The Difference Between a Record and a Reason

An image can have a detailed history and still resist a simple explanation. A record may tell us how it was made; a reason asks what it makes possible within a work.

In the previous essay, I considered the choice made before its explanation arrives. Suppose the choice has now been made and the files are safely kept. There may be a camera exposure, a prompt, a series of revisions, a date, and a list of rejected versions. Which of these tells us why the image belongs in a book?

Perhaps none on its own. That does not make the record unimportant. It asks us to distinguish the questions we put to it.

An instruction is not the finished work

Sol LeWitt's Wall Drawing 86 gives that distinction a useful form. According to MASS MoCA, the work's instructions define the basic conditions of its lines, while the person drawing decides aspects such as their orientation and intersections as the work progresses. The museum describes a contrast between the instruction's simplicity and the visual complexity that results. [1]

The instruction matters: it establishes the terms under which the drawing is realised. But it is not a photograph of every line on a particular wall. Nor does reading it alone settle what a viewer makes of the density, intervals or encounters between lines in an installation.

Another of LeWitt's works complicates the contrast. For Wall Drawing 159, MASS MoCA describes a more precise geometric construction and notes that previous solutions can provide a template, while each installation still leaves room for interpretation. [2] The two works do not divide neatly into freedom and control. They suggest that a set of instructions can be more or less specific, while its relation to a material result remains something to examine in each case.

This is not a claim that an AI prompt is equivalent to LeWitt's instructions, or that a model and a draughtsperson participate in the same way. The comparison is narrower: in both situations, knowing a verbal instruction and seeing a realised image are different kinds of access. One cannot simply be substituted for the other.

What a record can establish

Imagine a generated image of an empty room, kept as one candidate for a photobook. The accompanying file says that the maker requested a window, a table and oblique light. A later record shows that the image was cropped before being placed in a sequence. This is a hypothetical example, not a report of a particular work or experiment.

The prompt can show what was requested, if it is the actual prompt used. The revision record can show what was changed, provided the relevant versions survive. Neither automatically establishes why the crop was made, whether it improved the image, or what a reader will see in the finished book.

Now imagine that the uncropped image gives the window and table equal weight, while the crop removes part of the window and leaves a larger interval of wall beside the table. A reason for the crop might be that it makes the interval more difficult to pass over. That proposition refers to the visible arrangement. Another reader could find that the crop merely removes useful context. Both readers can look again at the images; neither needs to accept the first account because the maker kept a meticulous log.

There is an important difference between these statements: “I cropped the window” and “I cropped the window so that the interval would delay the reader.” The first describes an action. The second offers a purpose. A third question remains: does the resulting page give a reader grounds to notice that interval at all?

When the record changes the reading

It would be too easy to conclude that the image is all that matters. A record can change a responsible reading. A photograph presented as a scene witnessed by a camera invites questions about time, place and the circumstances of exposure. An AI-generated image may depict a similarly convincing room without recording such a room. The two images can both be considered for a book, but they should not be made to carry the same documentary claim merely because they look alike.

Here production information is not a decorative addition to interpretation. It helps a reader understand what sort of claim is being made. Equally, naming an image “AI-generated” does not by itself explain the choice of crop, sequence or subject. Disclosure and interpretation have different jobs; a short label can do the first without completing the second.

A prompt also has a practical use for the maker. It can make a process easier to revisit, reveal which descriptions were tried, and prevent a later account from erasing the record of earlier decisions. But an attractive prompt may begin to stand in for an attractive image. Conversely, a plain or awkward instruction may lead to an image worth keeping. The wording of the request and the achievement of the result need not have the same character.

Why the page changes the question

A record organised around files tells us which image came first in production. A book can put a later image before an earlier one. That change is not an error if the book is not presented as a chronology of making. It is an editorial decision, with consequences the production log cannot decide in advance.

Consider again the hypothetical room. Placed after a picture crowded with objects, its open wall might offer an interval. Placed after several similarly empty interiors, it might prolong a pattern or make the sequence feel repetitive. The same image and the same production history remain available in both positions. What changes is the relationship a reader encounters.

Neither version is automatically better. A book might need continuity rather than contrast. The point is that “this was the fifth attempt” cannot answer whether repetition is doing useful work on these pages. To answer, I would look at the neighbouring images, the page turn and the pace of the sequence, not only at the file history.

There is a counterargument. In process-based work, the successive trials may themselves be the work. Showing them in order can make revision, failure or the limits of a system visible. In such a book, chronology is not merely background information: it is part of what the reader is asked to consider. Even then, the maker chooses what to retain, how to display it and where a reader can enter. A complete record is also an arrangement.

A modest distinction to keep

I would keep three kinds of note separate when they are available: what was done, why a choice was made, and what that choice appears to do in its present setting. The first can include a date, a source file or a change of crop. The second might record an intention, including uncertainty at the time. The third remains open to another person's reading.

These notes do not have to accompany every published image. Sometimes the work needs room to be encountered before its making is explained. Sometimes an essential disclosure belongs close to the image. The position of a note is itself a decision about what the reader can know and when.

The previous essay left room for a first attraction that had not yet found its reason. A record can preserve that interval: “I kept this; I could not yet say why.” It need not turn the uncertainty into a theory. Later, a reason may emerge from the relation between the image and its neighbours. It will be stronger if someone else can inspect that relation—and if the maker is prepared to change the arrangement when the reason no longer holds.

References

1. MASS MoCA, Sol LeWitt, Wall Drawing 86. Museum description of the work and its instructions.

2. MASS MoCA, Sol LeWitt, Wall Drawing 159. Museum description of the work and its installation.

Related reading

Knowing What to Keep Before Knowing Why — On first attraction, later explanation and a decision's place in a sequence.

The Distance Between the Wall and the Page — On how the placement of words changes what a reader can inspect.

Young Chai

Visual Artist · Writer

https://luxlatens.com
Previous
Previous

What Can an Archive Supply?

Next
Next

Knowing What to Keep Before Knowing Why