Keeping records that may mislead

A photograph preserves a viewpoint with extraordinary detail, which can make its exclusions look accidental rather than chosen. I record where I stood, what was outside the frame and whether people changed behaviour when the camera appeared. The notes do not neutralise the image, but they prevent later viewers from treating it as an unlocated window.

Selection is unavoidable; invisible selection is the danger. The archive includes rejected frames when they reveal how the final composition was achieved. I preserve contact sheets, captions and the sequence in which images were selected, not only the polished photograph.

A later researcher can see how framing, access and editorial purpose shaped what appears to be direct observation. The final image remains evidence, but its production history prevents technical finish from being mistaken for an unmediated copy of the event. For every photograph I record where the camera stood, what was cropped and why the image was made.

Detail then remains valuable without being confused with neutrality, and rejected frames can show how the final view was constructed.

Instrument logs can create an illusion of continuity because the same variable name survives several software upgrades. I document changes in calibration, default filtering and operator training beside the numerical series. A smooth graph may join measurements that were never produced in exactly the same way.

Deleting the old data would be worse, but combining it without qualification is also misleading. The record must preserve comparability as a question, not assume it as a property. Missing notebook pages, undocumented calibration changes and staff transitions are described beside the numerical series.

Those omissions show which activities the laboratory treated as formal work and which were left in personal memory. A later manager can distinguish a gap in the phenomenon from a gap produced by institutional recording, and can limit comparison without discarding the surviving data. When software changes, the old values are retained with conversion notes and known discontinuities.

A familiar variable name cannot by itself guarantee comparability, so later analysts are told exactly which relationships survived the upgrade.

A transcript removes hesitation, gesture and the interview setting unless those features are deliberately represented. I keep the audio and a field note describing what prompted each part of the conversation. A polished transcript can make memory sound more settled than it was.

At the same time, excessive notation may turn a person’s account into a specimen. The record should preserve uncertainty without denying the speaker’s narrative agency. The public account may be organised thematically, but I retain the interview chronology and the moments when one recollection changed another question.

A later historian can recover the route from testimony to interpretation and detect where editing created an appearance of inevitability. A smooth story is not allowed to erase hesitation, correction or the order in which evidence acquired meaning. A transcript preserves pauses and revisions when they affect meaning, but the speaker decides whether vulnerable material remains.

Smooth prose is not automatically more truthful; sometimes it removes the sequence through which a memory became uncertain.

Training data become misleading when athletes alter their behaviour to satisfy the metric being watched. I rotate which measures receive attention and compare them with video and self-report. A rising number may reflect a better performance or a better strategy for producing the number.

The record is useful when it prompts a question, not when it closes evaluation. Athletes should know how measures are interpreted so that monitoring does not become secret judgement. When a player does not fit an established category, I record the competing classifications and the observations favouring each.

A later coach can update the label when performance, role or physical development changes, rather than inheriting one provisional judgement as fact. Classification remains usable for planning while its uncertainty stays visible to the people affected by it. Athletes are shown how each indicator will be interpreted and where it is unreliable.

If behaviour shifts merely to satisfy the dashboard, we change the measures and record the classification uncertainty rather than rewarding an illusion of progress.