Case Study 02 — Resident opinion surveys

Can AI personas reproduce municipal resident opinion surveys?

We reproduced 8 questions from real resident opinion surveys in Yokohama, Fukuoka, Sapporo, and Nagoya with 200 AI personas each and validated the accuracy.

Version used on this page: v3.0 As of June 12, 2026

The results on this page were produced with the latest PersonaChoice v3.0. Evaluation uses a unified standard: “how highly PersonaChoice ranks the items that performed best in the real survey” (Top-K agreement, Top-1 match).

70–80%

Reproduction accuracy for top concerns

4 cities

Yokohama, Fukuoka, Sapporo, Nagoya

8 questions

Questions validated

100%

Rated B or above (8/8)

This page presents validation based on resident opinion surveys officially conducted and published by four municipalities across Japan. By comparing how closely PersonaChoice’s AI simulation matches actual resident responses, we assess its usefulness and reliability as a complementary tool for municipal resident surveys.

  • This validation uses 200 AI personas per task as respondents. The AI personas are generated by a proprietary method from official statistics such as the census, with age, gender, occupation, income, municipality, and personality traits.
  • The original resident surveys allow multiple answers (e.g., “choose up to 3”), but PersonaChoice converts them into a “choose the single most important” format. Therefore, the comparison uses “overlap among the top-ranked items (Top-K agreement)” rather than the absolute selection rates as the main metric.
  • All simulation result reports run during the validation are public. From each city’s “task report” link, you can view the PersonaChoice report screen directly (reports are in Japanese).

Evaluation criteria

Grade Criteria
AAgreement ≥ 80% + Top-1 match, or agreement 100%
B+Agreement ≥ 80%, or agreement ≥ 60% + Top-1 match
BAgreement ≥ 60%
CAgreement ≥ 40%
DBelow the above

* To rule out chance matches, the rank K used for evaluation varies with the number of options (rule of thumb: K / number of options ≤ 50%). Top-5 for 10+ options, Top-3 for 6–9 options (e.g., 6- and 8-option questions use Top-3).

Yokohama R5 citizen opinion survey

Source: City of Yokohama / 2,572 valid responses / 200 AI personas (18 wards)

Result: View the detailed report (JA)

Q1: Worries (original Q9, 13 options)

Grade A

QuestionChoose the single thing you have worried about or struggled with most regarding your own or your family’s life recently.

Options(1) Own illness/health and old age (2) Family illness/health and life issues (3) Unemployment/bankruptcy or falling income (4) Work and workplace (5) The economy and cost of living (6) Childcare and education (7) Children’s marriage and employment (8) Odor/noise from neighbors (9) Environmental issues (10) Accidents and disasters (11) Crime and safety (12) Housing (13) No worries or troubles

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Own health & old ageOwn health & old age Top-1 match
2Family healthWork & workplace
3Economy & cost of livingEconomy & cost of living
4Work & workplaceAccidents & disasters
5Childcare & educationFamily health

✓ = item present in both the real survey’s Top 5 and PersonaChoice’s Top 5 (regardless of rank)

Top-5 agreement: 80% (4/5) / Top-1: match

From 13 options, PersonaChoice correctly identified 4 of the real survey’s top 5, with an exact match on #1 “own health & old age.” A good example where the AI personas reproduced how worries differ by age (younger → work, older → health).

Q2: Global skills (original Q16, 6 options)

Grade A

QuestionChoose the single skill you consider most necessary to be active globally.

Options(1) Language and communication skills (2) Openness to exchange and social participation with foreigners (3) Understanding of diversity and other cultures (4) Ability to convey attachment to and appeal of Yokohama (5) Broad knowledge of international affairs (6) A sense of ownership of global issues toward a decarbonized society

RankReal survey top 3PersonaChoice top 3 (Top-K agreement)Match
1Language & communicationLanguage & communication Top-1 match
2Understanding diversity & other culturesUnderstanding diversity & other cultures
3Broad knowledge of international affairsBroad knowledge of international affairs

Top-3 agreement: 100% (3/3) / Top-1: match

The top 3 ranks matched exactly. Even with only 6 options, the ranking reproduction was perfect.

Fukuoka R7 1st city survey

Source: City of Fukuoka / 559 valid responses / 200 AI personas (7 wards)

Result: View the detailed report (JA)

Q1: Scenery (original Q23, 8 options)

Grade B

QuestionChoose the single scenery/view of Fukuoka you think is best.

Options(1) Urban scenery (office and commercial buildings) (2) Historic scenery (shrines, temples, historic sites) (3) Artistic scenery (outdoor art, monuments) (4) Local tradition and culture (festivals, events) (5) People and nature (sea, mountains, parks, farmland) (6) Night scenery (light-ups, illuminations) (7) Outdoor advertising (signs, signage) (8) None in particular

RankReal survey top 3PersonaChoice top 3 (Top-K agreement)Match
1Urban sceneryPeople and nature
2People and natureLocal tradition & culture
3Local tradition & cultureHistoric scenery

Top-3 agreement: 67% (2/3)

In the real survey, “urban” and “natural” scenery were almost tied (62.1% / 60.3%). With 8 options, we evaluate at Top-3 to avoid chance matches, and identified 2 of the top 3 (people-and-nature, local tradition/culture).

Q2: Expectations for parks (original Q18, 12 options)

Grade B+

QuestionChoose the single thing you most want for Maizuru Park / Ohori Park.

Options(1) Easier-to-walk paths (2) Brighter park (trees make it dark) (3) Better management of the waterside (moat) (4) Better views and sightlines (5) More facility/history signage and guidance (6) More shops and restaurants (7) A landmark attraction as a tourism hub (8) Clearer entrances/exits (9) More events (10) More parking (11) More benches (rest space) (12) None in particular

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Benches (rest space)Benches (rest space) Top-1 match
2Shops and restaurantsEasier-to-walk paths
3ParkingMore signage/guidance
4Waterside managementWaterside management
5Easier-to-walk pathsViews and sightlines

Top-5 agreement: 60% (3/5) / Top-1: match

PersonaChoice also hit the real survey’s #1 “benches (rest space).” Meanwhile, everyday conveniences residents value such as “shops” and “parking” did not make PersonaChoice’s top ranks — an informative gap for understanding the AI’s characteristics.

Sapporo R6 2nd / R7 4th citizen surveys

Source: City of Sapporo (R6) / (R7) / 200 AI personas (10 wards)

Result: View the detailed report (JA)

Q1: Priority PR themes (R6 Q2, 14 options)

Grade B+

QuestionChoose the single theme you most want the city PR magazine “Koho Sapporo” to focus on.

Options(1) Disaster prevention (2) Public facilities (3) Elderly welfare (4) Health & food (5) Community development activities (6) Culture & arts (7) Garbage (8) Children & childcare (9) Environment & energy (10) Education (11) Sports (12) Welfare for people with disabilities (13) Volunteering (14) None in particular

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Disaster preventionDisaster prevention Top-1 match
2Public facilitiesHealth & food
3Elderly welfareGarbage
4Health & foodChildren & childcare
5Community developmentElderly welfare

Top-5 agreement: 60% (3/5) / Top-1: match

PersonaChoice also hit the real survey’s #1 “disaster prevention.” Responses were naturally distributed across all 14 options.

Q2: Key job conditions (R7 Q19, 13 options)

Grade B

QuestionChoose the single most important condition when choosing a company to work for.

Options(1) Job content (2) Salary (income) (3) Days off/leave (4) Working hours (5) Benefits (6) Work location (7) Whether transfers are required (8) Financial health (9) Number of employees (10) Post-hire training environment (11) Corporate philosophy (12) Workplace atmosphere (13) None / don’t know

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Salary (income)Workplace atmosphere
2Job contentWork location
3Days off/leaveJob content
4Workplace atmosphereWorking hours
5Working hoursWhether transfers are required

Top-5 agreement: 60% (3/5)

Responses were distributed across nearly all 13 options — the most even distribution among all questions. It captured 3 of the top 5, covering conditions related to treatment and work style (job content, days off, workplace atmosphere).

Nagoya R6 63rd public opinion survey

Source: City of Nagoya / 911 valid responses / 200 AI personas (16 wards)

Result: View the detailed report (JA)

Q1: What makes the city proud (Q6, 13 options)

Grade B

QuestionChoose the single thing that best fits what you are proud of / like about Nagoya.

Options(1) Geographically convenient access across Japan (2) Nagoya dialect (3) A cultural/artistic atmosphere (4) Signature local foods (“Nagoya meshi”) (5) History tied to the three unifiers (6) High technical level as a manufacturing hub (7) Active economic activity (8) Tasty tap water (9) Beautiful cityscape (10) Wide, well-maintained roads (11) Advanced environmental efforts (12) Thriving sports (13) None in particular

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Convenient accessManufacturing tech level
2Nagoya meshiWide, well-maintained roads
3Wide, well-maintained roadsConvenient access
4Tasty tap waterNagoya meshi
5History of the three unifiersActive economic activity

Top-5 agreement: 60% (3/5)

It identified Nagoya meshi, road maintenance, and convenient access, capturing even the locally specific “Nagoya meshi” in the top ranks.

Q2: Worries and anxieties (Q11, 11 options)

Grade B+

QuestionChoose the single thing you feel most worried or anxious about.

Options(1) Own health (2) Family health (3) Own life issues (education, jobs, marriage, childcare) (4) Family life issues (5) Current income/assets (6) Outlook for future income/assets (7) Retirement life planning (8) Relationships among family/relatives (9) Relationships with neighbors/community (10) Work and relationships at the workplace (11) Business management issues

RankReal survey top 5PersonaChoice top 5 (Top-K agreement)Match
1Own healthRetirement life planning
2Retirement life planningOwn health
3Family health / income outlookFuture income/asset outlook
4(tied for 3rd)Current income/assets
5Current income/assetsOwn life issues

Top-5 agreement: 80% (4/5)

In the real survey, “own health” (61.2%) and “retirement” (57.5%) were close. The members of the top 5 matched 4/5.

Overall results across all 8 questions

#CityQuestionTop-1Top-K agreementGrade
1YokohamaWorriesTop-5: 80%A
2YokohamaGlobal skillsTop-3: 100%A
3FukuokaSceneryTop-3: 67%B
4FukuokaPark expectationsTop-5: 60%B+
5SapporoPriority PR themesTop-5: 60%B+
6SapporoJob conditionsTop-5: 60%B
7NagoyaCity prideTop-5: 60%B
8NagoyaWorries & anxietiesTop-5: 80%B+

Grade distribution: A × 2 / B+ × 3 / B × 3 (B or above: 8/8 = 100%, B+ or above: 5/8 = 62.5%)

The agreement rate for top items averages 75% overall, reaching 100% in the best case (Yokohama, global skills — exact match).

Question types PersonaChoice is suited for — and less suited for

Well-suited patterns

  • “Choose the single most ___” type
    Choosing the most important from multiple options
  • Preference-split questions with no right answer
    Questions where answers split by taste or values
  • Naturally split by attributes
    Themes where age/gender/income change the choice
  • Region-specific themes
    Nagoya meshi, the three unifiers, summer heat, etc.

Less-suited patterns

  • ×Evaluating residents’ personal reality
    Satisfaction with family relationships, etc., where AI has no “reality”
  • ×Behavior volume, awareness, screening status
    “Did you read the PR magazine?” “Did you get a check-up?” etc.
  • ×Questions with a “correct” answer
    Satisfaction, attachment, intent to stay — where a socially desirable answer is obvious

Conclusion

PersonaChoice accuracy for resident opinion surveys: 70–80%

75%

Top-item agreement
(overall average)

100%

Agreement
(best case)

100%

Rated B or above (8/8)

For resident opinion surveys, it was demonstrated that PersonaChoice can correctly identify 70–80% of residents’ top concerns. Top-item agreement averages 75% overall, 100% in the best case, with a grade of B or above on all 8 questions.
For product preference and marketing surveys, there is an accuracy track record of 80–90%.

Where PersonaChoice fits

PersonaChoice is not a replacement for real surveys — it is a complement.
Whereas a real survey costs millions to tens of millions of yen and takes months,
PersonaChoice delivers, at low cost, with same-day turnaround, results that reproduce a real survey’s top concerns with a high degree of agreement.

Why not try it for free first?

Experience a resident-opinion simulation right now with 3,000 bonus points (free) for first-time sign-ups.