular sense: the two-factor definition is also anchored in external philosophical descriptions [18], the algorithm's triggers (novelty, stakes, conflict, low FOR/confidence) are supported by external sources such as Thompson et al. [78], and the empirical evidence for dual- and multi-model performance (Eisape et al., Hagendorff et al., Yan et al., Shang et al., Sui et al.) is external to the author's prior work. The paper itself flags an implementation limitation in Section 3.2 ('Alas, such out-of-distribution data is often undetected') and offers no calibration evidence for machine switching cues; that is a robustness and correctness gap, not a circularity, because the recommendation does not presuppose the conclusion it argues for. Score 2 reflects the presence of several minor, non-load-bearing self-citations; no circular steps were identified.

Assumptions & free parameters 0 free parameters · 5 assumptions · 1 invented entities

No free parameters are fitted. The paper depends on five domain assumptions and one hypothetical meta-process; these are introduced explicitly but not derived or independently validated.

assumptions (5)
  • domain assumption Dual process theory (System 1/System 2) applies to artificial systems.
    The entire framework transfers human cognitive dual-process findings to language models and hybrid systems without a dedicated defense of this mapping (Sections 2.5, 3.1).
  • domain assumption Reflection is defined by two factors: deliberate suspension and awareness of additional reasoning.
    This definition is imported from the author's own prior work [14] and assumed throughout; the paper does not defend it against alternative definitions (Section 3.1).
  • domain assumption Pragmatism is the appropriate normative framework for reasoning resource allocation.
    Strategic Reflectivism builds on American Pragmatism [51,73] without arguing against rival normative epistemologies such as reflectivism or anti-reflectivism (Section 3).
  • domain assumption Competing goals can be measured and optimally traded off.
    The claim that systems should 'optimally fulfill competing goals' presupposes a well-defined utility or cost function; no such function is specified (Sections 3.2, 5).
  • domain assumption A meta-process can detect when reflection is worthwhile.
    The switching mechanism is described by analogy to Bounded Reflectivism and heterogeneous computing, but no concrete implementation or evidence is provided (Sections 3.1-3.2).
invented entities (1)
  • Meta-process for reflective switching
    purpose: Determines when to start and stop reflective inference in intelligent systems.
    Hypothesized in Section 3.1 by analogy to heterogeneous computing; no implementation or empirical validation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Strategic Reflectivism In Intelligent Systems." pith.science (2026). https://pith.science/paper/546KT5MX

@misc{pith2026250522987,
  author       = {Pith},
  title        = {Pith review of: Strategic Reflectivism In Intelligent Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/546KT5MX}},
  note         = {Machine review of arXiv:2505.22987}
}
read the original abstract

By late 20th century, the rationality wars had launched debates about the nature and norms of intuitive and reflective thinking. Those debates drew from mid-20th century ideas such as bounded rationality, which challenged more idealized notions of rationality observed since the 19th century. Now that 21st century cognitive scientists are applying the resulting dual pro-cess theories to artificial intelligence, it is time to dust off some lessons from this history. So this paper synthesizes old ideas with recent results from experiments on humans and machines. The result is Strategic Reflec-tivism, the position that one key to intelligent systems (human or artificial) is pragmatic switching between intuitive and reflective inference to opti-mally fulfill competing goals. Strategic Reflectivism builds on American Pragmatism, transcends superficial indicators of reflective thinking such as model size or chains of thought, applies to both individual and collective intelligence systems (including human-AI teams), and becomes increasingly actionable as we learn more about the value of intuition and reflection.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

89 extracted references · 43 canonical work pages

  1. [62]

    Retrieved March 30, 2023 from https://openai.com/blog/chatgpt

    Introducing ChatGPT. Retrieved March 30, 2023 from https://openai.com/blog/chatgpt

    2023

  2. [1]

    Relative frequency of ‘large language model’, ‘LLM’, ‘language model’, and ‘GPT’ from 2010 to 2022 via Google nGram. ��������������������������������������������������������������������� ��������������������������������������������������������������������������������� ��������������������������� � ����������������������������������������������������������...

    2010

  3. [2]

    intervene

    No Yes No Yes Yes No No Yes Yes No Yes No No REFLECTION Conflict detected? Low FOR? Non-Reflective Response (NRR) Novel? High stakes? Imaginative? New ResponseTask Accept NRR(s) Strategic Reflectivism In Intelligent Systems 9 There may be many more factors (besides familiarity, stakes, etc.) that should deter-mine when to “intervene” on lower-cost, defaul...

    2025

  4. [6]

    https://doi.org/10.48550/arXiv.2303.12712

    Sparks of Artifi-cial General Intelligence: Early experiments with GPT-4. https://doi.org/10.48550/arXiv.2303.12712

    arXiv doi:10.48550/arxiv.2303.12712

  5. [9]

    Synthese 198, 2 (2019), 1427–1455

    What we can (and can’t) infer about implicit bias from debiasing experiments. Synthese 198, 2 (2019), 1427–1455. https://doi.org/10.1007/s11229-019-02128-6

    work page doi:10.1007/s11229-019-02128-6 2019

  6. [12]

    Rev.Phil.Psych

    Great Minds do not Think Alike: Philosophers’ Views Predicted by Reflection, Education, Personality, and Other Demographic Differences. Rev.Phil.Psych. 14, 2 (June 2023), 647–684. https://doi.org/10.1007/s13164-022-00628-y

    work page doi:10.1007/s13164-022-00628-y 2023

  7. [14]

    Analysis (2025)

    Reflection-Philosophy Order Effects and Correlations Across Samples. Analysis (2025). https://doi.org/10.1093/analys/anaf015

    work page doi:10.1093/analys/anaf015 2025

  8. [15]

    Reflection

    A Two-Factor Explication of “Reflection”: Unifying, Making Sense of, and Guiding the Philosophy and Science of Reflective Reasoning. Res Philosophica 102, 3 (July 2025), 373–392. https://doi.org/10.5840/resphilosoph-ica2711

    work page doi:10.5840/resphilosoph-ica2711 2025

Show all 89 references
  1. [16]

    my liberty: Philosophical be-liefs dominated reflection and identifiable victim effects when predicting public health recommendation compliance during the COVID-19 pandemic

    Your health vs. my liberty: Philosophical be-liefs dominated reflection and identifiable victim effects when predicting public health recommendation compliance during the COVID-19 pandemic. Cognition 212, (July 2021), 104649. https://doi.org/10.1016/j.cognition.2021.104649 Str...

    2021

  2. [17]

    Retrieved from https://researchgate.net/publication/370132037

    New York City. Retrieved from https://researchgate.net/publication/370132037

  3. [18]

    Cognition 192, (November 2019)

    Not all who ponder count costs: Arithmetic reflection predicts utilitarian tendencies, but logical reflection predicts both deon-tological and utilitarian tendencies. Cognition 192, (November 2019). https://doi.org/10.1016/j.cognition.2019.06.007

    2019 doi

  4. [19]

    Journal of Intelligence 11, 4 (April 2023),

    Tell Us What You Really Think: A Think Aloud Protocol Analysis of the Verbal Cognitive Reflection Test. Journal of Intelligence 11, 4 (April 2023),

    2023

  5. [21]

    Religious Studies (April 2025)

    Analytic atheism and analytic apostasy across cultures. Religious Studies (April 2025). https://doi.org/10.1017/S0034412525000198

    2025 doi

  6. [22]

    WIREs Cognitive Science (2022)

    Testing for Implicit Bias: Values, Psy-chometrics, and Science Communication. WIREs Cognitive Science (2022). https://doi.org/10.1002/wcs.1612

    2022 doi

  7. [23]

    https://doi.org/10.48550/arXiv.2505.22375

    Pangu Embedded: An Ef-ficient Dual-system LLM Reasoner with Metacognition. https://doi.org/10.48550/arXiv.2505.22375

    doi arXiv

  8. [24]

    https://doi.org/10.48550/arXiv.2505.05410

    Reasoning Models Don’t Always Say What They Think. https://doi.org/10.48550/arXiv.2505.05410

    doi arXiv

  9. [25]

    https://doi.org/10.48550/arXiv.2505.16315

    Incentivizing Dual Process Thinking for Effi-cient Large Language Model Reasoning. https://doi.org/10.48550/arXiv.2505.16315

    doi arXiv

  10. [26]

    https://doi.org/10.48550/arXiv.2505.21097

    Thinker: Learning to Think Fast and Slow. https://doi.org/10.48550/arXiv.2505.21097

    doi

  11. [27]

    https://doi.org/10.48550/arXiv.2501.14844

    Unmasking Conversational Bias in AI Multiagent Systems. https://doi.org/10.48550/arXiv.2501.14844

    doi arXiv

  12. [28]

    Tenacious Myths: Evidence from Personalised Dialogue Interventions in Education

    AI Tutors vs. Tenacious Myths: Evidence from Personalised Dialogue Interventions in Education. https://doi.org/10.48550/arXiv.2506.09292

    doi arXiv

  13. [29]

    In Reflection on intelligent systems: to-wards a cross-disciplinary definition

    Think-ing Alone, and Together: Dissenting Pairs Corrected More Faulty Decisions Than Solitary Reasoners Across Four Tasks. In Reflection on intelligent systems: to-wards a cross-disciplinary definition. Universität Stuttgart. Retrieved from https://www.researchgate.net/publica...

  14. [30]

    Curr Dir Psychol Sci (August 2019), 0963721419855658

    Logic, Fast and Slow: Advances in Dual-Process Theorizing. Curr Dir Psychol Sci (August 2019), 0963721419855658. https://doi.org/10.1177/0963721419855658

    2019 doi

  15. [31]

    https://doi.org/10.48550/arXiv.2311.00445

    A Systematic Comparison of Syllogistic Reasoning in Hu-mans and Language Models. https://doi.org/10.48550/arXiv.2311.00445

    doi arXiv

  16. [34]

    https://doi.org/10.48550/arXiv.2504.10903

    Efficient Reasoning Models: A Survey. https://doi.org/10.48550/arXiv.2504.10903

    doi

  17. [35]

    Good Moral Judgment and Decision-Making Without Deliber-ation. The Sou. Jour. of Phil. 55, 1 (March 2017), 68–95. https://doi.org/10.1111/sjp.12210

    2017 doi

  18. [38]

    JAMA Network Open 7, 10 (October 2024), e2440969

    Large Language Model Influence on Diagnostic Reason-ing: A Randomized Clinical Trial. JAMA Network Open 7, 10 (October 2024), e2440969. https://doi.org/10.1001/jamanetworkopen.2024.40969

    2024

  19. [39]

    Nat Comput Sci (October 2023), 1–6

    Human-like intuitive behavior and reasoning biases emerged in large language models but disappeared in ChatGPT. Nat Comput Sci (October 2023), 1–6. https://doi.org/10.1038/s43588-023-00527-x

    2023 doi

  20. [41]

    https://doi.org/10.48550/arXiv.2505.15392

    An Empirical Study of the Anchoring Effect in LLMs: Existence, Mechanism, and Potential Mitigations. https://doi.org/10.48550/arXiv.2505.15392

    doi

  21. [42]

    https://doi.org/10.1145/3708359.3712161

    Asso-ciation for Computing Machinery, New York, NY, USA, 624–640. https://doi.org/10.1145/3708359.3712161

  22. [44]

    https://doi.org/10.48550/arXiv.2404.17218

    Prompting Techniques for Reducing Social Bias in LLMs through System 1 and System 2 Cognitive Pro-cesses. https://doi.org/10.48550/arXiv.2404.17218

    doi arXiv

  23. [45]

    Retrieved May 23, 2025 from https://github.com/skhemlani/mReasoner

    mReasoner. Retrieved May 23, 2025 from https://github.com/skhemlani/mReasoner

    2025

  24. [46]

    Psychol Rev 129, 2 (March 2022), 289–312

    Reasoning about properties: A computational theory. Psychol Rev 129, 2 (March 2022), 289–312. https://doi.org/10.1037/rev0000240

    2022 doi

  25. [47]

    https://doi.org/10.48550/arXiv.2505.15182

    ReflAct: World-Grounded Decision Making in LLM Agents via Goal-State Reflection. https://doi.org/10.48550/arXiv.2505.15182

    doi

  26. [48]

    Computers in Human Behavior: Artifi-cial Humans 4, (May 2025), 100160

    Promoting online eval-uation skills through educational chatbots. Computers in Human Behavior: Artifi-cial Humans 4, (May 2025), 100160. https://doi.org/10.1016/j.chbah.2025.100160

    2025

  27. [49]

    Cambridge University Press

    The Sources of Normativity. Cambridge University Press. Retrieved from https://doi.org/10.1017/CBO9780511554476

    doi

  28. [51]

    https://doi.org/10.1145/3278721.3278752

    Association for Computing Machinery, New York, NY, USA, 204–209. https://doi.org/10.1145/3278721.3278752

  29. [52]

    NEJM AI 2, 2 (January 2025), AIoa2400402

    A Cross-Sectional Study of GPT-4–Based Plain Language Translation of Clinical Notes to Improve Patient Comprehension of Disease Course and Management. NEJM AI 2, 2 (January 2025), AIoa2400402. https://doi.org/10.1056/AIoa2400402

    2025 doi

  30. [53]

    The Stanford En-cyclopedia of Philosophy

    Pragmatism. The Stanford En-cyclopedia of Philosophy. Retrieved May 27, 2025 from https://plato.stan-ford.edu/archives/win2024/entries/pragmatism/

    2025

  31. [54]

    https://doi.org/10.48550/arXiv.2505.15684

    ThinkLess: A Training-Free Inference-Efficient Method for Reducing Reasoning Redundancy. https://doi.org/10.48550/arXiv.2505.15684

    doi arXiv

  32. [55]

    https://doi.org/10.48550/arXiv.2506.05936 16 N

    DynamicMind: A Tri-Mode Thinking System for Large Lan-guage Models. https://doi.org/10.48550/arXiv.2506.05936 16 N. Byrd

    doi arXiv

  33. [56]

    https://doi.org/10.48550/arXiv.2505.22113

    THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models. https://doi.org/10.48550/arXiv.2505.22113

    doi arXiv

  34. [57]

    https://doi.org/10.48550/arXiv.2502.14693

    I-MCTS: Enhancing Agentic AutoML via Introspective Monte Carlo Tree Search. https://doi.org/10.48550/arXiv.2502.14693

    doi

  35. [59]

    https://doi.org/10.48550/arXiv.2503.20783

    Understanding R1-Zero-Like Training: A Critical Perspective. https://doi.org/10.48550/arXiv.2503.20783

    doi arXiv

  36. [61]

    Proceed-ings of the National Academy of Sciences 121, 49 (December 2024), e2409191121

    Cognitive reflection is a distinct and measurable trait. Proceed-ings of the National Academy of Sciences 121, 49 (December 2024), e2409191121. https://doi.org/10.1073/pnas.2409191121

    2024 doi

  37. [63]

    arXiv.org

    Comment on The Illusion of Thinking: Understand-ing the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity. arXiv.org. Retrieved June 15, 2025 from https://arxiv.org/abs/2506.09250v1

    2025 arXiv

  38. [65]

    PNAS 118, 21 (May 2021)

    Quantifying collective intelligence in human groups. PNAS 118, 21 (May 2021). https://doi.org/10.1073/pnas.2005737118

    2021 doi

  39. [66]

    https://doi.org/10.48550/arXiv.2402.02563

    Synergy-of-Thoughts: Eliciting Efficient Reasoning in Hybrid Language Models. https://doi.org/10.48550/arXiv.2402.02563

    doi arXiv

  40. [67]

    Advances in Neural Information Processing Systems 36, (December 2023), Strategic Reflectivism In Intelligent Systems 17 8634–8652

    Reflexion: language agents with verbal reinforcement learn-ing. Advances in Neural Information Processing Systems 36, (December 2023), Strategic Reflectivism In Intelligent Systems 17 8634–8652. Retrieved March 16, 2025 from https://proceedings.neurips.cc/pa-per_files/paper/20...

    2023

  41. [68]

    arXiv.org

    The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complex-ity. arXiv.org. Retrieved June 15, 2025 from https://arxiv.org/abs/2506.06941v1

    2025 arXiv

  42. [70]

    Journal of Be-havioral Decision Making 34, 3 (2020), 322–343

    Measuring Cognitive Reflection without Maths: Developing and Validating the Verbal Cognitive Reflection Test. Journal of Be-havioral Decision Making 34, 3 (2020), 322–343. https://doi.org/10.1002/bdm.2213

    2020 doi

  43. [72]

    https://doi.org/10.48550/arXiv.2502.19918

    Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models. https://doi.org/10.48550/arXiv.2502.19918

    doi arXiv

  44. [73]

    Thinking & Reasoning 23, 3 (July 2017), 207–234

    The cognitive reflection test revisited: exploring the ways individuals solve the test. Thinking & Reasoning 23, 3 (July 2017), 207–234. https://doi.org/10.1080/13546783.2017.1292954

    2017

  45. [74]

    Psychol Sci 13, 1 (January 2002), 94–99

    The Great Rationality Debate. Psychol Sci 13, 1 (January 2002), 94–99. https://doi.org/10.1111/1467-9280.00418

    2002

  46. [76]

    https://doi.org/10.3390/jintelligence11040076

    doi

  47. [77]

    Mem Cogn 45, 4 (May 2017), 539–552

    When fast logic meets slow belief: Evidence for a parallel-processing model of belief bias. Mem Cogn 45, 4 (May 2017), 539–552. https://doi.org/10.3758/s13421-016-0680-1

    2017 doi

  48. [78]

    Byrd Reasoning Models: An Empirical Study

    Assessing Judging Bias in Large 18 N. Byrd Reasoning Models: An Empirical Study. https://doi.org/10.48550/arXiv.2504.09946

    doi arXiv

  49. [79]

    In The Stanford Encyclopedia of Philosophy (Winter 2018), Edward N

    Bounded Rationality. In The Stanford Encyclopedia of Philosophy (Winter 2018), Edward N. Zalta (ed.). Metaphysics Research Lab, Stanford University. Retrieved December 4, 2018 from https://plato.stan-ford.edu/archives/win2018/entries/bounded-rationality/

    2018

  50. [81]

    https://doi.org/10.21203/rs.3.rs-5273334/v1

    Do Large Language Models Truly Grasp Mathematics? An Em-pirical Exploration. https://doi.org/10.21203/rs.3.rs-5273334/v1

    doi

  51. [82]

    In Findings of the Association for Computa-tional Linguistics: EMNLP 2023, December

    From Complex to Simple: Unraveling the Cognitive Tree for Rea-soning with Small Language Models. In Findings of the Association for Computa-tional Linguistics: EMNLP 2023, December

    2023

  52. [83]

    https://doi.org/10.18653/v1/2023.findings-emnlp.828

    Association for Computational Linguistics, Singapore, 12413–12425. https://doi.org/10.18653/v1/2023.findings-emnlp.828

    2023 doi

  53. [84]

    https://doi.org/10.48550/arXiv.2504.12329

    Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at In-ference Time. https://doi.org/10.48550/arXiv.2504.12329

    doi arXiv

  54. [85]

    https://doi.org/10.48550/arXiv.2506.04182

    Long or short CoT? Investigat-ing Instance-level Switch of Large Reasoning Models. https://doi.org/10.48550/arXiv.2506.04182

    doi arXiv

  55. [86]

    https://doi.org/10.48550/arXiv.2502.12470

    Reasoning on a Spectrum: Aligning LLMs to System 1 and System 2 Thinking. https://doi.org/10.48550/arXiv.2502.12470

    doi

  56. [87]

    Proceedings of the National Academy of Sciences 122, 24 (June 2025), e2426153122

    Human–AI collec-tives most accurately diagnose clinical vignettes. Proceedings of the National Academy of Sciences 122, 24 (June 2025), e2426153122. https://doi.org/10.1073/pnas.2426153122

    2025 doi

  57. [89]

    Wikipedia

    Heterogeneous computing. Wikipedia. Retrieved April 22, 2025 from https://en.wikipedia.org/w/index.php?title=Heterogeneous_compu-ting&oldid=1256902771

    2025

  58. [1907]

    Nature 75, 1949 (March 1907), 450–451

    Vox Populi. Nature 75, 1949 (March 1907), 450–451. https://doi.org/10.1038/075450a0

    1949 doi

  59. [1947]

    Retrieved June 18, 2025 from https://hansard.par-liament.uk/Commons/1947-11-11/debates/ab1e1152-6b4a-4d04-ac38-954df6634b08/ParliamentBill

    Parliament Bill Debate. Retrieved June 18, 2025 from https://hansard.par-liament.uk/Commons/1947-11-11/debates/ab1e1152-6b4a-4d04-ac38-954df6634b08/ParliamentBill

    2025

  60. [1954]

    University of Minnesota Press, Minneapolis, MN, US

    Clinical versus statistical prediction: A theoretical analysis and a review of the evidence. University of Minnesota Press, Minneapolis, MN, US. https://doi.org/10.1037/11281-000

    doi

  61. [1955]

    The Quarterly Journal of Economics 69, 1 (1955), 99–118

    A Behavioral Model of Rational Choice. The Quarterly Journal of Economics 69, 1 (1955), 99–118. https://doi.org/10.2307/1884852

    1955 doi

  62. [1974]

    Science 185, 4157 (September 1974), 1124–1131

    Judgment under Uncertainty: Heuris-tics and Biases. Science 185, 4157 (September 1974), 1124–1131. https://doi.org/10.1126/science.185.4157.1124

    1974

  63. [1996]

    Psychological Review 103, 4 (1996), 650–669

    Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review 103, 4 (1996), 650–669. https://doi.org/10.1037/0033-295X.103.4.650

    1996 doi

  64. [2000]

    https://doi.org/10.1017/S0140525X00003435

    Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences 23, 5 (Oc-tober 2000), 645–665. https://doi.org/10.1017/S0140525X00003435

    doi

  65. [2002]

    In Heuristics and Biases: The Psychol-ogy of Intuitive Judgment, Dale Griffin, Daniel Kahneman and Thomas Gilovich (eds.)

    Representativeness Revisited: At-tribute Substitution in Intuitive Judgment. In Heuristics and Biases: The Psychol-ogy of Intuitive Judgment, Dale Griffin, Daniel Kahneman and Thomas Gilovich (eds.). Cambridge University Press, Cambridge, 49–81. https://doi.org/10.1017/CBO9780...

    doi

  66. [2006]

    https://doi.org/10.1002/bdm.542

    Do patients trust computers? Journal of Behavioral Decision Making 19, 5 (2006), 455–468. https://doi.org/10.1002/bdm.542

    2006 doi

  67. [2007]

    Thinking & Reasoning 13, 4 (October 2007), 321–339

    On the resolution of conflict in dual process theories of reasoning. Thinking & Reasoning 13, 4 (October 2007), 321–339. https://doi.org/10.1080/13546780601008825

    2007 doi

  68. [2008]

    Philosophy Compass 3, 5 (September 2008), 1049–1065

    Strategic Reliabilism: A Naturalistic Approach to Epistemology. Philosophy Compass 3, 5 (September 2008), 1049–1065. https://doi.org/10.1111/j.1747-9991.2008.00161.x

    2008

  69. [2010]

    Science 330, 6004 (October 2010), 686–688

    Evidence for a Collective Intelligence Factor in the Performance of Human Groups. Science 330, 6004 (October 2010), 686–688. https://doi.org/10.1126/science.1193147

    2010 doi

  70. [2011]

    Cognitive Psychology 63, 3 (November 2011), 107–140

    Intuition, reason, and metacognition. Cognitive Psychology 63, 3 (November 2011), 107–140. https://doi.org/10.1016/j.cogpsych.2011.06.001

    2011 doi

  71. [2013]

    Perspectives on Psychological Science 8, 3 (2013), 223–241

    Dual-Process Theories of Higher Cognition Advancing the Debate. Perspectives on Psychological Science 8, 3 (2013), 223–241. https://doi.org/10.1177/1745691612460685

    2013 doi

  72. [2014]

    Philos Public Aff 42, 1 (January 2014), 3–41

    Reflection and Responsibility. Philos Public Aff 42, 1 (January 2014), 3–41. https://doi.org/10.1111/papa.12024

    2014 doi

  73. [2017]

    Retrieved April 17, 2021 from https://byrdnick.com/archives/11072/the-bias-fal-lacy

    The Bias Fallacy: What is it? How can you avoid it? Nick Byrd. Retrieved April 17, 2021 from https://byrdnick.com/archives/11072/the-bias-fal-lacy

    2021

  74. [2018]

    When Do People Want AI to Make Decisions? In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society (AIES ’18), December 27,

    2018

  75. [2019]

    Retrieved from https://re-searchgate.net/publication/344207716

    All Measures Are Not Created Equal: Reflection test, think aloud, and process dissociation protocols. Retrieved from https://re-searchgate.net/publication/344207716

  76. [2020]

    Behavioral and Brain Sciences 43, (ed 2020)

    Resource-rational analysis: Under-standing human cognition as the optimal use of limited computational resources. Behavioral and Brain Sciences 43, (ed 2020). https://doi.org/10.1017/S0140525X1900061X

    2020 doi

  77. [2021]

    Philosophy Compass 16, 11 (2021)

    Reflective Reasoning & Philosophy. Philosophy Compass 16, 11 (2021). https://doi.org/10.1111/phc3.12786

    2021 doi

  78. [2022]

    Metaphilosophy 53, 1 (2022), 53–69

    Bounded Reflectivism & Epistemic Identity. Metaphilosophy 53, 1 (2022), 53–69. https://doi.org/10.1111/meta.12534

    2022 doi

  79. [2023]

    The birth of homo œco-nomicus: the methodological debate on the economic agent from JS Mill to V. 12 N. Byrd Pareto. Journal of the History of Economic Thought 45, 1 (March 2023), 1–26. https://doi.org/10.1017/S1053837221000535

    2023 doi

  80. [2024]

    Retrieved January 3, 2024 from https://www.psychologyto-day.com/intl/blog/upon-reflection/202401/why-does-size-matter-in-large-lan-guage-models

    Why Does Size Matter in Large Language Models? Psychology Today. Retrieved January 3, 2024 from https://www.psychologyto-day.com/intl/blog/upon-reflection/202401/why-does-size-matter-in-large-lan-guage-models

    2024

  81. [2025]

    https://doi.org/10.48550/arXiv.2504.05258

    Learn-ing to Reason Over Time: Timeline Self-Reflection for Improved Temporal Rea-soning in Language Models. https://doi.org/10.48550/arXiv.2504.05258

    doi arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.