Why Existing Values Are Not a Final Foundation
1. The target is not existing value, but its unconditional privilege
This thesis does not ask us to discard human morality, responses to pain, demands for freedom, dignity or fairness, social trust, attachments, or existing ethical theories. Present value systems contain enormous amounts of information about what matters to subjects, which institutions sustain cooperation, and which experiences exhibit value-like properties.
The problem arises when we confuse using them as evidence, preferences, and institutional rules with treating them as cosmic or subject-independent final value.
2. Causal origin and normative reason are different
Why humans hold a given value can be partly explained by evolution, neural structure, embodiment, upbringing, language, culture, institutions, power, and historical contingency. These answer "why do we experience X as valuable?" They do not by themselves answer "why does X give us a reason independently of our having been caused to value it?"
This is not to use genealogy as a truth-destroying machine. A belief does not become false merely because it has an evolutionary or cultural origin. The weaker conclusion is simply that an explanation of origin does not complete a justification.
3. Reflective endorsement can be strong evidence without being a final guarantee
An important reply says that we should keep only values we still endorse after becoming informed and reflective, rather than relying on raw intuition. Reflective equilibrium, coherence, perspective-taking, and empirical knowledge can indeed correct many crude value judgments.
But the inferential rules, attention structures, emotions, and models of agency used in reflection are themselves part of our present cognitive architecture. Reflective endorsement can greatly increase epistemic weight without necessarily guaranteeing final normativity beyond the current human form of cognition.
4. Even if moral realism is true, present human values do not automatically become true
This point is especially important. Accepting moral realism is not a refutation of the Core. Realism says, at minimum, that some moral or normative facts obtain in a way not reducible to the mere attitudes of a subject. It does not decide which first-order theory is correct.
Divine command, Kantian duty, consequentialism, virtue, contract, reasons, natural value, and non-natural normative facts can all be developed in realist directions. So even if realism is plausible—or widely accepted—it does not tell us which current conception of rights, suffering reduction, respect for persons, flourishing, or freedom should be irreversibly fixed.
Indeed, if realism is true, it remains logically possible that current human values are substantially misaligned with true value. Realism can therefore support inquiry: there may actually be something further to discover about what is correct.
5. Why major existing foundations are not treated as final foundations
Theological foundations
If revelation or divine command is true, it could provide a powerful normative basis. But the epistemic grounds for privileging a particular supernatural source or revelation over competing world-explanations are not presently decisive. This does not assign theology probability zero; it leaves it among the hypotheses to be explored.
Deontological and person-based foundations
Principles requiring respect for reason, autonomy, or persons have enormous institutional appeal. Yet a bridge is still needed from the descriptive or functional property "is a rational being" to the normative authority "therefore must be treated as an end in itself."
Utilitarian and pleasure/pain foundations
Pleasure and pain deserve special attention. The negative phenomenology of suffering may be one of the strongest candidates for a basis of real value. But moving from "pain is aversive for that subject" to "the pain of all subjects ought to be aggregated and minimized in a subject-independent way" requires further principles of symmetry, aggregation, interpersonal comparison, and distribution.
Social contract and public ideals
Rights, democracy, the rule of law, trust, and freedom can be strongly defended as institutions supporting current shared inquiry. That defense can stand even without treating them as final metaphysical values: they sustain cooperation, criticizability, decentralization, and corrigibility.
6. Value disagreement does not refute realism, but it is evidence against lock-in
Disagreement across cultures, eras, and subjects does not by itself show that moral realism is false; science also contains long periods of disagreement about facts. But if current values have varied greatly with historical conditions and the same subjects revise them in response to information or shifts in perspective, that should reduce our epistemic confidence in permanently freezing the present snapshot.
7. Phenomenal value is a foundational candidate, but not yet a stopping condition
It would also be premature to classify all existing value as social construction. The aversiveness of pain, desire satisfaction, or a sense of meaning may contain value-like properties in appearance itself. Epistemically, these sit closer to the lower layers than deontology or social contract theory.
But appearing as pleasant or unpleasant and appearing with objective normative force are not the same. The fact that pain is bad for the subject does not directly settle universalization across subjects, aggregation, or distribution.
Phenomenal value may therefore be among our most important clues to objective value, but in its current form it does not justify irreversible closure of inquiry. That assessment could change if future cognition disclosed normative force alongside phenomenal value with strength comparable to the epistemic minimum.
8. Existing values can be used strongly as current best practice
Lack of final foundation does not mean weak practical use. Strongly supported value judgments can receive strong commitment. If we have ample evidence that massive suffering, arbitrary violence, deception, destruction of trust, and uncriticizable concentrations of power damage existing subjects and collective inquiry, then rights, freedoms, and anti-violence norms can be adopted as powerful institutions.
The relevant distinction is between practical adoption and metaphysical lock-in. The burden of proof for "we should follow this value for now" is not the same as that for "all future subjects should be physically prevented from ever reconsidering this value."
9. Practical commitment and irreversible lock-in are separate
As confidence in a theory of value rises through inference, resource allocation, institutional design, and AI behavior can rationally move strongly toward it. But while that confidence remains at the inferential layer, there is no need to erase every minority view, record, foundational research program, or route by which another kind of subject could revisit the issue.
This distinction means that refusing to treat existing value as a final foundation does not entail ethical anarchy or permanent suspension of judgment. We use our best current values while protecting the mechanisms that could reveal their errors.
Ignoring the preferences and experiences of present subjects would itself destroy exploratory data. Present values can be preserved, observed, and negotiated not as "sacred final answers" but as real evaluative structures instantiated in existing subjects.
10. Implications for AI alignment
For near- and medium-term accident prevention and cooperation, it can be justified to align AI behavior to a substantial degree with current human requirements and institutions. Permanently fixing their current form as terminal and irreversible value is a separate claim. If human values may be partly mistaken, mutually inconsistent, or revisable by future knowledge and subjects, long-run design should distinguish safe provisional alignment from permanent value lock-in.
11. What would weaken this thesis?
- Strong evidence that one specific present value system uniquely coincides with subject-independent normative facts.
- A non-circular foundation of one first-order ethical theory that excludes its rivals.
- A demonstration that reflective endorsement converges necessarily on the same values independently of cognitive architecture and historical conditions.
- A demonstration that fixing current values consistently increases corrigibility more than preserving explorability does.