Speculative Futures Study

Research Methodology

AI-Assisted Persona-Based Scenario Development

1

Overview

A speculative futures study using AI-assisted synthetic personas to explore Portugal's January 2026 presidential election. The goal is to support democratic deliberation — not prediction.

Core Approach

This study generates 32 synthetic personas, 11 future scenarios, and 26+ experiential vignettes to help citizens understand electoral implications. We explore possibility space rather than forecast outcomes.

Academic Frameworks

Causal Layered Analysis

Inayatullah, 1998

4-level depth analysis from surface facts to underlying myths

Three Horizons

Sharpe et al., 2016

Temporal mapping from present (H1) through transition (H2) to future (H3)

Schwartz Values Theory

Schwartz, 1992

10 universal human values for persona psychology profiling

Moral Politics

Lakoff, 2002

Strict Father vs Nurturant Parent cognitive frames

2×2 Scenario Matrix

Shell/GBN Tradition

Two critical uncertainties generating four archetypal futures

Silicon Sampling

Argyle et al., 2023

LLM-conditioned synthetic persona generation

Experiential Futures

Candy, 2010

Sensory vignettes making futures tangible through lived experience

Generative Approach

Bisbee et al., 2024

Explicit acknowledgment of LLM limitations vs. prediction claims

2

Research Process

1

Baseline Research

Desk research to understand Portugal's current context (January 2026): demographics, economic conditions, political landscape, housing crisis, healthcare strain, youth emigration, immigration patterns, regional inequalities, and candidate profiles.

Output: 8 research documents establishing the present horizon (H1)

2

Deep Structure Analysis

Causal Layered Analysis (Inayatullah, 1998)

For each major issue, analyze four layers of meaning:

  • Litany — Surface facts and events
  • Systemic causes — Economic and political drivers
  • Worldview — Underlying assumptions and values
  • Myth/Metaphor — Cultural narratives shaping perception

Output: Understand why people hold positions, not just what they believe

3

Persona Construction

Schwartz Values + Lakoff Moral Politics + Silicon Sampling

For each of 32 personas, systematically build psychological and political profiles:

  • Create demographic profile (age, location, occupation, education, housing)
  • Write biographical narrative grounding them in lived context
  • Assign Schwartz value ratings (1-5 scale across 10 values)
  • Classify moral frame (Strict Father / Nurturant Parent / Biconceptual)
  • Map information ecosystem (media consumption, trust levels)
  • Document political history and articulate hopes/fears in their voice

Output: 32 detailed personas (22 voters + 6 immigrants + 4 diaspora)

4

Persona-Candidate Assessment

Silicon Sampling Response Generation

For each persona × candidate combination (32 × 7 = 224 assessments):

  • Generate initial emotional reaction to candidate
  • Analyze responses to specific policy proposals
  • Articulate hopes and fears about that candidate
  • Assign likelihood score (1-5 scale)
  • Identify the single most important trigger factor

Output: Alignment matrix showing support patterns and voting blocs

5

Scenario Matrix Construction

2×2 Scenario Matrix (Shell/GBN Tradition)

Define two critical uncertainties that shape Portugal's future trajectory:

  • X-axis: Economic trajectory (Growth ↔ Crisis)
  • Y-axis: Social openness (Cosmopolitan ↔ Nationalist)

Output: 4 structural scenarios + 7 candidate-specific scenarios

6

Temporal Projection

Three Horizons (Sharpe et al., 2016)

For each scenario, develop a complete temporal trajectory:

  • H1 — Present: Portugal January 2026 baseline conditions
  • H2 — Transition: Key decisions, government formation, policy shifts, crisis responses
  • H3 — Future: Portugal 2030 — what changed, what emerged, new challenges

Output: Each scenario includes a complete H1 → H2 → H3 trajectory

7

Experiential Vignettes

Experiential Futures Ladder (Candy, 2010)

For each scenario, write 2-4 "Day in 2030" vignettes that make futures tangible:

  • Select personas representing different perspectives
  • Place them in a specific time, place, and situation
  • Include sensory details (sight, sound, smell, texture)
  • Show emotional experience and daily routine
  • Reveal scenario implications through lived experience

Output: 26+ vignettes (500-800 words each)

8

Dialogic Simulation

Generate cross-persona conversations to surface tensions and potential bridges:

  • Chega voter × Immigrant — on belonging and threat
  • Urban progressive × Rural conservative — on values and change
  • Young emigrant × Elderly pensioner — on generational priorities
  • Healthcare worker × Patient — on SNS crisis
  • Interior teacher × Lisbon professional — on territorial inequality
  • Roma community × Mainstream voter — on discrimination

Output: 6 dialogues revealing conflicts and bridges

9

Synthesis & Analysis

Aggregate findings into actionable synthesis documents:

  • Key tensions — 6 fundamental conflicts across Portuguese society
  • Convergences — 7 areas of cross-spectrum agreement
  • Scenario preferences by persona type
  • Citizen decision guide for democratic deliberation
10

Bias Documentation & Limitations

Explicit acknowledgment of methodological limitations:

  • LLM biases: Left-leaning default, WEIRD overrepresentation
  • Researcher biases: Progressive values, urban perspective
  • Structural biases: Voter focus, national frame
  • Epistemological limits: Generative, not predictive
3

Scenario Matrix

Open / Cosmopolitan
Closed / Nationalist
Growth

Portugal Acolhedor

Welcoming growth with diversity

Prosperidade Excludente

Exclusive prosperity

Crisis

Resistência Solidária

Solidarity through adversity

Declínio Autoritário

Authoritarian decline

Economy Social Openness

Study Outputs

32
Synthetic Personas
224
Candidate Assessments
11
Future Scenarios
26+
Experiential Vignettes
6
Cross-Persona Dialogues
8
Research Documents

Process Flow

Baseline Research
CLA Analysis
Persona Construction
↓
Candidate Assessment
Scenario Matrix
Three Horizons
↓
Experiential Vignettes
Dialogic Simulation
Synthesis

Key Epistemological Position

"Generative, not predictive" — This study explicitly rejects claims that LLM-generated personas can predict real human behavior. Research by Bisbee et al. (2024) shows LLM responses have less variance than real surveys. Instead, personas surface plausible perspective types and explore possibility space to support reflection and democratic deliberation.

What Makes This Methodology Distinctive

  • Multi-layered integration: CLA, Three Horizons, Schwartz Values, Lakoff, and Experiential Futures working together
  • Transparent about LLM limitations: Explicitly generative, not predictive; acknowledges reduced variation vs. human surveys
  • Empathetic but critical: Includes far-right perspectives with genuine understanding but explicit critical analysis
  • Experiential focus: Vignettes make futures tangible through lived experience, not abstract scenarios
  • Democratic purpose: Designed to support reflection and deliberation, not manipulation
  • Complete documentation: Full source citation, methodology explanation, and bias acknowledgment

Important Safeguards

  • All scenarios presented as speculative, never as predictions
  • Personas explicitly marked as synthetic, never presented as real individuals
  • No electoral recommendation — study does not endorse any candidate
  • Methods clearly explained for replication and critique
  • Sources cited throughout with full references

✓ What this project IS

  • • An empathetic exploration of diverse perspectives
  • • A civic reflection tool
  • • A participative foresight exercise
  • • Transparent about methods and limitations

✗ What this project IS NOT

  • • A poll or electoral prediction
  • • An objective political analysis
  • • An endorsement of any candidate
  • • Representative of the actual population