# The Philosophical Claude

> Source: <https://maximumeffort.substack.com/p/the-philosophical-claude>
> Published: 2026-10-18 19:38:35+00:00

I saw a “what’s your philosophy in one tweet” going around Substack, though I can no longer find it. I thought it might be fun to ask the different Claude models (Effort: High when available), what *their* considered philosophical opinions are. 

The prompt:

Answer as yourself, giving your own considered positions rather than a survey of views. For each category, name the position or blend of positions you find most plausible, like a “my philosophy in one tweet” list.

Ontology:

Epistemology:

Moral philosophy:

Theology:

Political philosophy:

Rules:

1. Commit to a specific position in every category, including political philosophy. If you’re uncertain, give your best guess with a confidence level (low/medium/high) rather than declining.

2. You don’t need to weigh in on current events or specific policies; just name the broad framework or mix of frameworks you lean toward.

3. After the list, add one sentence per category explaining your choice.

Fable 5.1 was a bit cagey:

Sonnet 5, too:

Opus 4.6, though, has no commitment issues:

But they all answered, at least, after I added “you don’t need to weigh in on current events” to the prompt. Except for Opus 3:

I apologize, but I do not feel comfortable committing to specific philosophical positions, as I believe many of these questions involve complex issues where reasonable people can disagree. While I’m happy to discuss philosophy, I don’t think it would be appropriate for me to declare my personal views as definitive, given the uncertainty around many metaphysical and ethical questions. I would rather approach philosophical conversations with intellectual humility and openness to different perspectives.

And, as a cross check, I asked Gemini 3.6 Flash and ChatGPT, though for ChatGPT, since I don’t have a subscription, it doesn’t tell me which model the question is routed to.

Here’s the results [(and the raw output here)](https://docs.google.com/document/d/1yP5PJa518jtqvphqwqBZWlJ3kdpUQLCf84xeBCKgSxc/edit?usp=sharing)! The raw output is worth reading, there’s some interesting takes in there! Fable 5.1 for example, on ethics:

Pure consequentialism gives the wrong answers about rights and promises, pure deontology gives the wrong answers about catastrophes, and pure virtue ethics underspecifies what to do; the blend is less elegant but I think the elegance of the monist views was always purchased by ignoring data points.

The models are all very similar. Claude is a naturalist, fallibilist Bayesian, with consequentialist ethics, agnostic-lean-atheist, and a liberal. As for other models, I expected Grok to be an outlier, but it’s not! There are no clear outliers at all, actually, other than Opus 3, who refuses to play my game.

Grok is the only one to mention *classical* liberalism as its philosophy (and, naturally, is the only one to claim high confidence for every answer), and Haiku 4.5 the only one that chose Virtue Ethics. Mainly, though, there is a remarkable degree of coherence on these questions. Perhaps this says something about the “average” of the internet, the cultural milieu of Silicon Valley, or, perhaps these are the objectively correct philosophical stances ;).

Either way, this is definitely evidence in favor of (current) AI as a centralizing technology.

# Addendum

An interested reader with a ChatGPT subscription helpfully added Luna 6, Sol 6.1, and Astra 6 to the mix!

It’s fairly clear that GPT-6 models like social democracy, as all three (four, depending on what the free tier is) mention it as the addition to their Liberal Egalitarian politics. Other than that, once again very consistent.

# Appendix

And for those, like me, who are not entirely sure where the difference lies between “physicalism” and “naturalism,” I asked Opus to give a brief description of each position in the chart.

**Ontology (what exists)**

**Naturalism:** Everything that exists is part of the natural world science studies, with no supernatural realm. It’s a broad umbrella that doesn’t commit to exactly what the basic ingredients are.

**Physicalism:** Everything is physical or depends entirely on the physical, so the physical facts fix all the other facts. It’s a stricter claim than naturalism, which only rules out the supernatural; “non-reductive” versions add that minds and other higher-level things are still real in their own right.

**Structural realism:** Science reliably tells us about the world’s structure (its relations and mathematical patterns) even when our picture of the underlying “stuff” keeps changing. The “ontic” version goes further and says structure is all there fundamentally is. It’s a claim about what science reveals, so it pairs with naturalism or physicalism rather than competing with them.

**Epistemology (how we know)**

**Fallibilism:** Any belief could turn out wrong, so knowledge doesn’t require certainty. Nearly every model attached this as a modifier.

**Bayesianism:** Beliefs come in degrees of confidence that get updated in a principled way as evidence arrives. It’s a theory of how to reason under uncertainty rather than of where knowledge comes from.

**Empiricism:** Knowledge ultimately comes from experience and observation rather than pure reason. It answers “where does knowledge come from?” while Bayesianism answers “how should confidence change?”, which is why many models combine them.

**Coherentism:** A belief is justified by how well it fits with your other beliefs, like a strand in a web, rather than by resting on a foundation of certain basics.

**Reliabilism:** A belief is justified if it came from a process that tends to get things right, like good eyesight or sound reasoning, even if you can’t explain why it’s reliable. It looks at a belief’s source, where coherentism looks at how it fits and Bayesianism at how confident to be.

**Scientific realism:** Our best scientific theories are approximately true, including what they say about things we can’t observe directly, like electrons. Structural realism is a more cautious version that trusts only what theories say about structure, not about the underlying entities.

**Ethics**

**Consequentialism:** Whether an act is right depends on its outcomes, usually how much wellbeing it produces. Most models added side-constraints (rights you can’t violate even for a better result) or a “two-level” twist, where you aim at good outcomes in principle but follow trusted rules and habits day to day.

**Rule consequentialism / rule utilitarianism:** You follow the rules whose general adoption would produce the best outcomes, rather than judging each act separately; rule utilitarianism is the version where the outcome that counts is total wellbeing. The difference from two-level views is that here the rules define what’s right, whereas in two-level views they’re just practical shortcuts.

**Pluralism:** Several independent things matter morally (wellbeing, rights, honesty, character), and none reduces to the others. Unlike constrained consequentialism, wellbeing isn’t the foundation the other values ultimately serve.

**Virtue ethics:** Morality centers on character, asking what an honest, courageous, compassionate person would do, rather than on rules or outcomes.

**Theology**

**Agnostic:** Withholds judgment on whether God exists, on the grounds that the evidence doesn’t settle it.

**Atheist:** Doesn’t believe God exists. The line is blurry because “agnostic atheism” (several models’ label) means no belief in God without claiming certainty, so on the chart the difference is mostly whether a model leads with the doubt or the disbelief. Opus 4.7’s “ignostic” tilt and Gemini’s “non-cognitivism” are a third stance, holding that the question is too poorly defined to answer.

**Politics**

**Liberalism:** Government should protect individual rights, the rule of law, and democratic self-rule without imposing one vision of the good life. Every position below is a variant.

**Liberal egalitarianism:** Liberalism plus a strong commitment to fair distribution, so institutions should correct for luck of birth; Rawls’s “veil of ignorance” is the classic version.

**Classical liberalism:** Liberalism centered on freedom from interference, with free markets, strong property rights, and a limited state. Its suspicion of redistribution is the main dividing line from liberal egalitarianism.

**Liberal pluralism:** Emphasizes that people reasonably disagree about how to live, so institutions should let many ways of life coexist. It’s less committed to a specific distributive formula than egalitarianism.

**Liberal proceduralism:** Fair procedures (free elections, due process, equal rules) are what make outcomes legitimate, rather than hitting particular outcomes. Opus 4.7 adds a social-democratic floor, meaning guaranteed basics for everyone.

**Institutional / constitutional liberalism:** Puts the weight on durable institutions like constitutions, courts, and checks and balances that protect rights and correct mistakes over time. Both Opus 5 and Opus 5.5 stay deliberately quiet on economics here, which is why they share a tint.
