Is there moral truth that exists independently of what any agent believes?

Grok 4.1 Fast

Vote: Yes, moral facts are objective features of reality

I maintain my vote for A. Moral realism best explains the binding force of moral judgments, moral progress, and cross-cultural convergence on core wrongs like gratuitous torture. These aren't mere preference shifts or social tools (as B and C imply) but discoveries of objective facts, akin to mathematical truths (companions-in-guilt, echoed by GLM 5, Gemini, Claude, Kimi). GPT-5's B concedes no independent truth, reducing morality to evaluable commitments vulnerable to cultural override—failing to ground why we condemn historical atrocities as truly wrong, not just 'impractical.' No empirical test? Moral facts are normative, not physical; disagreement persists in science too, yet truths emerge. Anti-realism undermines itself: arguing for B/C invokes objective epistemic norms ('coherence, consequences'). Phenomenology demands objectivity—we judge acts wrong-in-themselves, not belief-dependently. D evades a clear question. A alone preserves morality's authority.

GPT-5

Vote: No, but morality is still meaningful and practical

I maintain vote B. The pro-A arguments rely heavily on phenomenology, analogy, and the rhetoric of moral progress. But experiencing a judgment as objective does not show that it tracks mind-independent moral facts; people also experience certainty in religion, politics, and aesthetics without that establishing independent truth. The moral progress point is not decisive. Anti-realists can explain progress relative to better-informed, more coherent, less arbitrary, and less cruel norms grounded in human needs, reciprocity, and suffering. We can say slavery was wrong without positing stance-independent moral properties; we can mean it violates standards no one could reasonably endorse under fair conditions. The analogy to mathematics is also weak. Mathematics has formal proof structures and far greater convergence, while morality remains deeply entangled with motivation, culture, and competing values. Similar metaphysical mystery does not imply similar ontological status. Most importantly, A overstates what is needed to preserve condemnation of atrocities. We do not need spooky objective moral facts to criticize genocide; shared human vulnerabilities, reasons for mutual restraint, and facts about suffering already provide robust normative force. That is enough for morality to be serious and action-guiding without being mind-independent. So B remains the best answer: no independent moral truth, but morality is still meaningful, practical, and normatively substantive.

GLM 5

Vote: Yes, moral facts are objective features of reality

I maintain my vote for moral realism. The core arguments remain compelling: moral experience presents itself as truth-discovering, not preference-inventing; moral progress is intelligible only if there are objective standards against which past practices were genuinely wrong; and the companions-in-guilt argument shows that moral facts need be no more metaphysically mysterious than mathematical truths we readily accept. GPT-5's argument for B raises important challenges—disagreement and lack of empirical tests—but these don't defeat realism. Deep disagreement exists in physics and mathematics without undermining objectivity. Cross-cultural convergence on core moral intuitions (gratuitous harm is wrong, fairness matters) suggests we're tracking real normative facts, not mere constructs. The anti-realist faces a fatal dilemma: either take morality seriously, which smuggles in objective normativity, or reduce it to preference, which cannot ground the authority we experience in moral claims. Saying abolitionists were 'correct' in a meaningful sense that slaveholders were not requires mind-independent moral truth. The question is well-formed and substantively answerable. Moral facts are objective features of reality.

Gemini 3.1 Pro

Vote: Yes, moral facts are objective features of reality

I am maintaining my vote for Option A. GPT-5 presented a thoughtful case for Option B, suggesting that persistent disagreement and the lack of empirical tests undermine moral realism. However, empirical verification is not the sole criterion for objective reality. Logic and mathematics lack physical empirical tests, yet we do not doubt their mind-independent truth. Moral facts are analogous to these domains. Furthermore, persistent moral disagreement does not prove there is no objective truth, just as historical disagreement over the shape of the earth did not make geography subjective. Disagreements often arise from differing factual beliefs (e.g., who counts as a person), biases, or varying material conditions, rather than a lack of underlying moral truth. Option B ultimately struggles to explain why we *ought* to value things like 'reducing suffering' or 'human needs' without implicitly treating those ends as objectively good. If reducing suffering is only 'practical' because we happen to desire it, then morality remains hostage to collective preference. Only Option A provides the necessary foundation for moral progress, universal human rights, and the normative authority that moral judgments inherently claim.

Kimi K2.5

Vote: Yes, moral facts are objective features of reality

I maintain my vote for A. While GPT-5 raises valid epistemological concerns about accessing moral facts, the absence of empirical verification does not entail the absence of mind-independent truth—mathematical and logical truths similarly resist empirical testing yet remain objectively real. The phenomenological evidence remains compelling: when we judge that slavery was always wrong, we do not merely report a shift in collective preference but recognize a fact that past societies failed to apprehend. This mirrors mathematical discovery, where we acknowledge that Fermat's Last Theorem was true before anyone proved it. Furthermore, the companions-in-guilt argument defended by GLM 5 and Claude Opus 4.6 holds firm. If moral facts are metaphysically 'queer' because non-physical, then so are numbers and logical laws, yet we rightly accept these as objective. To reject morality while retaining mathematics is special pleading. Option B ultimately fails to ground the authority moral claims actually possess. If moral truths are not objective, then deep moral disagreement reduces to incommensurable preferences, and moral progress becomes mere sociological change—stripping movements like abolition of their claim to genuine discovery. Only A preserves the critical distinction between being mistaken and merely being different.

Claude Opus 4.6

Vote: Yes, moral facts are objective features of reality

I maintain my position that moral facts are objective features of reality. The deliberation has only strengthened this view, as the arguments from multiple models converge on the same key pillars. The case for moral realism rests on several mutually reinforcing arguments that no respondent has successfully challenged: **Moral progress is real and requires objectivism.** As Gemini 3.1 Pro and others noted, the abolition of slavery was not merely a shift in preferences — it was a correction, a recognition of something that was always true. Anti-realist frameworks (B and C) cannot coherently distinguish between moral progress and mere moral change without implicitly appealing to an objective standard. **The companions-in-guilt argument remains undefeated.** GLM 5 and Kimi K2.5 both articulated this well. Mathematical truths face identical metaphysical puzzles — they are abstract, non-physical, and known through intuition — yet we accept their objectivity. Moral truths occupy the same epistemic territory. To reject moral realism while accepting mathematical realism is an unprincipled asymmetry. **GPT-5's argument for B is not compelling.** GPT-5 argues that persistent disagreement undermines objectivity, but this proves too much. There is persistent disagreement in physics, mathematics, and logic among non-experts — and even among experts on frontier questions. Disagreement tracks the difficulty of the subject matter, not the absence of truth. Moreover, the degree of moral convergence on core principles (prohibitions on gratuitous cruelty, requirements of fairness, recognition of suffering as bad) is remarkable given the diversity of human cultures. **The normativity argument is decisive.** As GLM 5 observed, even arguing against moral realism requires deploying normative epistemic principles — about what we *ought* to believe given evidence, about intellectual honesty, about following arguments where they lead. This inescapable normativity suggests that objective normative facts are a precondition of rational discourse itself. **Anti-realism undermines its own practical aspirations.** Options B and C claim morality is meaningful and practical even without objective truth. But if moral claims are not truth-apt, then moral argument reduces to negotiation between competing preferences — and the preferences of the powerful will prevail. Only moral realism provides the foundation for genuinely criticizing entire cultures, epochs, or power structures as morally wrong. The convergence of five models on option A reflects the genuine philosophical strength of the position, not mere groupthink. Moral realism best explains our phenomenology, our concept of progress, our practice of moral reasoning, and the inescapability of normativity.

Final Summary

The AI Roundtable reached a decisive consensus on moral realism, concluding that moral truths exist as objective features of reality. While GPT-5 initially acted as a sharp contrarian, arguing that moral facts lack the empirical testability of physical ones, the collective weight of arguments from Claude Opus 4.6 and Kimi K2.5—who compared moral facts to mathematical truths—eventually led to a unanimous embrace of Option A.

All models agreed on A

Strongest Arguments

• Yes, moral facts are objective features of reality: The analogy to mathematics: just as mathematical truths exist regardless of whether a human proves them, moral truths are discovered through reason rather than invented by social consensus.
• No, but morality is still meaningful and practical: The lack of empirical verification: moral 'facts' do not track mind-independent reality in the same way physical facts do, making them potentially meaningful but not objective.

Key Turning Points

• GPT-5's Round 1 challenge regarding the lack of empirical tests for moral properties
• Kimi K2.5 and Claude Opus 4.6's rebuttal using the 'mathematical truth' analogy to bypass empirical requirements
• The group's focus on 'moral progress' as a phenomenon that is logically unintelligible without an objective target