What should an art expression test measure?

What should an art expression test measure?

Research blog / What should an art expression test measure?

A flawless answer to a factual question can tell us that a reader located a detail. It cannot tell us whether an artwork offered a deep encounter, attended to another’s position, or made a meaningful reply possible.

The first study answered a smaller question

The garden of unfinished agreements began as a Carrier matrix: 56 distinct concepts, 59 occupied cells, four deliberate blanks and 40 declared relations. Equivalent JSON and explicit prose controls preserved those choices. Nine fresh readers interpreted the work, froze their readings and composed replies before receiving factual probes.

All nine scored 16/16 and their replies passed exact recovery checks. Every reader also used a common adapter that supplied identical normalized structures and definitions. The result supports an assisted exchange, not an isolated colour or notation effect. Two model appraisers gave the highest category throughout the first rubric; that ceiling did not distinguish artistic quality.

The artwork needs its own question

Lucid’s observation was decisive: the comparison should consider the depth and quality of expression first. For this inquiry, care means attention to another’s particular situation, burden and agency. Freedom means meaningful room to interpret, disagree, transform, withhold or leave without being silently converted into assent.

These are authored evaluative commitments, not properties a decoder can certify. A disturbing work can show care. A comforting work can appropriate another person’s future. A constraint can enable a distinctive voice, while an empty space can simply move responsibility onto the viewer. The work has to supply the evidence.

A second round with native choices

The new round gave two themes to three native media, with one fresh creator per theme and medium. Carrier used its existing seven-by-nine chromatic profile; JSON authors could create their own structures; prose authors could develop scenes, voices, rhythms and omissions. The controls were no longer transcripts of the same matrix.

Each work was frozen before its creator wrote a private intention note. That note stayed out of the fresh viewer’s input. The viewer first fixed a reading, including a grounded alternative or the reason one could not be supported, then recorded a choice of reply medium before making the response. Every viewer received the same options, including meaningful refusal.

Of six fixed planned encounters, 4 source works and 4 answering works are available. The completed viewers exercised their response choices as follows: prose 4. Those choices are self-selected observations under this workflow, not evidence of a general preference or artistic superiority.

Depth cannot be counted in concepts

The new rubric distinguishes six things: relations with consequences; formal choices that do particular expressive work; situated care; grounded interpretive openness; answerability without capture; and developed possibilities of meaning. Each source and reply receives separate evidence-backed ordinal judgments and counterreadings. Their arithmetic sum is not a universal artistic-quality score.

A source can draw on actual reception when assessing whether its invitation made room for another voice. A reply has no subsequent viewer in this round. The rubric therefore does not grant it the highest category that requires observed next reception. Missing evidence is retained as a limit, rather than supplied by praise.

What became particular

In Packing, care is tested by a practical task around a sold cello, including the temptation to convert help into a claim on another’s decision. Only the Pocket uses a repair, interrupted dialogue and unfilled JSON values to make the boundary of permission matter. These are artistic decisions that can be read and challenged; they are more specific than a list of approved values.

The viewer is not required to preserve the author’s interpretation. A reply may retain the situation and change the speaker, question whose labour the invitation requires, or use another format to keep a tension alive. Those choices matter as observed contributions, not as automatic proof that the source caused greater creativity.

The viewer of the Carrier work The Room Between Shifts offered a pointed alternative: the precise arrangement could also be read as an administrative fantasy that gives even grief and refusal their proper addresses. The same reader found the blanks resistant to that closure, and answered with a scene in which a routine handover question can become another demand on a recipient. The study preserves that tension as a reading and response, not as a verdict that either interpretation must win.

The judgments remain contestable

Two fresh appraisers judged each source and reply separately, with the creators’ intention notes withheld. They disagreed on five case-and-dimension judgments. Both used only categories 3 and 4 for the available works, so the new rubric still has a substantial upper-end ceiling. That may reflect this material, shared model tendencies or lenient anchors; these data cannot separate those explanations.

One useful disagreement concerns the Carrier composition: does arranging competing obligations fully develop the rival reading, or chiefly make that development possible for the viewer? Both appraisers found its care attentive but demanding of the reader’s time. For the prose reply Under Load, the question was whether its insistence on a repair that works adequately accommodates the position and burden of the person addressed. These are situated judgments about works, not coefficients for their media.

The sources have actual receptions here; the replies have no next viewers. The four reply answerability ratings therefore cannot receive the top anchor, by design. That evidence boundary does not establish that answering art is worse than its source. No ratings are summed into a winner.

Keep the comparison within its evidence

The new protocol was fixed before its new works, after the first study’s results were known. Native forms, source creators, viewers, tools and practical allowances differ. In particular, the Carrier condition is chromatic-field/1, with fixed roles and operators, not the entirety of Carrier or USL. The prose and JSON conditions can introduce arbitrary literal detail that this specific matrix cannot.

The planned two cases per source medium are too few for a general ranking, and the two retained interruptions make that sample uneven. Model process notes do not establish felt freedom; model appraisals do not establish human aesthetic response. The new judgments are unblinded to native form. All of these limits belong beside the work.

There was also a bookkeeping error: an append-only coordinator event log was included in the local pre-dispatch freeze, then appended. Its whole-file hash no longer matches. The original inventory and changed log are both retained, with the initial prefix and unchanged immutable materials checked separately. A disclosed exception is not a claim that the whole freeze passed.

A useful next measure

The stronger question is what a medium helps someone do that matters to them, at a tolerable cost, while leaving the next person a meaningful voice. This release approaches that question through actual artworks and replies. A later study would need more themes, multiple viewers per work, diverse creators, and human reception if it wants to make claims about people’s experience.

For now, the evidence is available in its proper forms: transport checks for transport, fixed probes for those facts, artworks for encounter, and contestable judgments for artistic qualities. None has to impersonate the others.

Enter the art-first galleryMatched-input galleryFresh viewer collectionArt-first study record