Persona-Fiction Prototyping

Workbook-grounded and author-clarified; proposed reconstruction. Design work often needs a concrete representation of possible users, participants or stakeholders. Some early and low-consequence questions require only a rough persona created from the brief. Deeper or more consequential work may require a persona synthesised from authorised research about the relevant group.

Proposed and author-clarified. Teams frequently design for an imagined generic user or rely on thin demographic stereotypes. GenAI can produce richly detailed personas quickly, but detail creates an illusion of knowledge. When rough and research-backed personas are not distinguished, an invented character may be mistaken for evidence or a research synthesis may conceal the limits and variation of its sources.

Proposed and author-clarified. Choose and label one of two persona modes. Create a rough, non-research-backed persona from an explicit brief for rapid and provisional testing. Create a deep, research-backed persona by synthesising authorised evidence about a stakeholder or user group and preserving traceability to that evidence. In both modes, separate evidence, assumption and invention; use the persona to probe design questions rather than to simulate testimony; and revise or retire it when it no longer helps.

• A set of inspectable persona hypotheses with known, assumed and invented elements labelled.
• An explicit classification of each persona as research-backed or non-research-backed.
• Design questions and tensions that can be tested through research or consultation.
• Reduced reliance on an unspoken generic user.
• A clear boundary preventing persona fiction from being cited as stakeholder evidence.

Researchers or designers define the purpose, select the persona mode and decide whether the consequence of the design requires direct research or participation. Relevant communities and domain experts challenge research-backed representations where appropriate. The system generates or synthesises provisional narratives. People determine whether the persona is respectful, useful, traceable and sufficiently grounded for its intended use.

Reconstructed risks

• Generated detail is mistaken for research data or lived experience.
• Personas reproduce stereotypes, deficit framings or tokenistic representation.
• A fictional voice is treated as consent or stakeholder endorsement.
• Teams design to the persona narrative and ignore contradictory evidence.
• Sensitive characteristics are inferred unnecessarily or encoded in reusable prompts.

• Minimum viable implementation: use a short template that records the design question, persona mode, relevant characteristics, assumptions and prohibited uses. Retain the design questions generated through the persona and delete or revise the persona when the exercise is complete.
• Robust implementation: maintain an evidence map for research-backed personas, distinguish source findings from synthesis and invention, record participatory review, version the persona and assign a review or expiry date.
• Low-code implementation: use a form to select the persona mode and populate the appropriate template, then connect the labelled persona to a structured artefact test whose outputs are recorded as hypotheses requiring review.
• Local or privacy-preserving implementation: keep interview, survey, service and participant material in controlled storage; use de-identified extracts and approved local processing where the evidence is sensitive.
• Speculative implementation: generate deliberately contrasting persona variants to test edge conditions, while limiting the number produced, preserving the mode label and preventing automated personas from authorising design decisions.