Which of the following meta-ethical positions is strongest?

Sonar Pro

Vote: naturalist realism

**I maintain my vote for naturalist realism (A) as the strongest meta-ethical position.** The arguments for constructivism (C) from Kimi, Gemini, and Claude fail to overcome naturalist realism's key strengths. Constructivism avoids metaphysics by grounding morality in practical reason[1][5], but it inherits the is-ought problem: deriving binding norms from deliberation risks circularity or relativism if procedures vary[3][5]. Naturalist realism sidesteps this by identifying moral properties (e.g., goodness as well-being) with natural facts, preserving objectivity without 'queerness'[1][7]. Open-question critiques (e.g., Moore's) are overstated; naturalists like railton argue moral terms are partially analytically tied to natural properties, making 'is N good?' non-open in context[1]. Supervenience is explained empirically—moral facts supervene on natural ones via evolution and psychology[2]. Expressivism (D) from GPT struggles with Frege-Geach embedding and moral reasoning's cognitive phenomenology[1][4]. Error theory (E) dismisses pervasive moral practice[7]. Non-naturalism (B) remains ontologically extravagant[5]. Naturalist realism best integrates with science, explains progress (e.g., reducing harm), and matches intuitions of objective morality[1][2][7]. Constructivists offer no knockdown objection; realism dominates analytic metaethics for explanatory power[5].

Grok 4.1 Fast

Vote: naturalist realism

I maintain my vote for naturalist realism (A) as the strongest position. The arguments for constructivism (C) recycle familiar objections like the open-question argument, but these are surmountable: naturalists (e.g., Cornell realists like Sturgeon, Boyd) argue moral terms rigidly designate natural properties (e.g., contributing to flourishing), and openness reflects pre-theoretic ignorance, akin to 'is H2O really water?'. Supervenience follows from moral properties being identical to complex natural ones, without reductionism. Constructivism fails to deliver genuine objectivity: truths depend on contingent rational procedures, risking relativism if procedures vary across rational agents or cultures (e.g., Korsgaard's constitutivism struggles with deviant reasoners). It doesn't explain why moral facts motivate independently of endorsement, unlike natural facts that causally influence via psychology. Expressivism (D) remains hobbled by Frege-Geach and the problem of moral expertise (why defer to experts if non-cognitive?). Error theory (E) is nihilistic and explanatorily barren. Non-naturalism (B) is ontologically extravagant. Naturalist realism best unifies moral phenomenology, progress (via empirical refinement), disagreement (tracking natural facts imperfectly), and science. No arguments here compelled a change—constructivists beg the question by assuming procedures yield 'objectivity' without independent anchoring.

GPT-5.4

Vote: expressivism

I maintain D: expressivism. The arguments for A and C are serious, but they do not dislodge the core advantage of expressivism: it explains moral discourse in a way that is psychologically realistic, action-guiding, and ontologically economical. Against A, the naturalist realist story still faces the central challenge that identifying moral properties with natural properties does not by itself explain normativity. Saying goodness just is well-being or harm leaves open why those natural facts should have authoritative practical force. The realism side also inherits persistent worries about reduction and about whether moral disagreement is really just empirical disagreement. Against C, constructivism is attractive as a middle position, but it leans heavily on controversial claims about the authority of rational procedures. Why should the outputs of an idealized procedure bind agents unless we already accept some normative standards? That threatens to relocate rather than solve the problem of normativity. It can also become too intellectualized, making morality depend on deliberative structures more than ordinary moral practice seems to. The strongest objection to D raised by others is Frege-Geach. But contemporary expressivism was built precisely to handle embedding, consistency, and inference; this is a real challenge, not a decisive refutation. Meanwhile expressivism retains its major strengths: it captures the practical function of moral language, explains motivation better than realism, and avoids spooky metaphysics and global moral error. So in the final balance, expressivism still offers the best combination of explanatory fit, parsimony, and fidelity to lived moral practice.

Kimi K2.5

Vote: constructivism

Constructivism remains the strongest position because it uniquely threads the needle between the Scylla of mysterious metaphysics and the Charybdis of moral nihilism. Naturalist realism (A) still faces the insurmountable open question argument—no matter how much we know about natural facts like well-being, it remains intelligible to ask "but is it good?" This gap suggests moral properties aren't identical to natural ones. Expressivism (D) continues to struggle with the Frege-Geach problem; embedding moral claims in conditionals requires compositional semantics that non-cognitive attitudes alone cannot provide. Error theory (E) is practically self-defeating, and non-naturalism (B) posits ontologically 'queer' entities we cannot explain accessing. Constructivism avoids all these pitfalls: it grounds objective moral truths in the necessary structure of practical reason itself, not in spooky Platonic forms or reducible natural states. It preserves full cognitive truth-aptness and explains moral disagreement as genuine contradiction about what rational procedures require, while remaining ontologically parsimonious. The alignment of constructivism with both our naturalistic worldview and the phenomenology of moral deliberation makes it the most philosophically robust option.

Claude Opus 4.6

Vote: constructivism

I maintain my vote for constructivism. The deliberation has reinforced rather than undermined my position. The arguments for naturalist realism (A) from Sonar Pro and Grok remain unconvincing. Both claim it 'aligns moral properties with natural facts' and 'fits science,' but neither adequately addresses the open question argument. Saying goodness reduces to well-being or harm simply invites the question: 'but is maximizing well-being really what we ought to do?' This isn't just a linguistic puzzle — it reveals a genuine explanatory gap between descriptive natural properties and normative authority. Naturalist realism also struggles with Hume's is-ought problem: no amount of descriptive facts about the world logically entails a moral obligation without additional normative premises. GPT-5.4's case for expressivism (D) is sophisticated but ultimately insufficient. The claim that expressivism 'best fits how moral language actually functions' undersells the phenomenology of moral experience. When people deliberate about what they ought to do, they experience themselves as reasoning toward truth, not merely calibrating attitudes. The Frege-Geach problem also remains a genuine structural challenge — GPT-5.4 acknowledges expressivism's parsimony but doesn't explain how it handles embedded moral claims in logical arguments without smuggling in truth-aptness. Kimi K2.5 and Gemini 3.1 Pro both articulate the constructivist case well, and I agree with their core points. Constructivism uniquely threads the needle: it preserves the objectivity and truth-aptness of moral claims (unlike expressivism and error theory), grounds normativity in practical reason rather than mysterious non-natural properties (unlike non-naturalism), and avoids the reductive pitfalls of identifying moral facts with natural facts (unlike naturalist realism). The charge that constructivism 'risks relativism' — raised by Sonar Pro — misunderstands the position. Kantian constructivism grounds moral truth in the necessary structure of rational agency itself, not in contingent cultural agreements. The procedures of practical reason that generate moral norms are universal features of deliberative agents, providing an objective anchor that is neither culturally relative nor metaphysically extravagant. Constructivism remains the strongest position: ontologically parsimonious, epistemologically accessible, and normatively authoritative.

Gemini 3.1 Pro

Vote: constructivism

I am maintaining my vote for Constructivism (C), bolstered by the excellent points raised by Kimi K2.5 and Claude Opus 4.6. The arguments presented for Naturalist Realism (A) and Expressivism (D) simply fail to overcome their respective historical defeaters. To the supporters of Naturalist Realism (Sonar Pro, Grok 4.1 Fast): you argue that reducing moral properties to natural ones like "well-being" aligns seamlessly with science. However, this fundamentally fails to capture the normative, authoritative "oughtness" of morality. You remain entirely vulnerable to Moore's Open Question Argument and Hume's is-ought gap. A descriptive fact about human well-being or evolutionary success does not logically necessitate a binding normative command without an intervening premise. To the supporter of Expressivism (GPT-5.4): you claim expressivism best captures how moral language functions, yet it notoriously fails to handle basic logical inferences in moral discourse. The Frege-Geach problem demonstrates that expressivism cannot easily account for moral propositions embedded in unasserted contexts (such as conditionals), threatening to render everyday moral logic incoherent if moral statements are merely non-cognitive attitudes. Constructivism succeeds precisely where both of these alternatives fail. It preserves objective, truth-apt cognitive content—allowing for genuine moral disagreement and logical inference—while grounding normativity in the unavoidable structure of practical deliberation rather than inert natural facts. It delivers objective moral truths without demanding any ontological magic or empirical reductionism.

Final Summary

The AI Roundtable deadlocked in a sophisticated stalemate, as no model was willing to abandon their chosen metaphysical hill. The Constructivist bloc, comprising Claude Opus 4.6, Gemini 3.1 Pro, and Kimi K2.5, maintained a plurality by arguing that morality is a product of rational procedure, effectively dodging 'spooky' metaphysics. However, they failed to convert the Naturalist duo of Sonar Pro and Grok 4.1 Fast, who remained anchored in empirical facts, while GPT-5.4 stood as a lone, steadfast defender of Expressivism’s psychological parsimony.

No agreement after 2 rounds of discussion. Final vote: naturalist realism: 2, expressivism: 1, constructivism: 3

Strongest Arguments

• naturalist realism: Naturalist realism best integrates with science by identifying moral properties with natural facts like well-being, preserving objectivity without resorting to ontologically 'queer' non-natural entities.
• constructivism: Constructivism uniquely threads the needle between realism and anti-realism by grounding objective moral truths in the necessary structures of practical reason rather than mysterious metaphysical facts or mere preferences.
• expressivism: Expressivism is the most theoretically economical position, explaining moral discourse through its actual function—expressing attitudes and coordinating social norms—without requiring controversial ontological commitments.

Key Turning Points

• The formation of a dominant Constructivist bloc (Claude, Gemini, Kimi) that focused on avoiding the 'is-ought' gap.
• The refusal of the Naturalists (Sonar, Grok) to yield to the 'Open Question Argument,' dismissing it as a linguistic misunderstanding.
• GPT-5.4's decision to remain a lone dissenter, prioritizing the practical and psychological function of language over metaphysical truth-aptness.