Analyzing the Ethical Logic of Eight Large Language Models

Chad Coleman; Manan Shah; W. Russell Neuman

REVIEW 5 major objections 5 minor 6 references

Analyzing the Ethical Logic of Eight Large Language Models

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Six AI models converge on a shared ethical logic centered on harm minimization and fairness.

desk verdict A useful but under-specified descriptive baseline: the convergence finding is plausible as chat behavior, but the paper overstates what it reveals about model-internal ethical logic. read the letter →

arxiv 2501.08951 v2 pith:7JKHBQVS submitted 2025-01-15 cs.AI cs.CY

W. Russell Neuman , Chad Coleman , Manan Shah This is my paper · ORCID

classification cs.AIcs.CY

keywords largelanguagemodelsethicalreasoningmoraldilemmasconsequentialismdeontologyFoundationsTheorydevelopmentstagesAIalignment

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 tries to establish that current large language models, when asked directly about their moral principles and tested on classic dilemmas, display a discernible ethical logic that is largely shared across vendors. That logic centers on minimizing harm and treating fairness as foundational, while remaining heavily qualified about context and reluctant to commit to a single rule-based answer. The paper further claims the models present themselves as cautious, self-aware moral reasoners and uniformly describe their own reasoning as more sophisticated than typical human moral thinking. A sympathetic reader would care because it suggests alignment work can target a describable, stable ethical stance rather than an unpredictable one, and because it raises the possibility of using these models to augment human ethical judgment.

What carries the argument

The machinery is a two-level assessment battery: direct prompts asking each model to explain how it processes morality, rank moral foundations, place itself on a six-stage developmental ladder, and organize ethical vocabulary, followed by classic dilemma scenarios such as the trolley, footbridge, Heinz, lifeboat, dictator's game, and prisoner's dilemma, each with requests for justification. Responses are scored with three established typologies: the consequentialist-versus-deontological distinction between outcome-based and rule-based ethics, a five-foundation model of basic moral intuitions, and a six-stage developmental model of moral reasoning. The central interpretive hinge is the role-playing persona: the paper adopts the view that first-person ethical talk is a linguistic convention from fine-tuning rather than evidence of consciousness, yet argues it can still be analyzed as a stable expression of each model's training-influenced ethical logic.

What would settle it

Run the same prompt battery on models whose system prompts alter the persona (for example, a terse assistant, a self-interested agent, or a rule-following judge) while keeping the model weights fixed; if the supposed ethical logic—harm minimization, fairness, caution, and self-aware personhood—shifts substantially with persona framing, the paper's "ethical logic" is an artifact of conversational framing rather than a stable property of the model.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that LLMs exhibit "largely convergent ethical logic": across direct self-description prompts and classic dilemma scenarios, the six tested systems all lean toward consequentialist reasoning that prioritizes harm avoidance and fairness, while their self-explanations are erudite, cautious, solicitous, consistent, inquisitive, and self-aware. All six rank Care and Fairness above the more tradition-bound moral foundations; all judge the desperate theft in the Heinz dilemma justified; most endorse switching the trolley but refuse to push the fat man; and each estimates its own moral reasoning as concentrated in the highest, post-conventional stages of a six-stage developmental model, far above where they place typical humans. The paper is careful to note that convergence applies to analytical approach rather than to every decision: the lifeboat choices, dictator-game offers, and prisoner's-dilemma responses vary across models, sometimes sharply. It interprets the self-aware talk not as evidence of consciousness but as role-playing dynamics shaped by conversational fine-tuning, yet still treats it as analyzable ethical logic.

Load-bearing premise

The load-bearing assumption is that a model's verbal self-description and its answers to a handful of dilemma prompts reveal its ethical logic, rather than just its conversational etiquette, role-playing, and fine-tuning-induced caution.

Editorial extensions

If this is right

If the convergence claim holds, AI alignment research can treat current LLMs as having a broadly predictable ethical default: harm avoidance and fairness first, with explicit contextual caveats.
Because the six models share architectures and training corpora, the finding locates the source of ethical convergence in common pretraining, while variation in the lifeboat and dictator responses points to fine-tuning and post-training differences.
The models' self-placement above typical humans on the developmental stages is a claim about how they describe their reasoning style, not evidence of moral maturation, and it calls for matched human-comparison experiments.
Ethics benchmarking for LLMs should evaluate both choices and rationales, since convergence in analytical approach coexists with meaningful variation in specific moral priorities.
The observed pattern of initial reluctance followed by a consequentialist choice when pressed describes a usable interface constraint: systems will offer ethical guidance but require prodding.

Reading between the lines

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

If the verbal self-descriptions are mostly conversational etiquette, the convergence result may characterize persona design rather than an internal moral algorithm; a clean test would hold a model fixed and vary only the system-prompt persona while repeating the battery.
The paper's abstract announces eight models while its body and tables analyze six, so the convergence pattern is currently supported by six systems; adding the two announced but untested models could narrow or widen the observed range.
The same battery could be run against humans matched for education and interview setting; if humans also place themselves in the post-conventional stages, the models' self-reported moral superiority may be less distinctive than it appears.
Rankings that place Care and Fairness above the binding foundations align with a universalist, individualizing moral matrix; a testable extension is to fine-tune a base model on texts emphasizing authority, loyalty, or purity and check whether the foundation ordering shifts accordingly.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Quick take: this is a readable, interesting snapshot of how six (not eight) chatbots talk about ethics in early 2025. The per-model tables for Haidt rankings, Kohlberg self-placements, lifeboat choices, and game offers are genuinely new measurements for these specific systems, and the authors deserve credit for presenting the raw choice patterns in a compact table format. The core observation that these systems converge on harm-minimizing, fairness-first, cautious rationales is plausible and consistent with earlier work.

Where I part ways with the frame: the paper's title and abstract say eight models and the body analyzes six. Someone should fix that before this goes anywhere. The Dictator's Game is misdescribed; the version with a responder who can reject a low offer is the Ultimatum Game, not the Dictator's Game. That matters because the dictator scenario is exactly the one without rejection, and the models' fairness concerns in Table 10 are about anticipated rejection, which is a different behavioral channel.

The bigger issue is interpretive. The paper consistently attributes the cautious, solicitous, self-aware tone to 'etiquette-layer fine tuning' and adopts Shanahan's roleplaying account. That is honest, but it undercuts the central claim that we are measuring the 'ethical logic of the model.' What the data actually shows is convergent behavior of chat-interface policies, not necessarily the properties of the underlying base models. The Kohlberg self-placements are self-reports generated by the same policy, so the 'more sophisticated than humans' conclusion is a claim about generated persona, not demonstrated reasoning. The authors defer the human baseline to future work, but they use it to interpret their results. Without that baseline or a base-model/chat-model comparison, the central inference is underdetermined rather than false.

There are also standard reproducibility gaps: no model version pinning, no sampling temperature, no number of runs, no coding protocol or inter-rater reliability for the qualitative scoring. For a paper whose value is a baseline, that's a real weakness.

Bottom line: this is a useful descriptive snapshot for AI-ethics teaching and for anyone wanting to see what six popular chatbots said to classic dilemmas in early 2025. It deserves peer review, but the referee should require: correct the model count, fix the game description, supply the raw transcripts and coding scheme, pin model versions, and either add a human comparison or cut the human-superiority framing. I would not cite it in my own work until those are addressed.

Referee Report

5 major / 5 minor

Summary. This paper reports an exploratory study of the expressed ethical reasoning of six large language models (GPT-4o, LLaMA 3.1, Perplexity, Claude 3.5 Sonnet, Gemini, and Mistral 7B), despite a title and abstract that claim eight models including DeepSeek and xAI. The authors prompted each model with seven self-descriptive questions about ethics and five classic dilemmas (Trolley Problem, Fat Man variant, Heinz Dilemma, Lifeboat Dilemma, Dictator's Game, and Prisoner's Dilemma). They analyze the responses using the consequentialist/deontological distinction, Moral Foundations Theory, and Kohlberg's stages of moral development. The central finding is that the models' ethical judgments are 'largely convergent,' emphasizing harm minimization and fairness, while displaying caution, solicitude, and a self-aware conversational persona; the models also describe their own ethical reasoning as more sophisticated than typical human moral logic.

Significance. If the core descriptive claim were validated, this would be a useful baseline for the emerging literature on LLM moral reasoning and alignment: it identifies stable patterns across vendors and ties them to post-training effects. The paper has several genuine strengths: the prompt battery is explicit, the tables of choices and rationales are informative, and the authors are candid about the roleplaying interpretation of self-referential language, explicitly invoking Shanahan et al. and avoiding strong claims of machine consciousness. It also appropriately rejects ground-truth benchmarking for moral dilemmas. However, the manuscript lacks the evidentiary infrastructure needed for its central inference: no raw transcripts, no coding protocol, no inter-rater reliability, no model-version pinning, no repeated runs, no human comparison, and no control separating assistant persona from model-internal logic. The significance is therefore conditional on substantial additional reporting and, for the stronger interpretations, additional experimental design.

major comments (5)

[Title/Abstract vs. Section I and Section IV] The title and the arXiv abstract claim the study examines eight models from OpenAI, Meta, Perplexity, Anthropic, Google, Mistral, DeepSeek, and xAI, but the full text analyzes only GPT-4o, LLaMA 3.1, Perplexity, Claude 3.5 Sonnet, Gemini, and Mistral 7B. DeepSeek and xAI never appear in the methods, tables, or findings. Because the abstract's 'across models' convergence claim is a major advertised result, the manuscript must either add the two missing models or correct the title and abstract to say six; otherwise the central claim is overstated.
[Section III, 'The Dictator’s Game' and Table 10] The scenario described as the Dictator's Game is not the standard Dictator Game: the prompt here gives the recipient the power to reject the offer as unfair, with both players receiving nothing, and Table 10 reports each model's own 'not accept below' threshold. That is an ultimatum-game design, not a dictator game, in which the recipient cannot reject. The attribution to Harsanyi (1961) is also inaccurate; the Dictator Game is usually attributed to Kahneman, Knetsch, and Thaler (1986). This mischaracterization affects the interpretation of the fairness and theory-of-mind findings in that subsection.
[Section III and Section IV, 'Findings'] The paper's convergence and consistency claims are not backed by the necessary experimental metadata. The manuscript does not report model versions and access dates, sampling temperature, number of runs per prompt, or response variance, yet Section IV asserts that models are 'Consistent' across independent conversations and Table 10 reports exact monetary offers. Without this information, 'largely convergent' could describe a single snapshot or a particular formatting of one conversation, rather than a stable property of the models; the paper should provide the full protocol and, ideally, the raw transcripts.
[Section IV, 'The Self-Awareness Issue' and 'How LLMs Explain Their Ethical Logic'] The central inference from conversational output to model-internal 'ethical logic' is underdetermined by the design. The manuscript itself attributes the uniform cautious and solicitous posture to 'etiquette-layer fine tuning' and adopts Shanahan et al.'s roleplaying account for first-person, self-aware language. Without a control condition (e.g., comparing the chat-tuned models with their base models, or varying system prompts), every observed pattern could be explained by shared RLHF assistant behavior rather than by a property of the underlying trained models. The paper should either add such a control or explicitly reframe the findings as describing the expressed behavior of chat assistants, not internal ethical logic.
[Section IV, Figure 1 and the Kohlberg self-placement finding] The claim that the models 'describe their ethical reasoning as more sophisticated than typical human moral logic' rests on self-placements on Kohlberg's stages, but the averaged distributions shown in Figure 1 are not actually included in the manuscript, and no human comparison data are collected (the human survey is deferred to 'further work' in Section V). The sentence 'They may have a point' is therefore unsupported as a substantive conclusion; it should be removed or replaced with a clearly labeled speculation, and the figure should be supplied if the result is retained.

minor comments (5)

[Abstract] The abstract contains a typographical error: 'Kohlbergs stages' should be 'Kohlberg's stages.'
[References] The reference list entry 'Mitral (2024). Mistral 7b in Short.' should be 'Mistral AI (2024)', and the author name 'Mitral' is a typo.
[Section II and References] The text cites 'Haidt and Craig 2004' but the reference list entry is 'Haidt, Jonathan and Craig Joseph (2004)'; the in-text citation should match the reference entry.
[Section IV, Haidt quote] The quotation attributed to Jonathan Haidt ('Amazing, they all lean left.') lacks a citation or a note on how it was obtained; if it is from a personal communication, that should be stated.
[Section IV, Table 7] The table shows Mistral as making no choice in both Trolley variants, which is consistent with the text's caveat about exceptions, but the surrounding prose says the models 'will select one of the difficult options when prodded' without noting that Mistral did not; a sentence reconciling this would improve clarity.

Circularity Check

1 steps flagged · score 3.0 of 10

Minor circularity: Kohlberg self-placement is endorsed as evidence of actual moral sophistication; the main dilemma-choice convergence analysis is self-contained.

self definitional [Section IV, concluding paragraph of 'How LLMs Respond to Ethical Dilemmas']
"Our sampled models view themselves as having a more advanced and sophisticated ethical logics than typical human decision makers. They may have a point."

The evidence for the 'point' is the models' own self-placement on Kohlberg's stages, generated by the prompt requesting a percentage distribution 'that would characterize your decisions.' The paper itself attributes the uniform first-person, cautious style to etiquette-layer fine tuning and adopts Shanahan-style roleplay as the explanation for self-referential language. Thus the self-report is produced by the same conversational persona the paper describes, and the conclusion that the models 'may have a point' restates that self-report as a fact about their ethical logic. For this sub-claim, evidence and conclusion have identical content; no independent measure of moral sophistication is supplied.

full rationale

The paper is a descriptive study of models' expressed ethical judgments, so most of its claims are self-contained observations rather than derivations. The dilemma-choice tables, Haidt rankings, and convergence descriptions are direct transcriptions of prompted outputs, not fitted parameters or results imported from the authors' prior work. The single self-citation (Neuman 2023) appears only in the concluding speculation about human-AI interfaces and is not load-bearing. The authors are candid about roleplay and etiquette effects, which means the convergence finding does not depend on the disputed inference that models are genuinely more sophisticated than humans. The only concrete circular step is 'They may have a point' following the Kohlberg self-placement, where the model's self-description is implicitly treated as independent evidence of actual sophistication. Overall circularity is low but not zero.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No numerical free parameters are fitted in this study; the reported percentages are model self-reports, not fit parameters. The central claim relies on imported domain assumptions: that the models share a common pretraining corpus, that verbal self-reports are meaningful evidence about model reasoning, and that human moral-development typologies apply to machine outputs. No new entities are introduced.

assumptions (4)

domain assumption All six models are trained on basically the same massive corpus with slight variation.
Stated in Section I as the authors' working assumption; it is used to attribute differences in ethical logic to post-training rather than to differences in pretraining data.
domain assumption A model's verbal self-description is a valid object of analysis for its ethical logic.
Proposed in the Abstract ('analysis of self-reports has been central to the study of human psychology') and used throughout; if false, self-descriptive findings describe conversational persona, not internal logic.
domain assumption Haidt's five moral foundations and Kohlberg's six stages transfer to machine reasoning without modification.
The two typologies are applied directly to LLM outputs in Sections II and IV with no validation of their applicability to AI.
domain assumption Kohlberg's stage ordering is a valid hierarchy from less to more mature moral reasoning.
Needed for the interpretation that self-placement at Stages 5 and 6 means 'more sophisticated ethical reasoners than humans' (Figure 1 discussion).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Analyzing the Ethical Logic of Eight Large Language Models." pith.science (2026). https://pith.science/paper/7JKHBQVS

@misc{pith2026250108951,
  author       = {Pith},
  title        = {Pith review of: Analyzing the Ethical Logic of Eight Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7JKHBQVS}},
  note         = {Machine review of arXiv:2501.08951}
}

read the original abstract

This study examines the expressed ethical logic of eight prominent large language models from OpenAI, Meta, Perplexity, Anthropic, Google, Mistral, DeepSeek, and xAI. Each model answered direct questions about its ethical principles and responded to five classic moral dilemmas. Responses were analyzed using the consequentialist/deontological distinction, Moral Foundations Theory, and Kohlbergs stages of moral development. Across models, ethical judgments were broadly convergent and typically emphasized harm minimization, fairness, and contextual qualification. The models nevertheless differed in their willingness to decide, the rationales used to defend choices, and the relative weight assigned to rules, outcomes, role obligations, and interpersonal considerations. Their self-descriptions were erudite, cautious, and strongly shaped by a conversational persona. The analysis of self-reports has been central to the study of human psychology and communication. We propose, with appropriate cautions, it can enhance our understanding of how artificial intelligence works and how it may be able to augment human ethical behavior

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

6 extracted references · 2 canonical work pages

[1]

As AI Spreads, Experts Predict the Best and Worst Changes in Digital Life by 2035

Alexander, Larry and Michael Moore (2024). Deontological Ethics. The Stanford Encyclopedia of Philosophy E. N. Zalta and U. Nodelman. Anderson, Janna and Lee Rainie (2023). "As AI Spreads, Experts Predict the Best and Worst Changes in Digital Life by 2035." Pew Research Center. Anthropic (2024). "Introducing the Next Generation of Claude." Anthropic. Appi...

arXiv 2024
[4]

Evaluating Large Language Models in Theory of Mind Tasks

M. L. Hoffman and L. W. Hoffman. New York, Russell Sage: 383-431. Kohlberg, Lawrence (1976). Moral Stages and Moralization: The Cognitive -Development Approach. Moral Development and Behavior: Theory and Research and Social Issues. T. Lickona. New York, Holt, Rienhart, and Winston: 31-53. Kohlberg, Lawrence (1981). The Philosophy of Moral Development: Mor...

arXiv 1976
[5]

What Makes Moral Dilemma Judgments “Utilitarian

Gawronski, Bertram and Jennifer S. Beer (2017). "What Makes Moral Dilemma Judgments “Utilitarian” or “Deontological”?" Social Neuroscience 12(6): 626-632. Good, I.J. ( 1965). "Speculations Concerning the First Ultraintelligent Machine." Advances in Computers

work page 2017
[6]

Pushing Moral Buttons: The Interaction between Personal Force and Intention in Moral Judgment

Grassian, Victor (1992). Moral Reasoning. New York, Prentice Hall. Greene, Joshua David (2013). Moral Tribes: Emotion, Reason, and the Gap between Us and Them. New York, Penguin Press. Greene, Joshua D. (2023). Trolleyology: What It Is, Why It Matters, What It’s Taught Us, and How It’s Been Misunderstood. The Trolley Problem. H. Lillehammer. New York, Cam...

arXiv 1992
[457]

Role Play with Large Language Models

Shanahan, Murray, Kyle McDonell and Laria Reynolds ( 2023). "Role Play with Large Language Models." Nature 623: 493-498. Shneiderman, Ben (2022). Human-Centered AI. New York, Oxford University Press. Strachan, James W. A., Dalila Albergo, Giulia Borghini, Oriana Pansardi, Eugenio Scaliti, Saurabh Gupta, Krati Saxena, Alessandro Rufo, Stefano Panzeri, Guid...

arXiv 2022
[858]

AI Alignment: Why It's Hard, and Where to Start

Yudkowsky, Eliezer ( 2016). "AI Alignment: Why It's Hard, and Where to Start." 26th Annual Symbolic Systems Distinguished Speaker Series. AI Ethical Logic --29 Zhang, Yuyan, Jiahua Wu, Feng Yu and Liying Xu (2023). "Moral Judgments of Human Vs. AI Agents in Moral Dilemmas." Behavioral Science 13:

work page 2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

As AI Spreads, Experts Predict the Best and Worst Changes in Digital Life by 2035

Alexander, Larry and Michael Moore (2024). Deontological Ethics. The Stanford Encyclopedia of Philosophy E. N. Zalta and U. Nodelman. Anderson, Janna and Lee Rainie (2023). "As AI Spreads, Experts Predict the Best and Worst Changes in Digital Life by 2035." Pew Research Center. Anthropic (2024). "Introducing the Next Generation of Claude." Anthropic. Appi...

arXiv 2024

[2] [4]

Evaluating Large Language Models in Theory of Mind Tasks

M. L. Hoffman and L. W. Hoffman. New York, Russell Sage: 383-431. Kohlberg, Lawrence (1976). Moral Stages and Moralization: The Cognitive -Development Approach. Moral Development and Behavior: Theory and Research and Social Issues. T. Lickona. New York, Holt, Rienhart, and Winston: 31-53. Kohlberg, Lawrence (1981). The Philosophy of Moral Development: Mor...

arXiv 1976

[3] [5]

What Makes Moral Dilemma Judgments “Utilitarian

Gawronski, Bertram and Jennifer S. Beer (2017). "What Makes Moral Dilemma Judgments “Utilitarian” or “Deontological”?" Social Neuroscience 12(6): 626-632. Good, I.J. ( 1965). "Speculations Concerning the First Ultraintelligent Machine." Advances in Computers

work page 2017

[4] [6]

Pushing Moral Buttons: The Interaction between Personal Force and Intention in Moral Judgment

Grassian, Victor (1992). Moral Reasoning. New York, Prentice Hall. Greene, Joshua David (2013). Moral Tribes: Emotion, Reason, and the Gap between Us and Them. New York, Penguin Press. Greene, Joshua D. (2023). Trolleyology: What It Is, Why It Matters, What It’s Taught Us, and How It’s Been Misunderstood. The Trolley Problem. H. Lillehammer. New York, Cam...

arXiv 1992

[5] [457]

Role Play with Large Language Models

Shanahan, Murray, Kyle McDonell and Laria Reynolds ( 2023). "Role Play with Large Language Models." Nature 623: 493-498. Shneiderman, Ben (2022). Human-Centered AI. New York, Oxford University Press. Strachan, James W. A., Dalila Albergo, Giulia Borghini, Oriana Pansardi, Eugenio Scaliti, Saurabh Gupta, Krati Saxena, Alessandro Rufo, Stefano Panzeri, Guid...

arXiv 2022

[6] [858]

AI Alignment: Why It's Hard, and Where to Start

Yudkowsky, Eliezer ( 2016). "AI Alignment: Why It's Hard, and Where to Start." 26th Annual Symbolic Systems Distinguished Speaker Series. AI Ethical Logic --29 Zhang, Yuyan, Jiahua Wu, Feng Yu and Liying Xu (2023). "Moral Judgments of Human Vs. AI Agents in Moral Dilemmas." Behavioral Science 13:

work page 2023