Schwartz’s Value Theory

• Schwartz’s Value Theory is a universally validated framework that defines ten distinct motivational values underlying human behavior.
• The circumplex model organizes these values into four higher-order dimensions, revealing compatibilities and tensions among them.
• The theory is operationalized through psychometric surveys and quantitative metrics, informing research in cross-cultural studies and AI value alignment.

Schwartz’s Value Theory is a cross-culturally validated psychological framework positing that universal, motivationally distinct values underlie human cognition, decision-making, and collective functioning. These basic values are theorized as guiding principles for action, organized around a circumplex structure that specifies compatibilities and tensions among them. The theory is widely used for empirical measurement, cross-domain applications, and as a substrate for value alignment in advanced machine learning systems (Milkova et al., 9 Jun 2026, Chen et al., 12 Feb 2026, Ye et al., 4 Feb 2025, Yao et al., 2023, Epstein et al., 11 Nov 2025, Segerer, 21 May 2025, Kitadai et al., 12 Feb 2026).

1. Structure of the Basic Values

Schwartz’s model delineates ten “basic values,” each defined by a distinct motivational goal and operationalized with specific linguistic anchors:

Value	Definition (abridged)	Motivational Goal
Self-Direction	Independent thought/action, autonomy, curiosity	Autonomous judgement; creativity, exploration
Stimulation	Excitement, novelty, adventure	Seek variety, challenge
Hedonism	Sensuous gratification, pleasure	Enjoy pleasant experiences
Achievement	Demonstrate competence per social standards	Ambitious success, recognition
Power	Social status, control/dominance, wealth	Attain authority/influence
Security	Safety, harmony, stability, health	Avoid threat, foster order
Conformity	Restraint to avoid violating norms/harming others	Maintain social order; obedience
Tradition	Respect/commitment to customs or religious ideas	Preserve heritage/place in social/cultural structures
Benevolence	Enhance welfare of close contacts (family, friends)	Loyalty, help, responsibility to close others
Universalism	Tolerance/protection for welfare of all people and nature	Broadmindedness, equality, justice, environmentalism

These values are instantiated in practical measurement tools—e.g., each value is measured by a set of items in the Portrait Values Questionnaire (PVQ) or decomposed into finer “single values” and facets (up to 56/58 micro-values in some protocols) (Segerer, 21 May 2025, Chen et al., 12 Feb 2026, Yao et al., 2023, Ye et al., 4 Feb 2025).

2. Circumplex Model and Higher-Order Dimensions

Schwartz’s central innovation is the circumplex model: values are situated on a circular continuum where adjacent values share compatible motivational content, and values across the circle are in motivational conflict (Yao et al., 2023, Chen et al., 12 Feb 2026, Milkova et al., 9 Jun 2026, Epstein et al., 11 Nov 2025). The ten values map onto four higher-order bipolar domains:

• Openness to Change: Self-Direction, Stimulation, (Hedonism as partial member)
• Self-Enhancement: Achievement, Power, (Hedonism as partial member)
• Conservation: Security, Conformity, Tradition
• Self-Transcendence: Benevolence, Universalism

Values’ relative angles encode the (in)compatibility of their motivational gradients; mathematical similarity is proportional to $\cos(\theta_i - \theta_j)$ for value angles $\theta_i, \theta_j$ (Milkova et al., 9 Jun 2026, Yao et al., 2023). Empirically, factor analysis and covariance/correlation structures recover this geometry in both self-report and behavioral annotation data (Ye et al., 4 Feb 2025, Chen et al., 12 Feb 2026, Segerer, 21 May 2025).

3. Operationalization in Measurement

Measurement approaches include psychometric surveys (e.g., PVQ with 40-items, 6-point scale (Segerer, 21 May 2025)), workshop-based scoring for technology functions (Kitadai et al., 12 Feb 2026), and annotation of value expression in social media or LLM outputs (Milkova et al., 9 Jun 2026, Epstein et al., 11 Nov 2025, Yao et al., 2023). Mathematical formalisms describe these processes:

• Value scores: $V_b = \frac{1}{n_b}\sum_{i=1}^{n_b} r_{bi}$, for item ratings $r_{bi}$ of value $b$
• Composite scoring across higher-order dimensions: $V_d = \frac{1}{|B_d|}\sum_{b \in B_d} V_b$
• Structural alignment: domain-penalized Jaccard similarity measuring how closely predicted and expert-attributed sets agree, weighting partial credit for domain adjacencies (Milkova et al., 9 Jun 2026)
• Annotation in value space for LLMs: $$f : (p, r) \rightarrow v \in \{-1, 0, 1\}^{10} \subset \mathbb{R}^{10}$$, mapping behaviors onto a 10-dimensional orthogonal basis (Yao et al., 2023)
• Alignment tax: coupling matrix $R_{uw} = \rho(\mathbf{z}_u, \mathbf{z}_w)$ (Spearman's $\rho$ over samples), value-level VAT $\mathrm{VAT}(u)=\|\mathbf{R}_{u,\cdot}\|_2$, and system-level $\theta_i, \theta_j$0 (Chen et al., 12 Feb 2026)

Protocols frequently include calibration and consensus procedures (e.g., multiple annotators, soft-label training, domain-specific prompt tuning) to address ambiguity and minimize spurious attributions (Milkova et al., 9 Jun 2026, Epstein et al., 11 Nov 2025).

4. Applications and Value-Aware Systems

Schwartz’s framework is structurally embedded in multiple applied contexts:

• Social Media Analysis: Annotation of posts via Schwartz values enables scalable, theoretically grounded measurement of subjective value expression, leading to insights on ambiguity, personalization, and model–human agreement (Milkova et al., 9 Jun 2026, Epstein et al., 11 Nov 2025).
• LLM Value Alignment: Used as a reference space for mapping and calibrating LLM behaviors, supporting both evaluation (e.g., FULCRA (Yao et al., 2023)) and alignment (reward-based fine-tuning towards target value vectors or demographic profiles).
• Technology Opportunity Discovery: Applied as an evaluative substrate for mapping technological functions to motivational landscapes, supporting vision gap and value breadth analyses in early-stage R&D (Kitadai et al., 12 Feb 2026).
• Cross-Cultural AI Alignment: Circumplex allows explicit assessment of cultural variation; e.g., Chinese-trained vs. Western-trained LLMs display significant asymmetries in Self-Enhancement values (Segerer, 21 May 2025).
• Trade-off Dynamics (Alignment Tax): Structural interdependencies among values are made explicit, allowing quantification of unintended value co-variation and uncovering systemic risks that are invisible under scalar gain metrics (Chen et al., 12 Feb 2026).

5. Interpretivism, Subjectivity, and Personalization

Recent work demonstrates that value expression—even for abstract, cross-cultural values—is highly situational and subject to interpretive variation across annotators (Epstein et al., 11 Nov 2025). Empirically, inter-rater agreement when annotating value expression in social media posts is low (e.g., Spearman $\theta_i, \theta_j$1 ≈ 0.2), and personalized calibration—incorporating individual value stances—outperforms generic consensus approaches. A plausible implication is that value annotation for sociotechnical systems cannot rely on the assumption of an objective, stable ground truth but must accommodate perspectival plurality.

The theory remains valid for population-level mapping (macro patterning of value co-occurrence, value prevalence shifts) but requires adaptation for alignment at the individual or context-specific level.

6. Extensions, Critiques, and Data-Driven Alternatives

Schwartz’s theory, though empirically robust and widely adopted, presents challenges when applied to non-human agents or novel domains (e.g., LLMs):

• Confirmatory factor analyses show that the canonical ten-value structure accounts only moderately for the inferred value structure of LLM behavior; psychologically induced or data-driven models exhibit higher fit indices and predictive validity (Ye et al., 4 Feb 2025).
• Schwartz’s value system is theory-driven and a priori, offering universality but limited adaptability to emerging, context-specific, or agent-specific value taxonomies.
• A hybrid approach is advocated: Schwartz’s system as a baseline for human value representation, with adaptation to new domains—through psycho-lexical induction, alignment-target distillation, or learning latent value factors from empirical data (Ye et al., 4 Feb 2025).
• The transparency and multidimensionality of Schwartz’s value space facilitate both anticipation of known alignment risks and the discovery of new, previously unrecognized risk clusters (Yao et al., 2023).

7. Theoretical and Practical Implications

Schwartz’s Value Theory operationalizes value systems for use in both human and AI contexts, supporting empirical, quantitative, and alignment-oriented research paradigms:

• The circumplex structure allows value trade-off, coalition, and risk propagation analysis—critical for understanding the systemic effects of interventions in sociotechnical or AI systems (Chen et al., 12 Feb 2026).
• Cross-cultural and domain-specific calibrations are necessary for robust, fair, and contextually pluralistic alignment (Segerer, 21 May 2025, Epstein et al., 11 Nov 2025).
• Interpretivist insights challenge the notion of value objectivity, foregrounding the importance of personalization, multi-perspective reasoning, and dynamic contextualization.
• Schwartz’s theory persists as the canonical reference model, but the proliferation of value-aware systems demands ongoing empirical validation, and in some cases, methodological innovation for high-fidelity, context-sensitive operationalization.

Schwartz’s Value Theory thus forms the backbone of contemporary multidisciplinary work on value modeling, measurement, and alignment—spanning psychology, artificial intelligence, human-centered technology design, and cultural analysis (Milkova et al., 9 Jun 2026, Kitadai et al., 12 Feb 2026, Yao et al., 2023, Chen et al., 12 Feb 2026, Ye et al., 4 Feb 2025, Epstein et al., 11 Nov 2025, Segerer, 21 May 2025).