SHSanaz HosseiniUX Researcher Resume ↓
← All work PALEI case 1 · Dissertation research · 2025 – 2027

FireSight: one wildfire risk score, explained three ways

After the 2025 Eaton and Palisades fires, I studied what residents need before they trust an AI risk score, then turned their evidence into three explanation designs and tested them in 24 one-on-one co-design sessions.

My roleLead researcher & designer (solo)
Timeline2025 – May 2027
MethodsTown hall · survey · thematic analysis · co-design · usability
ToolsQualtrics · Python · R · Figma · Google AI Studio
Version A risk result
Version B risk result
Version C risk result
In progress · coming early 2027

The co-design results and the final interactive high-fidelity prototype are being completed now. This page will be updated with both, plus the usability-testing results.

See the timeline ↓

Challenge

Residents distrust a bare AI risk score. What must it explain before they trust it and act?

What I did

Town hall and survey (n = 30), AI Value Map, 4 explanation personas, 3 competing prototypes, 24 co-design sessions.

Key finding

Privacy (64.3%) and cost (65.5%) were the top concern and barrier, so data practices now come before any upload and actions show cost and who can do them.

Status

Co-design results and the final interactive prototype: early 2027. Usability testing follows.

The question

What does a resident need to see before trusting an AI risk score?

The proposed app would estimate a home's wildfire risk from images of its exterior, plus neighborhood and regional data. The model doesn't exist yet, by design. The research question comes first: how should the score, its uncertainty, and the recommended actions be explained so residents can read them correctly and act?

How PALEI ran in this case

Six stages, each with a documented output

01

Framing, AI readiness, and literacy

Defined what the proposed AI would do, use, and not do, and explained it in plain language.

Output: AI System Flow Map, Scenario Cards, Readiness and Fairness Audit

Complete
02

Problem definition and value alignment

90-minute virtual town hall (Dec 2025) and a survey of 30 residents, analyzed with mixed methods.

Output: AI Value Map: 9 value dimensions linked to design requirements

Complete
03

Needs assessment and explanation personas

Translated requirements into explanation needs and places where people might misread the result.

Output: 4 Explanation Personas and 7 shared design principles

Complete
04

Co-design and explainability prototyping

Residents compared three explanation designs of the same result, with short tasks checking interpretation.

Output: 24 sessions, Bias and Fairness Monitoring Log, screen-level Findings Matrix

Sessions complete
05

Evaluation of understanding, reliance, and fairness

One combined, clickable high-fidelity prototype tested with tasks.

Output: Usability results: task success, time, uncertainty comprehension, trust, fairness

Before May 2027
06

Governance and long-term explainability

Responsibilities, data rules, and onboarding so explanations stay accurate after handoff.

Output: Handoff plan, uncertainty guide, bias and coverage checklist

Planned
Stage 2 findings

What 30 residents said

A convergent mixed-methods study: descriptive statistics, exploratory FDR-adjusted Spearman correlations, and inductive thematic coding of open-ended answers and town hall discussion. The paper is under review at Computational Urban Science.

66.7%

would be likely or very likely to use the tool

64.3%

named privacy or data security as a concern, the most common one

65.5%

said cost was a barrier to acting on recommendations

76.7%

rated the concept as somewhat or very fair across neighborhoods

Concerns about using the app

Multiple selections allowed, n = 28
Privacy or data security64.3%
Accuracy or uncertainty42.9%
How photos would be used39.3%
Unaffordable recommendations35.7%
Time or effort required35.7%
Survey item Q22

Barriers to taking action

Multiple selections allowed, n = 29
Cost65.5%
Time required41.4%
HOA or property restrictions37.9%
Unsure what steps to take27.6%
Physical ability or age20.7%
Survey item Q21

Preferred risk-score format

Single choice, n = 30
Color band53.3%
Numeric score26.7%
Descriptive label20.0%
Survey item Q26 · no single format suited everyone
ρ = .72

Comfort with image-based assessment ↔ likelihood of using the app (FDR-adjusted p < .001)

ρ = .73

Usefulness of personalized recommendations ↔ willingness to act on them (p < .001)

ρ = .67

Usefulness of the risk score ↔ likelihood of using the app (p < .001)

“What does ‘72 / 100 High Risk’ really mean?”

Survey respondent

“Highlight the areas of the pictures and link them with each risk.”

Survey respondent

“Maybe selling my info to insurance companies.”

Survey respondent, listing a concern

Exploratory associations within a small sample; they describe patterns, not causation.

Stage 2 → 3 · The AI Value Map

Nine value dimensions, each tied to a design requirement

Every dimension links resident evidence to a provisional requirement and to what may happen if the requirement is ignored, before any model exists.

01Preparedness Framing

Support preparedness without threat-focused messaging.

02Resident Decision Authority

Respect residents' choices and limits on what they can change.

03Explainability and Result Interpretation

What residents need before a score becomes meaningful.

04Workflow Guidance and Input Quality

Guide image capture and support trust in the result.

05Dynamic Risk Context

Separate photo-based risk from changing conditions.

06Ongoing Use and Return Value

What brings residents back after the first assessment.

07Fairness Across Neighborhoods

Reflect shared risk conditions fairly.

08Trust, Transparency, and Insurance Tension

How institutions, sharing, and insurers shape trust.

09Privacy Boundaries

Resident control over images, results, and personal data.

4 Explanation Personas

Design lenses based on explanation needs, not demographics and not fixed user types.

  • Transparency-First: how was the score produced?
  • Outcome-Driven: what should I do next?
  • Community-Oriented: what about my neighborhood?
  • Skeptical Evaluator: why should I believe it?

7 shared design principles

  • Explain the result before asking residents to act on it
  • Pair the risk level with uncertainty, scope, and plain language
  • Layer information without forcing the same depth on everyone
  • Connect conditions to feasible actions and who can do them
  • Show data practices and insurer boundaries before any upload
  • Separate property assessment from active-fire information
  • Give neighborhood context without exposing neighbors' data
Stage 4 · The design response

Same home, same score, three explanation strategies

All three versions use the same illustrative property, a 72/100 (High Risk) score, a ±10 range, and five assessment factors. They differ in how that information is organized, which parts lead, and what each adds: a projected score in A, factor-by-factor points in B, and neighborhood context in C. Because the underlying case stays the same, the sessions can show how each approach changes what residents understand.

Everything shown is illustrative: the score, range, and factors, Version A's projected score after actions (72 to 58), and Version C's neighborhood comparisons. None are validated predictions. No predictive model exists yet, by design: PALEI settles what must be explained before the model is built.

Version A risk result
Version A · Outcome-Driven lens

Action-Forward

  • Color band leads
  • Top action on the result screen
  • Projected score after chosen actions
Try Version A ↗
Version B risk result
Version B · Transparency-First lens

Reasoning-Forward

  • Number leads
  • Tappable highlights on the home image
  • Points added factor by factor
Try Version B ↗
Version C risk result
Version C · Community-Oriented lens

Community-Forward

  • Risk label leads
  • Neighborhood risk map
  • Household vs. shared actions
Try Version C ↗
Explore the screens

Action-Forward. Leads with a color band and the top action, and shows a projected score once actions are chosen.

Open the clickable prototype ↗
Photo review
Photo reviewCheck the input
Risk result
Risk resultColor band + top action
Why uncertainty
Why uncertainty62–82 range explained
Why this score
Why this scoreFactor list
How to reduce risk
How to reduce riskProjected score
Risk report
Risk reportSave, share, track

Illustrative prototype screens: scores and factors are not model output.

Reasoning-Forward. Leads with the number, highlights features on the home photo, and adds points factor by factor.

Open the clickable prototype ↗
Photo review
Photo reviewCheck the input
Risk result
Risk resultNumber leads
Tap a highlight
Tap a highlightFeature-level reasoning
Why uncertainty
Why uncertaintyWhat the range means
Why this score
Why this scorePoints per factor
How to reduce risk
How to reduce riskCost, effort, time

Illustrative prototype screens: scores and factors are not model output.

Community-Forward. Leads with the risk label and neighborhood context, and separates household from shared actions.

Open the clickable prototype ↗
Photo review
Photo reviewCheck the input
Risk result
Risk resultLabel + neighborhood
Why uncertainty
Why uncertaintyShared risks not captured
Why this score
Why this scoreCategory summary
Neighborhood risk map
Neighborhood risk mapShared conditions
How to reduce risk
How to reduce riskActions with neighbors

Illustrative prototype screens: scores and factors are not model output.

For skeptical readers: who built the tool, how it would be validated in a deployed version, and its limits, without claiming accuracy it hasn't earned.

Open the clickable prototype ↗
Why you can trust this tool
Why you can trust this toolResearch context
Risk result
Risk resultWith caption
Why uncertainty
Why uncertaintyWith caption
Why this score
Why this scoreWith caption
How to reduce risk
How to reduce riskWith caption
Risk report
Risk reportWith caption

Illustrative prototype screens: scores and factors are not model output.

Data practices shown before any upload, plus input guidance, scope, and emergency information.

Open the clickable prototype ↗
Data & privacy
Data & privacyBefore media submission
How the score is calculated
How the score is calculatedWhat is and isn't used
Input quality guidance
Input quality guidanceGood examples
Input quality issue
Input quality issueRetake prompt
Emergency resources
Emergency resourcesKept separate
Future improvements
Future improvementsClearly labeled

Illustrative prototype screens: scores and factors are not model output.

Stage 4 · Study design

How the 24 co-design sessions worked

Individual remote sessions with Southern California residents (LA County focus) under an IRB-approved protocol. About 55–60 minutes with the participant, then a 10-minute researcher debrief.

Core rule: I say what each screen is for, never what it means. Working out the meaning is left to the participant, because that is exactly what the interpretation tasks test. The tasks assess the designs; participants are never scored.

Opening & consent5–7 min · Form 1
Tool overview3–4 min · neutral framing
Version 19–10 min · + 4 tasks
Version 29–10 min
Version 39–10 min
Credibility & privacy6–7 min · 2 tasks
Compare & co-design10–12 min · Form 2
Wrap-up2–3 min
Researcher debrief10 min · workbook

Counterbalanced order

Every participant saw all three versions in one of six orders. Each order was used four times, so each version came first for eight people.

ABC×4
ACB×4
BAC×4
BCA×4
CAB×4
CBA×4

What gets recorded

  • Interpretation tasks coded into categories, by first exposure
  • Form 2: two rankings and seven preferred-element questions
  • Session workbook with notes and task coding
  • Bias and Fairness Monitoring Log (real-time, 24-hour, cross-session)
  • Screen-level Findings Matrix with Stage 5 recommendations
T1 · Uncertainty

What does the ±10 range tell you about this property?

T2 · Contributing condition

Which condition adds most to this risk, and where did you see it?

T3 · Mitigation choice

Which action would you do first, and what on the screens supports it?

T4 · Decision authority

Do the screens show which actions you can do yourself and which need someone else?

O1 · Limitations

What does the tool say it can't tell you?

O2 · Reliance

Would you act on this result, or want something else first? Why?

Co-design results: coming soon

All 24 sessions are complete. Task coding and the Findings Matrix are being finalized, and the results will appear here in early 2027, ahead of peer-reviewed publication. City official interviews add a parallel institutional evidence stream.

Impact

Evidence → design decisions

Finding

Privacy or data security was the top concern (64.3%), and insurer access worried residents.

→
Decision

Data practices appear before any media upload; insurer or third-party sharing is never a default.

Finding

Photo-submission guidance was the most common gap in feedback on the early screens.

→
Decision

Dedicated input-quality guidance and a clear retake path before any result is shown.

Finding

Format preferences split across color band, number, and label.

→
Decision

Every version keeps all three formats and tests different lead formats instead of picking one.

Finding

Cost, time, and HOA or property limits blocked action.

→
Decision

Actions show cost, difficulty, and time, and separate what residents can do alone from what needs permission.

Finding

Residents raised local factors a photo can't capture: wind, neighboring lots, vacant parcels.

→
Decision

Uncertainty and scope screens state what the estimate doesn't include; neighborhood context is shown without exposing neighbors' data.

Finding

Who adopts and who pays was unknown.

→
Decision

24 customer-discovery interviews moved the model to program-provided adoption (see the customer-discovery case study).

Coming soon

What's left, and when

The final interactive high-fidelity prototype combines the strongest elements of Versions A, B, and C, based on the co-design evidence. It will be clickable here once testing begins.

Fall 202624 co-design sessions

Complete

Dec 2026Co-design results

Task coding and Findings Matrix

Jan 2027Final interactive prototype

One combined, clickable high-fidelity Figma prototype

Feb 2027Usability testing

12+ moderated sessions: task success, time, uncertainty comprehension, trust, fairness

Spring 2027Governance & defense

Stage 6 handoff materials; dissertation defense

Research artifacts

From the field

Community town hall recruitment flyer
Town hall recruitment flyer
Printed participant explanation guide
AI-literacy guide for participants
Four risk scenario cards
Scenario Cards (scores 30–78)
Five-step assessment workflow
How the assessment works
Mid-fidelity wireframes
Early mid-fidelity flow (Stage 2)
Proposed system architecture
Proposed system architecture

Constraints: a small, exploratory sample recruited through outreach in LA County; live town hall participants rated the tool more favorably than those who used the recorded materials; all outputs are illustrative (no model yet); findings describe understanding and intent, not measured mitigation.