The Eight Lenses › Lens 7: Marginalised Voices

Lens 7: Marginalised Voices

What this lens looks for

The final lens is the most direct, and it functions as the framework's integrity check. For any communication it asks: who are the stakeholders of this matter, which of them speak in it, which are spoken about without speaking, and which appear nowhere at all? "The table" is concrete: the sources quoted, the experts consulted, the authors credited, the reviewers acknowledged, the data subjects consenting. Absence from the table is recorded as a finding with the same evidentiary discipline as any other — naming who is missing, why their presence was reasonably expected, and what the communication loses by their absence.

Why it is an architectural constraint

This lens operationalises the framework's standpoint-theory grounding at its most literal: marginalised positions reveal structural blind spots the centre cannot see. The practical consequence is that an absent voice is not merely an unfairness — it is a detection failure. Whatever that position would have seen is now invisible to the communication and to its readers, and no amount of accuracy in the present voices compensates, because they are accurate about a smaller world. This is also why the lens sits at the end of the outward-facing sequence: after the six before it have mapped specific kinds of absence, Lens 7 asks the summary question — whose presence would have changed everything the audit just found? (Only Trickster runs after it, turning the audit on the audit itself.)

Distinguishing the three absences

Spoken about, not speaking

The group is the communication's subject — "gig workers", "patients", "affected communities" — but every quoted voice belongs to someone else: executives, analysts, officials. Presence as topic; absence as agent.

Present as token

One representative voice included, positioned as garnish; quoted late, framed by others, given reaction rather than analysis. Detectable by structural position, not by counting.

Structurally absent

Not mentioned at all, though materially affected. The hardest to detect and the reason this lens requires the auditor to build a stakeholder map from outside the communication before reading it — using only the text's own cast list guarantees inheriting its exclusions.

Characteristic gaps this lens detects

• All-spokesperson sourcing — institutional voices only, on a matter whose weight falls on non-institutional people.
• Affected-party silence — workers in automation stories, moderators in platform stories, data subjects in dataset stories.
• Credential gatekeeping — lived experience excluded because it lacks the accepted credential, in exactly the contexts where lived experience is the relevant expertise.
• Aggregate erasure — affected people present only as statistics, with no route from the number back to a voice.
• Reply-channel absence — no mechanism for the spoken-about to respond; the gap the right-of-reply pattern exists to close.

When the lens binds

It binds strongly on essentially all public communication — every matter has stakeholders, and the lens's expectation scales with the stakes: the more a communication's subject-matter falls on a group, the stronger the expectation that the group speaks. It binds on Preservative AI itself: the limitations page records whose voices this framework still lacks, and adoption requires compensated advisory roles rather than extracted critique.

Scoring guidance