Most interfaces treat hesitation as friction to eliminate. This note proposes aporia as a designed state: incompatible premises made visible, the call to action suspended, and the user helped to name what would have to be true before proceeding.
Product language has a preferred emotion for decision: smooth. Smooth means the next click arrives before the previous doubt finishes forming. Smooth is sold as respect for the user's time. Sometimes it is. Sometimes it is respect for the shipping calendar wearing a UX badge. The institute is interested in a different state — not confusion as default, not paralysis as virtue, but aporia as something an interface can hold without pretending the hold is a bug.
Aporia, in the long philosophical and rhetorical tradition that includes Socratic elenchus and later deconstructive attention to thresholds, names a difficulty that is not merely a missing fact. It is a bind: conditions of possibility that also function as conditions of impossibility; a question that cannot be answered from inside the terms that made it urgent. Keats's negative capability — the capacity to remain with uncertainties without irritable reaching after fact and reason — is not a productivity tip. It is a description of a disciplined refusal to fake closure.
Source ·Aporias (opens in new tab)· Jacques Derrida / Stanford University Pressatlas
Source ·Negative Capability (opens in new tab)· Poetry Foundation / John Keatsatlas
Interfaces rarely afford that discipline. They afford progress indicators. The institute's proposal is not to replace every funnel with a seminar. It is to recognise a class of moments where the honest next state is neither "choose A" nor "choose B" nor "loading," but a structured bind: two or more commitments that cannot be jointly satisfied under the current description of the world, and a decision that still presses.
1 — Hesitation as treated today
In contemporary decision support, hesitation is usually modelled as latency, incomplete information, or insufficient persuasion. The remedies follow: more defaults, more recommendations, more social proof, more "people also chose," more gentle urgency that is not gentle. The system assumes the user is stuck on the way to a choice, and that the designer's job is to shorten the way.
Sometimes that diagnosis is correct. Choosing a shipping option is not an aporia. Choosing whether to delete a dataset that cannot be reconstructed, or whether to deploy a model whose failure modes are unevenly distributed, or whether two institutional commitments can both be honoured, often is. Treating those moments as friction is not neutral helpfulness. It is a theory of mind that denies the bind.
There is a quieter variant of the same mistake in so-called intelligent assistants: the system offers a synthesis that splits the difference, or a ranked list that pretends commensurability where the premises refuse it. Ranking is a powerful solvent. It dissolves qualitative conflict into ordinal comfort. Users feel helped. The conflict has not been resolved; it has been formatted.
The comic failure mode is familiar: a modal that asks you to confirm an irreversible action in the same visual language used to confirm a newsletter subscription. The interface has one volume knob for seriousness and it is broken at eleven for everything.
2 — Aporia as designed state
An aporetic interface does not withhold the button forever. It changes what the button means until the bind has been named.
At minimum, the state does three things. It reveals incompatible premises rather than averaging them into a compromise nobody believes. It suspends the primary call to action long enough for the incompatibility to be readable. It asks the user — or the operator, or the committee — to state what would have to be true for each path to be justified. The third move is easy to fake with checkbox theatre. The institute's standard is harsher: the named conditions must be specific enough that a later audit could tell whether they obtained.
Proceed now: treat the remaining uncertainty as ordinary risk and accept responsibility for acting under incomplete information.
Wait with structure: refuse the call to action until the incompatible premises are listed, owned, and assigned conditions under which each would license a decision.
Neither option is "being careful" in the vague sense. Both are commitments. The first is often correct under time pressure with reversible stakes. The second is often correct when the interface's smoothness would otherwise launder a conflict into a preference. Designed aporia is the second commitment made legible as a first-class state, not as a loading spinner with better copy.
3 — What the state is not
Aporia is not a dark pattern for engagement. Keeping people undecided so they remain on the page is the opposite of the proposal. Aporia is also not romanticised indecision. The institute does not praise dithering, and it does not treat contradiction as an aesthetic. Hallucinated both-sides fluency — the model that can argue any horn of a dilemma without noticing it is a dilemma — is a failure mode adjacent to this topic and not a friend of it.
Negative capability is not infinite patience. It is the capacity to stay with a real uncertainty without manufacturing a fake resolution to soothe irritability. An interface that implements the soothing and calls it decisiveness has failed twice: once as epistemology, once as manners.
There is a second objection from the optimisation culture that funds most interfaces: if you can measure completion rate, you can improve it, and anything that reduces completion looks like regression. We accept the metric's honesty about what it measures. We deny that completion is always the right thing to measure. A completed decision that erased the bind is not a success with caveats. It is a different event wearing the success costume.
4 — Naming before proceeding
The operational heart of the proposal is linguistic and procedural. Before the primary action unlocks, the system should force a short articulation: which premises conflict; what each premise protects; what evidence or authority would license choosing one over the other; what will be logged as the reason for proceeding if the user overrides. This is not a philosophy seminar stuffed into a form. It is closer to a pre-mortem that admits the conflict is conceptual, not merely predictive.
Socratic practice produced aporia by cross-examination. Modern systems produce false certainty by autocomplete. The institute is not asking product teams to become Plato. It is asking them to stop treating the moment before irreversible action as an awkward pause in the funnel. Sometimes the pause is the work. Sometimes the incompatible options are the most honest UI the organisation will ship all year.
A practical test helps. If removing the hesitation-state would change only the time-to-click and not the quality of what can later be said about why the click occurred, you did not have aporia; you had delay. If removing it would erase the only place where incompatible premises were forced into the open, you had a designed state doing epistemic work — and optimising it away is not a UX win. It is a documentation deletion performed in the present tense.
Related notes on explanation-without-cause and on logic as a local maximum belong beside this one. An unfaithful rationale can fake the exit from aporia by narrating a path that was never taken. A local logical triumph can fake the exit by proving brilliantly inside one horn of a dilemma. Designed aporia is the refusal of both fakes until the bind has a name.
“Charming. Please add a toggle to skip the aporia for power users, by which we mean everyone.”
We will not add that toggle as a default. Power users are often the people most practised at laundering binds into preferences. The skip, if it exists, should be costly, logged, and socially visible inside the institution that claims to care about the decision. Smoothness can return after the naming. Smoothness that arrives before the naming is not speed. It is amnesia with a progress bar.