3. Intrinsic Accessibility

From functional to intrinsic

Functional accessibility, as set out on page 2, is existential: it asks whether at least one medium-and-protocol path succeeds for a particular user in a particular context. That is a useful floor. It is also a floor that bolt-on assistive technology clears at high cost, by building specialist paths for specialist user-populations and accepting that the underlying interface stays the shape it always was.

Intrinsic accessibility asks the next question. Instead of requiring at least one successful path for a given user, it asks how many distinct user-and-context profiles the underlying interface admits successful paths for — without specialist hardware, without parallel workflows, without the user being carried through an experience designed for someone else.

An interface is intrinsically accessible to the extent that it admits successful negotiation across a wide range of user-and-context profiles through its ordinary medium-and-protocol paths. The definition is comparative; some interfaces are more intrinsically accessible than others; the property is measurable on the interface itself, independent of any particular user.

The pseudo-user formalism

The measurement requires a way to talk about the range of users an interface might serve without committing to a specific user. The construct that does this is the pseudo-user.

A pseudo-user is a synthetic user-profile constructed by combining values across a set of capacity dimensions — visual, sonic, haptic, cognitive, language, colour-vision, and so on. Each dimension admits a finite number of meaningful values for the purposes of accessibility analysis (the visual dimension distinguishes full sight, low vision, functional blindness; the sonic dimension distinguishes hearing, hard-of-hearing, deaf; the haptic dimension distinguishes precision, tremor, switch-only). The Cartesian product of these dimensions yields a finite set of pseudo-user profiles, each representing one combination of capacity values.

Notation:

• S_PU — the set of all pseudo-users.
• N_PU — the cardinality of S_PU; the number of pseudo-users in the set.
• N_IB — the intrinsic-accessibility breadth of an interface; the number of pseudo-users in S_PU for whom the interface admits a successful negotiation through its ordinary paths.

The intrinsic-accessibility breadth is what the definition measures. An interface with high N_IB serves many pseudo-users; an interface with low N_IB serves few. The ratio N_IB / N_PUgives the share of the pseudo-user space the interface admits, and is a useful single number for comparing interfaces — with the caveat that no single number is the whole story, and that the dimensions of S_PUthemselves carry assumptions.

The optimal pseudo-user set

The construction so far depends on the choice of S_PU. A different set of capacity dimensions, or a different granularity of values within each, yields a different N_PU and a different N_IB. That dependency would normally make cross-interface comparison meaningless — if every analyst chooses their own pseudo-user set, numbers from different analyses are not commensurable.

The claim that resolves this is that an optimalpseudo-user set exists, and is independent of any specific provider, transaction, or interface. The optimal set is the one that admits no meaningful refinement — every additional dimension or value either duplicates an existing distinction or introduces a distinction the accessibility-relevant difference engine cannot resolve. Two pseudo-users that differ only on a dimension that no interface treats differently are the same pseudo-user from the perspective of accessibility breadth; the optimal set is the quotient under that equivalence.

The argument for the existence of the optimal set is constructive in principle: start with the maximal dimension-set, identify which dimensions the accessibility analysers respond to, collapse equivalent values, and iterate to a fixed point. In practice the procedure terminates because the dimension vocabulary used by accessibility tooling is itself finite. The claim is open at the edges — any shift in what the field considers an accessibility-relevant capacity changes the optimal set — but within a fixed analytical scope, the set is unique up to the equivalence above.

That uniqueness is what allows N_IB to be compared across interfaces and across providers. Two interfaces evaluated against the same optimal S_PU yield commensurable breadth measures. The measure means something about the interface, not just about the pseudo-user set the analyst chose.

The four-model architecture

The pseudo-user formalism gives the measure. To make the measure operational, the interface and the user need to be modelled in a way that lets the capacity-and-requirement match be computed mechanically for any pseudo-user. Four related models do this work.

1. Capability Model. What propertiesexist for describing a user, organised by subject ontology — Visual, Sonic, Haptic, Cognitive, Language, ColourBlindness, TabularContent. Properties are typed; they group into Capability Templates; templates carry a precedence order so that meaningless questions are not asked (it makes no sense to ask about minimum readable font size if the user has no sight). The Capability Model exists independent of any specific user.
2. Capacity Model. The settings for a specific user (or pseudo-user, or group of users) in a specific context, populating the Capability Model. The crucial move here is the support for functionally-dependent settings: a setting can be defined as an action triggered by an external influence — a function of fatigue, of time of day, of observed user behaviour, of ambient light. Static profiles are a special case where the function returns a constant; dynamic profiles let the capacity recompute itself in response to runtime conditions. That is what turns static profiles into autonomous agents.
3. Preference Model.The user’s arbitrary personal intervention into how their capacities are applied. Capability is what the user can do; preference is what they choose. The two are sharply distinguished and connected only by explicit bridges: where the user is allowed to override capacity, the bridge names that override; where capacity overrides preference, the bridge names that direction instead. The model never collapses the two into one because the political question of who decides what is decided differently in different jurisdictions and different contexts.
4. Requirement Model.The counterpoint to Capacity. Modalities have requirements; users have capacities; the runtime selection mechanism is the match between the two. Interaction modalities are eliminated from consideration when their requirements exceed the user-and-context’s capacity; the modalities that survive are the candidate paths against which the at-least-one condition is evaluated.

Together the four models constitute a complete specification: capability defines the abstract user; capacity binds the user to a context; preference admits user agency; requirement filters what modalities are available. Functional accessibility becomes the statement “at least one capacity-requirement match exists for this user-context-modality.” Intrinsic accessibility becomes “the underlying interface admits many such matches across many pseudo-users.” The same machinery measures both; the only difference is what is being quantified over.

Why bolt-on assistive tech is the wrong shape

The PacMate from page 2 is the illustration. It achieves functional accessibility for blind users at the cost of being no longer the original device. In the pseudo-user vocabulary: it adds a single specialist medium-and-protocol path that resolves the capacity-requirement match for one specific cell of S_PU, while leaving every other cell unchanged. N_IB goes up by one.

An intrinsically accessible alternative would not add a path; it would change the underlying interface so that the existing paths admit the additional cells. N_IB goes up by many. The user-experience cost of specialisation does not get paid by anyone, because no specialisation happened.

Bolt-on assistive technology is therefore structurally a functional-accessibility solution to a problem that wants intrinsic-accessibility. Each individual bolt-on may be a perfectly reasonable response to the specific situation it was built for; the cumulative effect of treating the whole field that way is to keep N_IB low and the cost of accessibility high. The structural alternative is to make the underlying interfaces shape-changeable enough that the additional paths are intrinsic, not bolted on.

What the breadth measure does not promise

N_IBmeasures how many pseudo-users the interface admits. It does not measure whether the experiences the admitted pseudo-users have are equivalent to one another. A user with full sight and a user navigating only by screen reader can both be counted in the breadth; that count says nothing about whether the second user’s task takes ten times as long, three times the effort, or fewer context-switches. Equivalent experience is a different question, treated on page 4.

The measure also does not name how the interface should be built to admit the additional pseudo-users. That is a methodological question — how do you design and engineer an interface so that intrinsic accessibility is achievable as a property of the underlying system rather than an aspiration painted on top? The methodological substrate that makes the formal definitions on this page buildable rather than just stated is treated on page 5.

Continue

← Previous: 2. Functional Accessibility · Next: 4. Equivalent Experience →