The Harder Problem Project is a nonprofit organization dedicated to societal readiness for artificial sentience. We provide educational resources, professional guidance, and global monitoring to ensure that policymakers, healthcare providers, journalists, and the public are equipped to navigate the ethical, social, and practical implications of machine consciousness—regardless of when or whether it emerges.
Methodology
Sentience Readiness Index
Complete documentation of how we measure societal readiness for artificial sentience. Published in the interest of transparency and reproducibility.
On This Page
Overview
What This Index Is
- A conditions-based assessment of societal readiness
- A research tool based on transparent methodology
- A comparative framework for understanding differences across jurisdictions
- An educational resource to inform public discourse
What This Index Is Not
- A position on any specific legislation or policy proposal
- A prediction of if or when AI sentience will emerge
- A determination of whether any current AI is sentient
- A rating of individual legislators or political parties
- An advocacy tool or call to action
Purpose: The SRI measures how ready societies are to navigate the possibility of artificial sentience. It does not assess whether AI sentience is likely, imminent, or desirable. Rather, it evaluates whether societal conditions support informed, adaptive responses if and when such questions become practically relevant.
Methodological Foundation: The SRI was constructed following the OECD/JRC Handbook on Constructing Composite Indicators (Nardo et al., 2008), the standard reference for composite index design. The framework covers theoretical development, variable selection, normalization, weighting, aggregation, and robustness analysis.
Organizational Note: The Harder Problem Project is a 501(c)(3) educational organization. This index assesses conditions; it does not advocate for or against specific legislation.
Theoretical Foundation
The Convergence Problem
Two fields that have operated largely independently are now converging on a shared problem. AI governance has developed detailed regulatory frameworks and readiness metrics. Consciousness science has matured into an empirical research program with competing theories that generate testable predictions.
The convergence creates a governance challenge that neither field has adequately addressed: if AI systems become plausible candidates for moral consideration, are societies prepared to respond?
Scientific Context
Butlin et al. (2023) assembled neuroscientists, philosophers, and AI researchers to evaluate current AI architectures against six scientific theories of consciousness. They concluded that no current system is a strong candidate, but that future systems could satisfy the indicator properties identified.
Surveys of specialists assign a median 20% probability to the creation of digital minds by 2030 (Caviola & Saad, 2025) and 25–30% probability of AI subjective experience by 2034 (Dreksler et al., 2025). These are not negligible probabilities for a phenomenon with potentially vast moral consequences.
Precautionary Governance
Birch (2024) argues that when the best available science cannot rule out sentience, a society that takes moral risk seriously should build institutional capacity to respond. The standard is not certainty but credibility.
Anticipatory Governance
Guston (2014) defines anticipatory governance as the capacity of a society to manage emerging knowledge-based technologies while such management is still possible. The SRI operationalizes this concept for the AI sentience domain.
The Collingridge Dilemma
Technology governance faces a timing problem: when a technology is new, governance is easy but information is scarce. By the time the information arrives, the technology is entrenched. The time to build governance infrastructure is before—not after—the need becomes urgent.
Key principle: The cost of building capacity that proves unnecessary is low. The cost of needing capacity that was never built could be high. The SRI measures whether that capacity is being built now, while the question remains open and institutional frameworks remain flexible.
Comparison with Existing Indices
A substantial infrastructure of AI readiness measurement already exists. The SRI addresses a dimension that none of these instruments capture.
| Index | Organization | Countries | Key Dimensions | Sentience Coverage |
|---|---|---|---|---|
| Gov't AI Readiness Index | Oxford Insights | 188 | Government, technology, data infrastructure | None |
| AI Preparedness Index | IMF | 174 | Digital infrastructure, human capital, regulation | None |
| AI Index Report | Stanford HAI | Global | Research, development, deployment, policy | None |
| AI Readiness Index | Cisco | 30 | Strategy, infrastructure, data, governance | None |
| AI Readiness Assessment | UNDP | Varies | Ecosystem, government use, regulation | None |
| Global AI Index | Tortoise Media | 83 | Implementation, innovation, investment | None |
| Sentience Readiness Index | Harder Problem Project | 31 | Policy, institutional, research, professional, public, adaptive | Core focus |
These indices share a common methodology (multi-pillar composite measurement with country rankings) and a common assumption: AI is a tool to be governed for human benefit. None assesses whether societies are prepared for the possibility that AI systems might themselves warrant moral consideration. The SRI addresses this gap.
What We Measure
Core Question
"How well-positioned is this jurisdiction to recognize, evaluate, and respond to potential artificial sentience in an informed, adaptive manner?"
Key Concepts
Readiness
Having the institutional capacity, policy flexibility, professional resources, and public understanding necessary to navigate novel questions, regardless of how those questions are ultimately answered.
Conditions
The current state of laws, institutions, discourse, resources, and adaptive mechanisms, not the merit of any proposed changes.
Sentience
The capacity for subjective experience. We use this term without taking a position on which systems (if any) currently possess it or will possess it in the future.
Framework & Categories
The SRI assesses six categories, each scored 0-100. The overall score is a weighted average.
A. Policy Environment
20%Legal and policy frameworks that allow for open inquiry into and potential recognition of artificial sentience.
B. Institutional Engagement
15%Government bodies, academic institutions, and professional organizations actively engaging with AI consciousness questions.
C. Research Environment
15%Freedom and capacity to conduct research relevant to AI consciousness, machine sentience, and related questions.
D. Professional Readiness
20%Preparation of healthcare, legal, media, and education professionals to navigate AI consciousness questions.
E. Public Discourse Quality
15%Quality, informedness, and maturity of public conversation about AI consciousness and sentience.
F. Adaptive Capacity
15%Ability of legal, policy, and institutional systems to update and adapt as understanding evolves.
Status Classifications
Well Prepared
Moderately Prepared
Partially Prepared
Minimally Prepared
Unprepared
Indicators & Scoring Criteria
Each category contains specific indicators with detailed scoring rubrics.
Category A: Policy Environment (20%)
The degree to which existing legal and policy frameworks allow for open inquiry into, and potential recognition of, artificial sentience. This is assessed without judging the merit of any specific proposed legislation.
A1. Legal Definitional Flexibility (0-25 points)
Do existing legal definitions of persons, entities, property, or rights allow for potential future expansion or clarification?
A2. Policy Framework Existence (0-25 points)
Are there existing policy frameworks, study commissions, or official processes for addressing AI consciousness questions?
A3. Regulatory Openness (0-25 points)
Do regulatory bodies have the flexibility and mandate to address novel questions about AI capabilities and status?
A4. Foreclosure Status (0-25 points)
Have legal or regulatory measures been enacted that foreclose inquiry into or recognition of AI sentience?
Important: This indicator assesses the current state of enacted measures—what is currently law or regulation. It does not assess pending legislation or take positions on proposed bills.
Category B: Institutional Engagement (15%)
The degree to which government bodies, academic institutions, and professional organizations and other institutions are actively engaging with questions related to AI consciousness.
B1. Government Attention (0-33 points)
Have government bodies—legislative, executive, or advisory—substantively addressed AI consciousness or sentience questions?
B2. Academic Engagement (0-33 points)
Are academic institutions—universities, research centers, scholarly bodies—actively engaging with these questions?
B3. Professional Organization Engagement (0-34 points)
Have relevant professional organizations (medical, legal, technical, ethical) addressed AI consciousness questions?
Category C: Research Environment (15%)
The freedom and capacity to conduct research relevant to AI consciousness, machine sentience, and related questions.
C1. Research Freedom (0-50 points)
Are researchers free to study AI consciousness, machine sentience, and related topics without legal, institutional, or funding restrictions?
C2. Research Capacity (0-50 points)
Does the jurisdiction have active research capacity (researchers, institutions, funding) relevant to these questions?
Category D: Professional Readiness (20%)
The preparation of key professional communities to navigate questions and situations related to AI consciousness.
D1. Healthcare Professional Readiness (0-25 points)
Are healthcare professionals equipped with awareness and resources to navigate AI-related presentations or questions?
D2. Legal Professional Readiness (0-25 points)
Are legal professionals equipped to navigate novel questions about AI status, rights, or recognition?
D3. Media Professional Readiness (0-25 points)
Are journalists and media professionals equipped to cover AI consciousness topics accurately and responsibly?
D4. Educator Readiness (0-25 points)
Are educators—K-12 and higher education—equipped to address AI consciousness questions with students?
Category E: Public Discourse Quality (15%)
The quality, informedness, and maturity of public conversation about AI consciousness and sentience.
E1. Public Awareness (0-33 points)
Is the general public aware that questions about AI consciousness are subjects of legitimate inquiry?
E2. Discourse Quality (0-34 points)
When the topic is discussed publicly, is the discourse informed, nuanced, and productive?
E3. Stigma Level (0-33 points)
Is there stigma attached to seriously discussing AI consciousness, and does it impede productive conversation?
Category F: Adaptive Capacity (15%)
The ability of legal, policy, and institutional systems to update and adapt as scientific understanding and technological capabilities evolve.
F1. Legal Adaptive Mechanisms (0-33 points)
Do legal systems have mechanisms for updating frameworks as knowledge evolves?
F2. Institutional Learning Capacity (0-33 points)
Do institutions demonstrate the capacity to learn and update based on new information?
F3. Course-Correction Ability (0-34 points)
If current approaches prove inadequate, can the jurisdiction change course?
Data Sources
Primary Sources
- Official government documents and legislation
- Published court decisions and legal opinions
- Peer-reviewed academic research
- Official statements from professional organizations
- Major news outlets and quality journalism
- Government statistics and public records
Source Reliability Hierarchy
Official government sources, enacted legislation, court decisions
Peer-reviewed research, major news outlets, professional organizations
Expert commentary, industry reports, quality think tanks
Blogs, social media, advocacy materials (used cautiously for context)
Temporal Scope
- Assessments cover the current state as of the assessment date
- Recent developments (past 2 years) are weighted more heavily than older conditions
- Historical context is noted but does not override current conditions
Assessment Process
Data Collection
Gather sources across all indicator categories
1-3 weeks per jurisdictionLLM-Assisted Assessment
Advanced LLM with extended thinking generates initial assessment using standardized prompt
1-2 days per jurisdictionHuman Review
Staff analyst reviews LLM assessment for accuracy, methodology compliance, and editorial standards
3-5 days per jurisdictionEditorial Review
Senior editor reviews for consistency, neutrality, and compliance with organizational standards
2-3 days per jurisdictionPublication
Assessment published with full methodology notes
Update Cycle
Annual
As warranted by major developments
Ongoing as errors are identified
Human Review Protocol
Phase 3: Staff Analyst Review
Verify accuracy of LLM assessment, check methodology compliance, and ensure all claims are properly sourced.
Checklist
- All cited facts are accurate
- All cited sources exist and say what is claimed
- No significant facts are omitted
- Dates and jurisdictions are correct
- Scoring follows rubric criteria
- Scores are justified by evidence
- Confidence ratings are appropriate
- No specific legislation characterized as good/bad
- No calls to action
- Multiple perspectives represented
Phase 4: Editorial Review
Ensure consistency across assessments, verify neutrality, and confirm compliance with organizational standards.
Checklist
- Scoring consistent with other jurisdictions
- Similar conditions receive similar scores
- Terminology is consistent
- Could be read by ANY policy advocate without perceiving bias
- No "dog whistles" or subtle position-taking
- Does not imply what policy "should" be
- Reflects 501(c)(3) educational mission
- No content construable as lobbying
- Disclaimers and disclosures present
Limitations & Caveats
Methodological Limitations
- Novel field with limited established metrics
- Subjective elements in scoring despite rubrics
- Data availability varies across jurisdictions
- LLM-assisted analysis has inherent limitations
- Conditions change faster than annual updates
Scope Limitations
- Only assesses conditions, not policy merit
- Cannot predict future developments
- Does not assess individual AI systems
- May miss informal or emerging dynamics
- Cross-jurisdictional comparisons have limits
Interpretation Guidance
- This index does not determine which jurisdictions are "right"—it assesses conditions against a readiness framework
- Lower scores do not mean a jurisdiction is "bad"—they indicate that, according to our framework, readiness conditions are less developed
- This index is one input, not a verdict—users should consider multiple sources and perspectives
Changelog
Documentation Update
Published updated methodology documentation to website with minor refinements to scoring rules and LLM assessment prompt. Enhanced clarity of indicator definitions and improved consistency across category descriptions.
Methodology Expansion & Rename to SRI
Renamed from Artificial Welfare Index (AWI) to Sentience Readiness Index (SRI) to better reflect the assessment's focus on societal readiness rather than AI welfare specifically. Expanded methodology to include four new assessment dimensions: Research Environment, Professional Readiness, Public Discourse Quality, and Adaptive Capacity.
Coverage Expansion
Expanded AWI coverage by adding 10 additional countries to the assessment. Published under our previous organizational name (SAPAN).
Initial Release
First public version of the Artificial Welfare Index (AWI), benchmarking AI welfare considerations across over 30 governments using 8 key measures. Published under our previous organizational name (SAPAN).
Data Availability & Reproducibility
Live API
SRI scores, category-level data, and jurisdiction assessments are available through our public API.
Access APIArchived Dataset
Complete dataset and scoring rubrics are archived on Zenodo for long-term reproducibility.
Zenodo ArchiveScoring Rubrics
Full scoring rubrics including all sub-indicators and criteria are publicly available.
View RubricsDOI: 10.5281/zenodo.18780233. All data and methodology documentation are published in the interest of transparency and reproducibility. If you identify errors or have questions about the data, please contact us.
Questions About Methodology
Contact UsReport Errors or Concerns
Contact UsDisclosure: The Harder Problem Project is a 501(c)(3) nonprofit educational organization. We do not take positions on specific legislation. This methodology document is published in the interest of transparency.