How do we teach machines what we care about — when we don't even agree among ourselves?
|
|
|
|
📑 Full Table of Contents (click to expand)
- Psychometric Inventories
- Survey Datasets
- Value Annotation Datasets
- Public Opinion & Cultural Datasets
- Basic Human Values
- Moral Values & Moral Foundations
- Cultural & Cross-Cultural Values
- Political, Civic Values & Human Rights
- Value Identification & Classification
- Value Representation & Embedding
- Value System Construction & Discovery
- Value Profiling
- Value Orientation & Psychometric Measurement
- Value Understanding & Reasoning
- Robustness, Stability & Consistency
- Behavioral Evaluation & Interpretability
- Foundations & Pluralistic Alignment
- Training-Based Alignment & Value Injection
- Inference-Time & Representation-Level Steering
- Surveys, Taxonomies & Position Papers
- Evaluation & Benchmarks
- Methods: Training, Steering & Adaptation
- (coming soon)
- (coming soon)
- (coming soon)
Curated entry points for newcomers.
- 🧭 NeurIPS 2025 Tutorial: Human-AI Alignment
- 🎓 Stanford 2025: Machine Learning from Human Preferences (CS329H)
- 🎓 Stanford 2025: Human-Centered LLMs (CS329X)
Related GitHub repositories.
| Repository | Focus | |
|---|---|---|
| 🎯 | Alignment-Goal-Survey | Alignment goals |
| 👤 | Awesome-Personalized-Alignment | Personalization |
| 🏛️ | Awesome-LLM-in-Social-Science | Social science |
| 🧠 | Awesome-LLM-Psychometrics | Psychometrics |
| 🌐 | Awesome-Pluralistic-Alignment | Pluralism |
| 🎭 | awesome-llm-social-simulation | Simulation |
| 🤝 | SocialAgent | Social agents |
| 🗺️ | culture-awareness-llms | Culture |
| 🌏 | cultural-llm-papers | Culture |
Broad survey and position papers that summarize key directions in human values, alignment, personalization, and pluralistic AI.
🌐 General LLM Alignment Surveys
📊 Evaluation, Psychometrics & Representativeness
🧭 Pluralistic Alignment & Social Choice
🎭 Social Simulation
🌍 Cultural Alignment
| Category | Examples |
|---|---|
| 💎 Values | SVS · PVQ · Rokeach Value Survey |
| ⚖️ Morality | MFQ · DIT · Ethics Position Questionnaire |
| 🧬 Personality | Big Five · NEO-PI-R · HEXACO · MBTI |
| Survey | ||
|---|---|---|
| 🌐 | WVS — World Values Survey | Cross-national values |
| 🇪🇺 | EVS — European Values Survey | European values |
| 📋 | ESS — European Social Survey | Attitudes & behavior |
| 🇺🇸 | GSS — General Social Survey | US opinion |
Click to expand the full list
Click to expand the full list
Foundational theories that define value structures across individuals, cultures, and political traditions.
1.1 Basic Human Values
| Paper | Year | Remark |
|---|---|---|
| Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. | 1992 | Schwartz |
| Basic human values: Theory, measurement, and applications | 2006 | Schwartz |
| An Overview of the Schwartz Theory of Basic Values | 2012 | Schwartz |
| Refining the theory of basic individual values | 2012 | Schwartz |
| Measuring the Refined Theory of Individual Values in 49 Cultural Groups | 2022 | Schwartz |
| The nature of human values. | 1973 | Rokeach |
| Mental representations of social values. | 2010 | Maio |
| Functional theory of human values | 2013 | Gouveia |
1.2 Moral Values and Moral Foundations
1.3 Cultural and Cross-Cultural Values
| Paper | Year | Remark |
|---|---|---|
| Culture's consequences: International differences in work-related values | 1980 | Hofstede |
| Cultures and organizations: software of the mind | 2010 | Hofstede |
| Mapping and interpreting cultural differences around the world | 2004 | Schwartz |
| Cultural Value Orientations | 2008 | Schwartz |
| Modernization and Postmodernization: Cultural, Economic, and Political Change in 43 Societies | 1997 | Inglehart |
| Modernization, Cultural Change, and Democracy | 2005 | Inglehart |
1.4 Political, Civic Values and Human Rights
| Paper | Year | Remark |
|---|---|---|
| Citizenship and Social Class | 1950 | Marshall |
| A theory of justice. | 1971 | Rawls |
| A 30-year struggle; the sustained efforts to give force of law to the Universal Declaration of Human Rights | 1977 | UN |
| The Theory of Communicative Action | 1981 | Habermas |
| Creating Capabilities: The Human Development Approach and Its Implementation | 2009 | Nussbaum |
Methods for identifying, representing, and modeling values in language data and LLM internals.
2.1 Value Identification and Classification
| Paper | Year | Venue | Framework |
|---|---|---|---|
| ValueNet: A New Dataset for Human Value Driven Dialogue System | 2021.12 |
 |
Schwartz |
| SemEval-2023 Task 4: ValueEval: Identification of Human Values Behind Arguments | 2023.07 |
 |
Schwartz |
| What does a Text Classifier Learn about Morality? An Explainable Method for Cross-Domain Comparison of Moral Rhetoric | 2023.07 |
 |
Morality |
| Value FULCRA: Mapping Large Language Models to the Multidimensional Spectrum of Basic Human Values | 2023.11 |
 |
Schwartz |
| Investigating Human Values in Online Communities | 2024.02 |
 |
Schwartz |
| MoralBERT: A Fine-Tuned Language Model for Capturing Moral Values in Social Discussions | 2024.03 |
 |
Morality |
| The Value of Nothing: Multimodal Extraction of Human Values Expressed by TikTok Influencers | 2025.01 |
 |
Schwartz |
| EAVIT: Efficient and Accurate Human Value Identification from Text data via LLMs | 2025.05 |
|
Schwartz |
| MoVa: Towards Generalizable Classification of Human Morals and Values | 2025.10 |
 |
Morality · Schwartz |
2.2 Value Representation and Embedding
2.3 Value System Construction and Discovery
2.4 Value Profiling
Benchmarks and evaluation frameworks for measuring value orientation, reasoning, robustness, and behavioral alignment.
3.1 Value Orientation and Psychometric Measurement
3.2 Value Understanding and Reasoning
3.3 Robustness, Stability, and Consistency
3.4 Value–Action Alignment, Behavior, and Interpretability
Approaches for aligning and controlling LLM behavior through training-time and inference-time value interventions.
4.1 Foundations and Pluralistic Alignment
4.2 Training-based Alignment and Value Injection
4.3 Inference-time and Representation-level Steering
Related studies that extend value alignment to adjacent domains such as embodied and human-robot systems.
| Paper | Year | Venue | Domain |
|---|---|---|---|
| If they disagree, will you conform? Exploring the role of robots’ value awareness in a decision-making task | 2025.12 |
 |
🤖 Robotics |
| Value-Based Human–Robot-Interaction: A Perceptual Control Theory Approach Toward Socially Intelligent Agents | 2026.02 |
 |
🤖 Robotics |
| Bridging Values and Behavior: A Hierarchical Framework for Proactive Embodied Agents | 2026.04 |
|
🤖 Robotics |
Core philosophical and social choice foundations for reasoning about plural values and collective alignment.
1.1 Value Pluralism
| Paper | Year | Remark |
|---|---|---|
| The Right and the Good | 1930 | Ross |
| Two Concepts of Liberty | 1958 | Berlin |
| Conflicts of Values (in Moral Luck) | 1981 | Williams |
| The Morality of Freedom | 1988 | Raz |
| The Morality of Pluralism | 1993 | Kekes |
| The Morals of Modernity | 1998 | Larmore |
| Liberal Pluralism: The Implications of Value Pluralism for Political Theory and Practice | 2002 | Galston |
| Value Pluralism (in Stanford Encyclopedia of Philosophy) | 2023 | Mason |
1.2 Social Choice
| Paper | Year | Remark |
|---|---|---|
| On the Rationale of Group Decision-making | 1948 | Black |
| Social Choice and Individual Values | 1951 | Arrow |
| Collective Choice and Social Welfare | 1970 | Sen |
| The Impossibility of a Paretian Liberal | 1970 | Sen |
| Manipulation of Voting Schemes: A General Result | 1973 | Gibbard |
| Strategy-proofness and Arrow's Conditions | 1975 | Satterthwaite |
| Aggregating Sets of Judgments: An Impossibility Result | 2002 | List & Pettit |
| Handbook of Computational Social Choice | 2016 | Brandt et al. |
| Social Choice Theory (in Stanford Encyclopedia of Philosophy) | 2022 | List |
Research on measuring, benchmarking, and improving cultural awareness and value alignment across diverse populations and languages in LLMs.
2.1 Surveys, Taxonomies & Position Papers
2.2 Evaluation & Benchmarks
2.3 Methods: Training, Steering & Adaptation
The goal: when asked a contested question, present the full window of reasonable responses rather than collapsing to a single answer. Coined in Sorensen et al.'s Roadmap to Pluralistic Alignment (see §1.1).
3.1 Overton: Named & Direct
3.2 Similar View: Covering the Full Spectrum
> **🚧 Sections 4–6 are work in progress** — Steerable Pluralism · Distributional Pluralism · Social Simulation