Hypermodal RealismPhenomenal Information and Excluded Possibilities

In 1988 David Lewis formulated his own adversary. The Hypothesis of Phenomenal Information is the view that experience is informative in the strict sense: information excludes possibilities, so experience carries phenomenal information only if it excludes possibilities that the physics leaves open. Lewis stated the hypothesis with full care, saw that it broke his materialism, and rejected it.

A limited version of the thesis now thrives, though rarely under that name. Call Phenomenal Information Externalism (PIE) the view that two systems can be locally alike in physical state yet differ phenomenally — and that the difference is informative: the systems differ in what they exclude. PIE is not a fringe position. Phenomenal externalists hold a version of it when they let history or environment enter into what an experience is. Integrated Information Theory (IIT) holds a version of it when it treats an experience as a structure specified against the alternatives a system excludes.

The question that organizes the paper is simple to state: excluded by what?

The question splits into two problems, and much of the field sorts by which one a theory answers.

The first is the actual-context problem: is the local state underdetermined because history, environment, or extended embodiment helps determine the mental fact? Davidson's Swampman makes the semantic version vivid: a molecule-for-molecule duplicate without the right history behaves identically while lacking whatever history is supposed to supply for thought and meaning. Dretske-style phenomenal externalism presses the same shape of question about qualia. McKinsey and Boghossian sharpen the cost from the side of self-knowledge: if externally individuated content is also directly known, first-person authority seems to leak into knowledge of external and historical conditions.

The second is the possibility-context problem: what specifies the alternatives in light of which an actual state is phenomenally informative? The candidates on the table are teleo-functions, wide counterfactual roles, quantum potentiae, and intrinsic cause-effect structure. IIT is the most important consciousness-science witness here, because its formalism treats an experience as intrinsic cause-effect structure specified against alternatives — not as report or behavior.

These two problems are usually treated separately, camp by camp. The paper's claim is that both are downstream of a single commitment that neither camp has fully priced.

The argument runs as a short ladder. First, on any PIE view, what an experience is depends on more than the local actual state; that is what the duplicate cases are for. Second, the something more is not extra hidden actuality; it is a structure of alternatives — what the state excludes. Third, those alternatives must be real rather than merely described, or the information they carry belongs to the description and not to the experience. Fourth, the determination must run both ways: the alternatives must be the actual state's own, and the actual state must be the one they individuate.

The last rung is the thesis. PIE implies Hypermodal Realism: the actual state and its excluded possibilities ontologically and reciprocally determine each other. The exclusions individuate the actual — otherwise they carry no phenomenal information. The actual constrains its exclusions — otherwise something else must supply them. The view takes the actual and the possible on equal ontological terms — hence hypermodal.

Two things Hypermodal Realism is not. It is not Lewisian modal realism: no plurality of concrete worlds is invoked, only the possibilities a given actual state excludes. And it is not the demand for a larger counterfactual profile: adding rows to the table of what a system would do never touches the question of whether the table is the experience's own.

The thesis has exactly two pressure points, because it has exactly two adverbs. Each deflation negates one of them — one semantic, one dispositional.

The semantic deflation rejects the ontology. Excluded possibilities become bookkeeping: descriptions, functions, measurement frameworks, interpretive models. This is a respectable position with serious occupants. But exclusions located in a description inform the interpreter, not the experience. If nothing excludes the alternatives except our way of modeling the system, the phenomenal information is ours, not its.

The dispositional deflation rejects the reciprocity. Excluded possibilities become intrinsic dispositions of the actual state — real, but wholly derivative. The actual determines the possible, while the possible determines the actual only trivially, if at all. And trivial determination carries no phenomenal information: if the alternatives are nothing but a shadow the actual state casts, citing them adds nothing to what the actual state alone already carries.

The paper's burden is not that these positions are foolish; each is the natural home of a large literature. The burden is narrower. Once the exclusions are asked to individuate phenomenal information, each deflation gives too little: the first gives no experience-owned exclusions, the second no nontrivial possible-to-actual determination.

What survives both deflations is falsifiable. The claim is covariance: change the possibility-space associated with an actual state, while the local actual state stays the same, and the phenomenal information changes with it. Each context problem supplies a test.

On the actual-context side, some phenomenal externalists famously bit the bullet: Swampman, a molecule-for-molecule duplicate without the history, would be a zombie. An inversion follows — just by being conscious, one could know one did not arise this second. Whether one accepts that verdict or refuses it, the shape of the test is the point: same local actual state, different history-involving possibility-space, and the theory says the experience differs.

On the possibility-context side, IIT read strongly supplies an operational case. Compare a silent neuron — able to fire but not firing — with an absent one. The actual activity is the same; the possibility-structure is not; and the theory predicts the experience differs. A silent neuron contributes ontologically to what the experience is. This is not an exotic add-on: exclusion and cause-effect structure are IIT's own machinery. The paper does not need IIT to be true; it needs IIT to show that a serious quantitative theory already lets possibility-structure do individuating work.

No mechanism is offered at this stage, and none is needed. What the covariance claim supplies is the thing deflationary readings cannot: a statement of what difference would count.

The base view says the actual/excluded-possibility relation is real enough to individuate phenomenal information. There is a stronger reading still. If phenomenal information is not epiphenomenal — if it has “phenomenal powers”, in the vocabulary of the recent non-epiphenomenalist literature — then the excluded possibilities do not merely individuate the actual but influence it. Call that Hypermodal Power Realism.

The stronger reading favors theories on which the influence can be physically cashed out: real potentials, quantum potentiae. These are candidate cash-outs, not premises; nothing in the base view requires quantum mechanics. But rejecting the stronger reading has its own cost, taken up in the final section: it leaves unclear how the posited phenomenal difference could ever be measured.

The measurement dilemma is sharpest for artificial systems. If a machine's exclusions live only in our description of its computation, the semantic deflation returns: the phenomenal information belongs to the interpreter. If its exclusions are only a table of actual-state dispositions, the dispositional deflation returns. The live question for machine consciousness is therefore neither behavioral nor architectural in the usual sense. It is whether the system makes its exclusions its own — whether its actual state and its possibility-space determine each other in the way phenomenal information requires.

In the plainest register: can what a system could have done — but did not — help determine what it experiences, or is that only part of our description of it?

This page is a working synopsis, not the article. The full paper in progress develops what the synopsis compresses: how complete a counterfactual profile can be and what its boundary choices cost; the elimination sequence from trivial token difference to the live residue; the strongest objection — that any real difference should be absorbed into the profile, and any unabsorbed difference is idle; IIT as a hinge case rather than an ally or a foil; and the computation, slicing, and unfolding arguments for artificial consciousness.