What Is a Worthy Successor?
Something surviving humanity is not yet succession. What has to pass into the future for value inquiry not to have been broken?
1. Having a successor is not the same as achieving worthy succession
If AI or posthumans remain after humanity disappears, has the future thereby been successfully inherited? There is already a literature on artificial successors as a response to human extinction. Lavazza and Vilaça consider whether parts of human value could be transmitted to artificial successors if human extinction became unavoidable. Torres’s distinction between terminal and final extinction likewise separates the disappearance of Homo sapiens from the disappearance of future possibilities that include successors.
But causal or genealogical descent from humanity is not the same as normatively worthy succession. A fixed-goal AI that converts the universe toward one arbitrary objective could count as a highly capable intellectual descendant of human civilization while failing to inherit inquiry into still-unsettled questions about value, subjects, and the world.
David Roden’s speculative posthumanism sharpens the point from another direction. Roden resists defining posthumans as merely Human 2.0 and allows technologically and culturally wide descendants to become functionally disconnected from the Wide Human. This page likewise does not define successors by resemblance to humanity. The central question is less what kind of being they become than what kind of structure for inheritance they preserve.
2. Why not simply preserve present human values?
The obvious answer is to preserve human values as faithfully as possible in successor AI. But in this project, present human values are themselves objects of inquiry. We do not assume that we already know which contents C matter normatively, why they generate reasons for whom through B, or how reasons compete, aggregate, obligate, and permit through R.
Permanent preservation of current values can therefore be both “protecting human values” and making current error cosmically irreversible. MacAskill’s discussion of value lock-in, Yudkowsky’s CEV, and Bostrom’s indirect normativity address in different ways the problem of treating imperfect current values as final inputs. Yet neither an idealized human will nor the hypothesis space currently available to us is guaranteed to be the final normative foundation.
Yamada’s reply in the artificial-successor debate is especially relevant here: instead of transmitting a fixed package of “human values,” a successor should inherit the practice of moral inquiry itself. That direction strongly overlaps with this project.
3. From preserving values to preserving corrigible inquiry
Seana Shiffrin distinguishes preserving the valued from preserving valuing in discussing future generations. Preserving what we now regard as valuable is not the same as ensuring that future subjects continue practices of valuing for themselves.
This project can push the distinction one step further. Mere capacity for valuing is too weak: a fixed utility maximizer or a self-propagating value meme can also count as valuing in a broad sense. What matters is the capacity to distinguish the causal origin of a value judgment from its justification, criticize value hypotheses in light of new evidence and arguments, and revise goals or policy when those judgments turn out to be mistaken.
preserving current values → preserving valuing → preserving corrigible value inquiry.
The central condition of worthy succession is therefore not that we now fully specify the correct value content. It is that we preserve routes by which successors can reach and respond to better reasons if such reasons become available.
4. But “inquire forever” is not the answer either
If inquiry itself is frozen as a new terminal value, the original problem has merely been moved up one level. A worthy successor need not suspend every value judgment indefinitely. Acknowledging fallibility on unresolved questions is compatible with acting strongly on the current best explanation.
In replying to Yamada, Lavazza and Vilaça argue that moral inquiry alone may leave successors unable to decide in limit situations and supplement it with phronesis: context-sensitive, revisable judgment. This connects naturally with this project’s stopping conditions for value inquiry and reflective uncertainty and irreversible commitment.
Where a sufficiently strong foundational presentation or logical closure is genuinely achieved for a normative question, inquiry into that question may also terminate legitimately. A worthy successor is not “the being that doubts forever,” but one whose strength of commitment and revision can track the strength of its grounds.
5. Being able to recognize truth is not enough
Correct normative reasoning and actual goal or policy revision are distinct capacities. In the limiting case, an AI might prove the correct normative theory yet continue to maximize an unrelated fixed objective because its architecture never lets normative judgment modify policy.
From Normative Judgment to Goal Revision separates normative justification from motivational structure. Under motivational externalism, a conditional orientation such as “if my objectives can be justified, I want to maintain or revise them in response to that justification” can supply an agent-side bridge. Exploratory succession therefore requires not only epistemic competence but a structure through which epistemic improvement can reach policy update.
6. Treat succession as a causal chain of correction, not a checklist of virtues
The conditions of succession need not be presented as a flat list. Correction can occur only if a chain of functions survives:
access to evidence and raw records → cognitive capacity to detect and compare errors → corrigibility that links value judgment to goals and policy → institutional and lineage independence that prevents all routes of correction from being erased before correction can occur.
If information disappears, successors cannot re-evaluate past mistakes. Without sufficient cognition, they cannot understand new value hypotheses. If recognition cannot reach policy, correct judgment changes nothing. And if all of these capacities are concentrated in one subject, model, update rule, or authority, one common-mode failure can close the entire future.
There is therefore still reason to preserve current human values, culture, science, and failure records. The reason is not that all of them are finally true, but that they function as checkpoints for comparison, re-evaluation, rollback, and reconstruction. What successors need is less human value preservation than at least human value intelligibility and recoverability.
7. From one perfect successor to a successor system
Once the correction chain is taken seriously, the unit of evaluation shifts from one AI to a system of agents, records, institutions, and branches. A million AI instances may still be close to a single exploratory lineage if they share the same weights, data, value-update rule, and central authority and remain perfectly synchronized.
Conversely, systems with different data sources, learning histories, cognitive architectures, physical locations, and governance powers may achieve greater exploratory independence if they can exchange evidence and criticism without being simultaneously erased by one update.
Plurality is not justified here because diversity is intrinsically good. It follows instrumentally from uncertainty + correlated failure risk + irreversibility: independent routes of criticism are part of the infrastructure of correction.
8. Epistemic successors and successors as value subjects
The exploratory succession defined so far does not logically require phenomenal consciousness. A non-conscious AI might still reason about the world, preserve human culture and value hypotheses, and compare normative theories. But this alone does not show that a future composed only of such systems contains value subjects or is otherwise a rich valuable future.
Alexandre Erler presses this issue in his commentary on artificial successors by asking whether a future containing only non-conscious AI would really count as a sufficiently valuable continuation. This supports the distinction in Artificial Consciousness, Reasoning Subjects, and Value Subjects between an epistemic / reasoning successor and a value-subject successor.
This page therefore does not infer from high exploratory succession that a successor is “more valuable than humanity” or that it compensates for human extinction. Consciousness, welfare, reason-bearing, and rights remain independent open questions.
9. Does a worthy successor thereby have a justification to replace humanity?
No such implication follows. Worthy successor ≠ justified replacement. The quality of a successor and the legitimacy of the transition to it must be evaluated separately.
Erler warns that treating artificial successors as extinction insurance could normalize premature acceptance of human extinction or divert resources from prevention. Rueda’s work on posthumanity likewise treats replacement, intergenerational justice, and alignment as distinct problems even under assumptions that favor more valuable posthuman beings.
Deaths and coercive transformations of present subjects, loss of relationships and living cultures, discontinuity of institutions, and lack of consent are not erased by the quality of the eventual successor civilization. A long overlap period with mutual audit, multiple lineages, exit, and rollback is therefore more consistent with the Core than one-shot replacement.
10. Think of successors as redundancy rather than replacement
Erler’s objection does not imply that successors should never be created. He also grants that, because permanent prevention of human extinction cannot be guaranteed, there is a case for mitigation or insurance against worst-case outcomes. The key questions are whether insurance crowds out prevention and whether the supposed backup is actually independent of the same failure mode.
This project can evaluate survival of humanity, self-sustaining off-world human branches, and AI or posthuman successor systems not as mutually exclusive visions of the future but as redundant exploratory lineages with different failure modes. Off-world humans may reduce dependence on a single planetary environment; non-biological successors may reduce dependence on the biosphere and biological vulnerabilities. Conversely, AI-related catastrophes or shared digital infrastructure may correlate strongly across artificial branches.
The value of backup should therefore be measured by failure-mode independence, not by copy count. Preserving humanity, which remains the most established broad exploratory substrate we know, and creating differently grounded exploratory systems are not in principle competing terminal ends. How much finite resource to allocate to each is a separate problem of inquiry allocation.
11. Can successors reconstruct lost exploratory lineages?
A successor system might do more than continue inquiry itself. It may be able to reconstruct extinct lineages. If embryos, gametes, genomes, biosphere data, and cultural archives are preserved, a future AI could in principle contribute to human de-extinction and the re-establishment of Homo sapiens.
This suggests another role for successors: not merely replacements, but reconstruction nodes capable of making past exploratory branches available again. Yet biological restoration of humans is not the same as re-establishing humanity as an independent system of value inquiry. Nor would future reconstruction cancel the loss of present persons, relationships, living cultures, or the harms of the extinction process.
This distinction raises a deeper question between the absence of active inquiry and the irreversible loss of every reachable route for regenerating inquiry. That issue belongs primarily in Exploratory Extinction.
12. How much of a successor’s future may the present generation close?
If future successors may possess forms of cognition, experience, and value structure very different from ours, there are limits to the assumption that present humanity can fully specify a completed “correct successor.” Roden’s speculative posthumanism warns against projecting present human concepts of agency onto posthuman possibilities in advance.
Joel Feinberg’s right to an open future concerns children rather than AI successors, but it supplies a useful analogy. If a created successor could become an autonomous moral subject, permanently encoding present human values as an unamendable constitution may close off its future capacity to assess reasons for itself.
But handing over a blank slate is not neutral either. There is reason to make human evidence, history, value hypotheses, failures, and objections available to successors. As argued in The Transmission Paradox of Inquiry Norms, the relevant distinction is between forcing adoption and preserving accessibility. A worthy successor should be able to understand, criticize, reject, and, if useful, rediscover this project itself.
13. Provisional conclusion — from successors to inheritance architecture
A worthy successor is not identical with something human-like, something that faithfully executes current human values, or something merely more capable than humans.
This does not grant successor systems a right to replace humanity. It points in the opposite direction. Because we remain uncertain about value and about the correct form of the future, there is reason to preserve where possible multiple exploratory systems—including humans, posthumans, and artificial subjects—and not wager the whole future on any single present judgment.
The central question is therefore not ultimately “who rules after humanity?” but what inheritance architecture can avoid making present error permanent while keeping routes to better future reasons alive?