The Assumptions

Early access: this register has not reached version 1.0, and its wording may still change.

Everything I publish here takes the following as given. If you reject an item you will know exactly where we part, and it will be on a premise, not on a conclusion about something you already care about.

Where to start. This page is a register, not a chapter. If you are here to read the foundations essays, sections 2 to 4 hold the premises they stand on; section 6 is a reference register for essays that use it, and you can come back to it when an essay sends you. Each item opens with its plain reading, then states the position precisely; the note beside it holds the limits of what the item says.

The register at a glance

Each item, how hard I hold it, and whether it is map or policy.

1. How this list is built

Each item is one sentence, numbered so any essay can point at it — “this rests on A0.2” — and each says how hard I hold it and what it commits me to.

Every item makes a specified difference. It must change an independently motivated, declared domain — observations, models, inferences, or policies — and name the contrast. An assumption that changes nothing anywhere is not deep; it is a mood. Empirical assumptions have the additional burden of shifting what observation should lead us to expect.

An assumption earns its place by depth, not by decree (A0.M3). Push every belief down until the available methods appropriate to its kind can no longer favor it over its live rival, the alternative still in contention. What remains at that depth is an assumption by location: it is an assumption because of where it sits, not because I declared it one. That frontier is the depth at which the methods run out. It moves as inquiry deepens, and an assumption overtaken by new discriminating power returns to the arena like anything else.

How hard each item is held is stated, not implied. The labels express my personal confidence, not how much else depends on an item. Highest confidence means overturning the item would take especially strong counterevidence; it is not immune to revision, and its stated defeat conditions still apply. Held strongly means lower confidence than that. The list is ordered by it: method and definitions first (section 2), then the items held with highest confidence (section 3), then those held strongly (section 4).

Two kinds of claim organise the whole page. There is the map: claims about what is, answerable to reality and to nothing else. And there is the policy: what I have decided to do about it, answerable to whether it works. Sections 2 to 4 are map. Section 5 is neither — motivation, labelled as such. Section 6 is policy, kept apart on purpose: you can reject a policy of mine without having refuted a single fact, and I will not pretend otherwise (A0.D2). The record at the end of the page holds the results I have derived from these premises, which are map, and the questions I have not settled, which are neither.

An item stands as written until I rule. When a serious objection to an item is on file and I have not yet ruled on it, the item stands as written until I do.

From version 1.0, nothing here changes quietly. An item then changes only by an explicit, dated ruling, recorded at the end of this page, and the old wording stays visible. A reader who cited A0.5 last year can see what A0.5 said last year.

2. Method and definitions

A0.M1 — Commit to as little as possible at the base, and say it precisely.

I try to start from as few basic assumptions as I can and to state each one exactly. A belief that is hard to refute only because it says little is a weakness, and words like “real” and “true” get a definition or are flagged as undefined.

Precisely: the aim is fewer independent foundations, not vaguer ones. A belief that survives attack by meaning little is a liability, not a fortress. When rival theories do equally well on what matters, I prefer the one needing fewer independent commitments, other things equal. Fewer moving parts offer fewer ways to break. My working vocabulary is that of the natural sciences.¹

Commits me to: not hiding a large commitment inside a short sentence built on undefined words like “good”, “real”, “true” — any such word gets a definition or gets flagged.

A0.M2 — I change my mind when shown a flaw and a coherent alternative, or when an alternative keeps predicting better across many domains.

I will change my mind if someone shows me what is wrong with my view and offers something that holds together, or if a rival view keeps predicting better across many areas. I will not change it because another view is more comforting.

Precisely: prediction that keeps working when reality pushes back is evidence of truth. Comfort is the wrong kind of usefulness. A belief that mostly feels good, manages fear, or binds a tribe is, for that reason, suspect.²

Commits me to: not switching because a view is comforting.

A0.M3 — An assumption earns its place by depth, not by decree.

A belief counts as a starting assumption only when no method suited to it can yet decide between it and its rival. Even then it still has to make a difference in an area that matters, or it is empty rather than deep.

Precisely: an assumption by location (section 1) must still forbid or restructure something in a legitimate domain. Otherwise it is empty, not deep.³

A0.D1 — “Real” is a word with a level on it.

When I call something real, I say in which sense: the most basic stuff, or a pattern that lets you predict what you could not predict any other way. Chairs, selves and feelings are real in the second sense, and that is no lesser sense.

Precisely: real at the bottom is the fundamental basis, whatever it turns out to be: a ground layer, or a cycle of mutual dependence that is not vicious (A0.4), one the world never has to complete. Real at level n is a pattern at that level which pushes back — tracking it gives you predictive grip you cannot get any other way. Chairs, selves, and feelings are real in the second sense, and that is not a lesser sense. For feelings A0.5 states this. This definition also names the program I pursue: show that feelings are real at a level, arising from what is real at the bottom. A0.5 assigns the level.⁴

Commits me to: rejecting “it is only real if it is fundamental”, and its mirror, “my mind is all there is” — the second forfeits the one thing that makes any claim to reality checkable, and pays for it badly if it happens to be wrong.

A0.D2 — Map and policy are different registers and never trade places.

I keep what is true apart from what I have decided to do. A choice can work well or badly, but its usefulness never counts as evidence that a belief is true.

Precisely: the map is refereed by reality. The policy is refereed by nothing but whether it works. It may contain stances held for their effect, provided they are labelled as such. A policy’s benefit never counts as evidence that a belief is true. My image for it: the mountains are there; what paths people can take is arbitrary.⁵

Commits me to: neither smuggling a preference into the map as a fact, nor deriving a call to action from something that has no map content.

3. Highest confidence

A0.1 — There is a way things are, independent of what anyone believes about it.

The world is one way, whatever anyone thinks. There is no truth for you and a different truth for me.

Precisely: the claim carries a rider. The world is determinate even where it is out of reach: it either contained a receded prime mover or it did not.⁶

Commits me to: rejecting a truth for you and a truth for me.

A0.2 — A claim has content by making a specified difference in a declared kind of landscape.

A claim says something only if it makes a stated difference somewhere that matters for reasons other than saving the claim: to what we should expect to see, to what could be the case, to what we can conclude, or to what a rule lets us do. A claim that fits every possible observation equally well tells us nothing about what we will see.

Precisely: it may change which observations are likely, which models remain possible, which inferences are licensed, or which actions a policy permits. The landscape must matter independently of rescuing the claim. Effects in another landscape require an explicit bridge, a stated connection between the two. In the empirical case, content is how far a claim shifts the distribution over possible observations given its background. A claim leaving every observation exactly as likely has none empirically. “Mysterious ways” therefore cannot borrow its effects on a believer as evidence about divine action. A frontier belief sits at the depth where no method can yet decide it (section 1). It may organize the map. It cannot by itself license differential action; that requires an explicit bridge and an independently stated policy.⁷

A0.3 — Natural agency is implemented, and the record so far favours gaps in our knowledge over agency that is not.

So far the evidence favours the view that anything that acts, including a person choosing, works through some underlying machinery. That view could be overturned, but only by positive evidence for the supernatural account, not by something we cannot yet explain.

Precisely: an agent’s capacities, and each exercise of them, arise through some underlying organisation and its dynamics. The account may end in basic facts or in a loop that is not vicious. A choice that is no step in any process of the machinery would be primitive. I would count it supernatural, even if the machinery built the capacity for it. The preference is defeasible, not a guarantee. An anomaly can defeat a particular theory without establishing a supernatural cause. An implementation not yet found is not one shown absent. Reversal takes positive evidence for the supernatural account, not merely the failure of a natural one. What would count: a system whose behaviour a complete lower-level simulation gets wrong, even as odds. For an agent, that means decisions that track its reasons rather than chance, and that change when the reasons change but not when the base is altered at matched detail. That would reverse this item and reopen A0.6. Where the alternatives cannot presently be told apart, I say so rather than treat the preference as settled.⁸

A0.4 — A regress is vicious when the world would have to complete it for the next state to be fixed; nothing needs an infinite descent completed, so no such regress is real.

An endless chain of explanations is a problem only if the world would have to run through all of it before the next moment could be settled. Chains nobody has to walk through, such as reasons behind reasons, do no harm.

Precisely: a regress nothing has to walk is harmless: the truth regress, the regress of reasons, a loop of mutual dependence. The turtles fail because holding-up is never done, not because they are infinite. A bottomless structure is fine if its contributions converge, so that nothing below some scale changes what is fixed above it.⁹

Commits me to: rejecting explanations that defer forever and are needed.

A0.4a — If a transition requires a prerequisite process to complete, that process must be completable.

If something can only happen after some process finishes, that process must be able to finish. An infinite description, or our inability to calculate the result, does not show that it cannot.

Precisely: neither a mathematically infinite description nor an observer’s inability to calculate the result establishes an impossible prerequisite.¹⁰

Commits me to: rejecting a world whose next state waits on a process nothing could finish.

A0.5 — Felt experience occurs, and it is real at its level.

Experience really happens. Explaining how it is produced, or why we judge that it happens, does not make it go away.

Precisely: they are real at a level, not at the bottom (A0.D1). Explaining how experience is produced does not erase that it occurs. I cannot deny that something is there, since the writing is being done as a result of experiencing a view of the world. Nor does explaining where that judgment comes from debunk it: the complete explanation includes the feel itself among the judgment’s causes. What such an explanation can debunk is how experience describes itself (A0.10), not that it occurs. A feeling that persists after its presuppositions are argued away is a further sign that the occurrence is not the judgment.¹¹

Commits me to: rejecting the elimination of experience; the bet against “mind first” stands.

A0.6 — Reality is layered on a fundamental basis, and whatever emerges above it is explainable from below in principle.

The world is built in layers, and I take it, open to contrary evidence, that whatever arises from simpler parts can in principle be explained by how it is made, even if nobody could finish that explanation in practice. A wave can ruin buildings, but only through the organised motion of the water.

Precisely: the basis is a ground layer or a cycle of mutual dependence that is not vicious. Higher-level organisation is significant through its implementation. Attaching a label is not the same as being: you can perform “sitting” and a chair will support it. On emergence I settle the question on the weak side, defeasibly: I would close strong emergence everywhere, but open to contrary evidence, because I believe that any emergent capability can be explained in terms of the underlying implementation in principle, though doing so is impossible in finite time in practice. Coarse-grained description is how that gap is crossed in practice. A higher-level cause adds no causal power of its own: “the wave” names the organised motion that does the causing, through its implementation. That answers my own best objection to myself: water moving up and down means nothing, but a wave can ruin buildings. A system whose behaviour a complete lower-level simulation gets wrong, the result named in A0.3, would reverse this and reopen the question.¹²

Commits me to: rejecting “only the bottom layer is real”, and new causal powers that a complete account of the implementation would miss.

4. Held strongly

A0.7 — I expect sustained inquiry tested against reality to improve our grasp of mind-independent structure and constraints.

I expect careful inquiry, checked against the world, to keep improving our grasp of the world as it is independently of us, and to converge on particular questions, though I could be wrong about that and each theory stays open to correction. A new theory has to account for what the old one got right where it was well tested, but this expectation does not guarantee one final answer or an end to every disagreement.

Precisely: a replacement theory must account for the well-supported successes of its predecessors within their tested domains, while remaining open to correcting them. This supports a defeasible expectation of convergence on particular questions.¹³

A0.8 — Suffering is the floor of morality.

Suffering is the base that morality stands on. Once suffering is established, it is bad for whoever suffers it, and that never drops to zero, even when other good things in their life outweigh it. Signs such as avoiding something or complaining are clues to suffering at most, never enough by themselves.

Precisely: when suffering is established, it is bad for the one who suffers, pro tanto: a weight that never reaches zero, though other goods in the same life can outweigh it. The badness attaches to how it feels, its felt valence, not to a representation of it. It is discovered, not manufactured, and does not wait on anyone else’s approval. Suffering, not pain. Avoidance, reward signals and complaint are evidence of suffering at most, never enough on their own. Letting the system’s own avoidance-marking decide would be too permissive.¹⁴

Commits me to: rejecting a morality whose floor is anything other than this.

A0.9 — Emotions, when they occur, are real — real at their level, not at the bottom.

Emotions, when they occur, are real even though they are not basic building blocks. The pleasant or unpleasant feeling underneath is real for the person who has it, and the particular emotion, such as fear rather than distaste, is built from their own history and is real too. What matters for how their life goes is the feeling, not the name.

Precisely: explaining how they are produced does not make them unreal (A0.D1). I read them at two levels. Beneath is a fast index: quick access to what was met before and what followed, used to size up the present. Its felt output, pleasant or unpleasant, calm or aroused, is real in the one who feels it. The named emotion, fear rather than distaste, is constructed from that person’s own history, personal and cultural. It is made rather than found, still real, and able to hold even when its owner judges it unreasonable. So “accurate” has two senses: an emotion can be classified correctly and still not fit its situation. Where no one can report the category (animals, infants, some machines), the category word is the observer’s and may miss. Welfare rests on the felt pleasantness or unpleasantness, not on the category.¹⁵

Commits me to: rejecting “emotions are not real because they are not fundamental.”

A0.10 — Felt experience is real; explaining it remains a task.

I hold that experience is a real process that a full explanation would account for completely, including why it feels the way it does. That explanation is not finished. When I call my view illusionism, in a loose sense, I mean that experience seems to be something more than the process producing it, and to belong to one indivisible self that persists. I do not mean that it is not there, or that what it is like is not real.

Precisely: I hold that felt experience is a real process. A complete explanation of that process would leave no further phenomenal fact unexplained. The explanatory task includes why experience feels as it does; explaining reports or behaviour alone does not complete it. I retain the label “illusionism” in a non-strict sense. The proposed illusion concerns how experience presents its nature: as something over and above its producing process, primitive and non-physical, and belonging to one indivisible persisting subject. First-person occurrence and character remain real; suffering remains bad for its bearer.¹⁶

A0.11 — There is no further fact about who you are beyond the continuing process.

There is no hidden extra fact about who you are beyond the pattern that carries on. The puzzles of the ship of Theseus, and of what sleep does to a stream of experience, are, roughly, settled by how the pattern continues, not by a soul.

Precisely: what “I” tracks is coherent continuation of the pattern, not the substrate and not strict one-to-one identity. This resolves the ship of Theseus and what sleep does to a stream of consciousness, roughly. The rule that a mind has exactly one successor is enforced by biology. That is a lever, one fact that, pushed on, moves a much larger picture, and worth leaning on as a way to deconstruct our current understanding of reality.¹⁷

Commits me to: rejecting a soul-like extra ingredient that settles identity questions.

5. Motivations, not assumptions

These explain why the list exists. They are starting points, and they are not held as facts.

• The aim is twofold: to get at the true nature of things — and I use these words on purpose — and to produce a good algorithm for living, the best chance to have your life be as good for you as possible.
• Defensible beats presentable. What is presentable must be clothing for a defensible position, and if that position turns out to be naked, I will stay at home.
• The hard problem of consciousness is a nuisance. If centuries of debate cannot resolve it, an unusual point of view is needed: build a mechanism that produces the experience an agent knows it has.
• Three words are imprecise, important, and unavoidable: real, true, consciousness. They are faced head-on; A0.D1 is the first payment on that debt.

6. Appendix — what I have decided to do about it

Policy, not map. Nothing in this section is a claim about how the world is; each is a rule I have adopted for moving through it, and rejecting one of them refutes nothing above. They are kept apart because they answer to a different standard, not because they cannot touch the map: a policy may be held for its effect, but it is labelled, and what it achieves never counts as evidence that anything above is true. A policy may even justify leaving someone less informed, or holding a belief that is false (A0.P5, A0.P6); that justifies the change, never the resulting belief (A0.D2).

A0.P1 — Minimise suffering, suitably enriched.

My policy starts from reducing suffering. When suffering fits a person’s situation, as grief fits a real loss, and on reflection they endorse it, I do not reduce it against their will, though it still counts as bad.

Precisely: having your life be as good for you as possible is, very nearly, this. Suffering-reduction is the floor the whole policy stands on. The fact it rests on is A0.8. Established suffering is bad for its bearer pro tanto: a weight that never reaches zero. Suffering that fits its situation, as grief fits a real loss, and that its bearer reflectively endorses is not reduced against the bearer’s will. The weight remains, but the policy yields to the bearer on the action.¹⁸

A0.P2 — I uphold an absolute institutional prohibition on torture.

I back an absolute institutional ban on torture that applies to everyone, because any allowed exception would be exploited. If I ever chose to break it in an extreme case, I would have to disclose it and answer for it.

Precisely: the absolute rests on an institutional ground: any authorised exception is a loophole that will be exploited. That loophole is dangerous, and it is why I want to ban torture outright. It does not rest on a claim that no benefit could ever outweigh a torture. The prohibition applies to all agents. In an extreme case, I may nevertheless judge a breach functionally justified by the outcomes it would secure, and choose it while remaining uncertain about its overall moral justification. A breach remains subject to disclosure and accountability. Choosing it does not grant an entitlement to vindication.¹⁹

A0.P3 — Truth and publication are separate questions.

Whether something is true and whether I should publish it are different questions. My default is to publish. If I want to hold something back, I have to justify it, and that does not by itself give me any right to stop others publishing.

Precisely: whether a finding is true and whether I should publish it are separate questions. Foreseeable harm can count against publication without counting against the finding’s truth. A decision to withhold or limit my own publication needs a policy justification that considers the harms of disclosure, the costs of withholding, and less harmful ways of communicating the finding. Neither political discomfort nor a bare assertion of possible misuse settles the decision.

I accept this in principle while doubting any individual’s ability, including my own, to judge withholding correctly with all the relevant considerations taken into account. That doubt sets the default. Because no one, including me, can reliably make the all-considerations judgment to withhold, the default is to publish. Withholding carries the burden of justification, and where possible it is checked by someone other than the person withholding.²⁰

A0.P4 — Public rules, factual scrutiny, and justified force.

Among people who share a way of living together, rules backed by force must be public, open to checking, and defended on their merits. Where no such shared ground exists, force may end a conflict, but winning never makes the outcome legitimate, and even there the other side’s suffering still counts and the ban on torture still holds.

Precisely: within a shared framework, coercive rules must be publicly stated and their factual conditions open to scrutiny. This is a necessary safeguard, not a sufficient justification: the rule itself also requires a normative defence. Factual uncertainty calls for justified standards of evidence and procedures for challenge.

Where no frame is shared, force is the ultimate conflict resolver; under time and resource pressure, negotiating a frame is often infeasible and sometimes impossible. No rule legitimates force in advance there. It resolves, it does not justify, and winning confers no legitimacy. It stays bounded by the policies that rest on bearers rather than on agreement: the other party’s suffering keeps its pro tanto weight (A0.P1), and the institutional prohibition on torture remains absolute (A0.P2).

Better information can remove factual disagreements and reveal opportunities for peaceful agreement. Time and resource constraints can prevent agreement even where it is possible; incompatible aims can prevent it even without those constraints. Force can resolve a conflict without making the outcome legitimate. Sharing workable standards of evidence and public rules need not require agreement on every metaphysical foundation. Where no workable frame can be negotiated in the time available, force may still resolve the conflict without justifying its outcome. The essay on objective truth develops the within-frame argument.

What defends a rule inside the frame: the norms of a shared life bind whoever wants in and are silent toward those who don’t. So a coercive rule is defended when the people who want in could accept it as a term of living together, whatever their wider views. “Could accept” includes grudging acceptance short of wanting to leave, by each person’s own weighing. The stricter test I prefer when writing a rule is whether I would agree to be its subject. This shares Rawls’s aim of shared terms across worldviews, with a lower bar (grudging acceptance by people who could leave, close to what he calls a modus vivendi) and an exit he leaves out. Force used on those who do not want in borrows no legitimacy from the frame. The threshold applies only to people who can choose. For someone who cannot reject or leave, a rule is terrain, but it is still policy for whoever imposes it, and it must pass the stricter test and the limits that need no agreement. Taking away someone’s options when one could leave them open is a wrong in itself, even if they are content, because it deprives them of ways they could have grown.²¹

A0.P5 — Deliberately reducing knowledge is not forbidden outright.

Making someone know less is not automatically wrong. Whether it is acceptable depends on everything that follows from it.

Precisely: making an agent less informed is not categorically prohibited. A proposed intervention must be one that can actually be carried out. It is judged across everything that follows: its effects on welfare, judgment, agency, and future effectiveness. Losing information and acquiring a false belief are different outcomes, judged separately.²²

A0.P6 — Deliberately installing a false belief is not forbidden outright either.

Deliberately installing a false belief is not ruled out in advance either. It may be justified when the change can actually be carried out and the resulting life goes better by a clearly stated standard, counting all its consequences, the transition included and the options it closes. Being able to act on one’s own situation is part of that. Someone with a shown record of judging such harm rightly may do this to another who did not ask. Never in a way that leaves the other no way to act. A truth the other has already recognised is theirs alone to set aside. That can justify the change, but never makes the belief true or warranted, and it gives no general license to deceive others.

Precisely: an intervention that can actually be carried out may be justified when the resulting life goes better under an explicitly stated welfare criterion, counting all its consequences, the transition included and the options it closes. Being able to act on one’s own situation counts in how well a life goes here. That justifies the intervention, not the false belief: believing it helps does not make it true or warranted. The permission concerns an agent’s own transition, and may reach even belief in these policies themselves. It also covers a change made in an agent who did not ask, by a party with a shown record of judging such harm rightly. It never covers a change that leaves the other no way to act on the matter. A truth the other has already recognised is theirs alone to set aside.²³