Structuring the Space of Sociotechnical Alignment
Not on the map. Why?
Abstract (arXiv)
Sociotechnical alignment concerns the social desirability of AI behavior and is thus inherently normative, not merely technical. While NLP research increasingly addresses its technical aspects, it often leaves underspecified what such "social desirability" entails. We argue that this reflects a fundamental gap: the absence of a systematic way to specify how sociotechnical alignment defines, justifies, and evaluates socially desirable AI behavior. To address this gap, we introduce a human-centered framework for specifying sociotechnical alignment. We draw on social-scientific accounts of sociobehavioral desirability to ground the basis for behavioral desirability judgments and use this framework to analyze how alignment is specified in practice. Our systematic literature review identifies recurring patterns: normative concepts grounding desirability judgments are often unspecified or conflated with alignment targets for (desired) system behavior, target populations are underdefined, and design choices are rarely theoretically justified. These findings point to a lack of conceptual specificity that limits cumulative progress. We therefore offer recommendations that link social-scientific frameworks to alignment design choices, supporting more conceptually precise approaches to sociotechnical alignment.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 181 of its 305 references that are in the map. First posted on arXiv 2 Jun 2026.
Builds on
181 of 305 references.
References · 181
- 2020Artificial Intelligence, Values and Alignmentcited by 443
- 2024PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preferencecited by 252
- 2023Defending Against Alignment-Breaking Attacks via Robustly Aligned LLMcited by 173
- 2023Removing RLHF Protections in GPT-4 via Fine-Tuningcited by 165
- 2022Probing Pre-Trained Language Models for Cross-Cultural Differences in Valuescited by 147
- 2024Arithmetic Control of LLMs for Diverse User Preferences: Directional Preference Alignment with Multi-Objective Rewardscited by 145
- 2020Moral Stories: Situated Reasoning about Norms, Intents, Actions, and their Consequencescited by 133
- 2024Direct Preference Optimization with an Offsetcited by 122
- 2023Cultural Alignment in Large Language Models: An Explanatory Analysis Based on Hofstede's Cultural Dimensionscited by 111
- 2023SteerLM: Attribute Conditioned SFT as an (User-Steerable) Alternative to RLHFcited by 107
- 2024Controllable Preference Optimization: Toward Controllable Multi-Objective Alignmentcited by 105
- 2023SeaLLMs -- Large Language Models for Southeast Asiacited by 104
38 more
- 2024Systematic Biases in LLM Simulations of Debatescited by 98
- 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contextscited by 92
- 2023Moral Foundations of Large Language Modelscited by 84
- 2024Evaluating Large Language Model Biases in Persona-Steered Generationcited by 78
- 2024Defending Large Language Models Against Jailbreak Attacks via Layer-specific Editingcited by 77
- 2023AceGPT, Localizing Large Language Models in Arabiccited by 75
- 2023Aligning Large Language Models with Human Preferences through Representation Engineeringcited by 71
- 2024PERSONA: A Reproducible Testbed for Pluralistic Alignmentcited by 71
- 2024Hidden Persuaders: LLMs' Political Leaning and Their Influence on Voterscited by 71
- 2024CodeAttack: Revealing Safety Generalization Challenges of Large Language Models via Code Completioncited by 70
- 2024The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harmcited by 67
- 2024Unintended Impacts of LLM Alignment on Global Representationcited by 66
- 2023Ethical Reasoning over Moral Alignment: A Case and Framework for In-Context Ethical Policies in LLMscited by 64
- 2024Intention Analysis Makes LLMs A Good Jailbreak Defendercited by 64
- 2024Aligning LLMs with Individual Preferences via Interactioncited by 63
- 2024Benchmarking Distributional Alignment of Large Language Modelscited by 58
- 2024One fish, two fish, but not the whole sea: Alignment reduces language models' conceptual diversitycited by 57
- 2023Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generationcited by 55
- 2024Beyond prompt brittleness: Evaluating the reliability and consistency of political worldviews in LLMscited by 55
- 2023The Past, Present and Better Future of Feedback Learning in Large Language Models for Subjective Human Preferences and Valuescited by 53
- 2024WPO: Enhancing RLHF with Weighted Preference Optimizationcited by 53
- 2024Context-DPO: Aligning Language Models for Context-Faithfulnesscited by 50
- 2023Value FULCRA: Mapping Large Language Models to the Multidimensional Spectrum of Basic Human Valuescited by 49
- 2024Ethical Reasoning and Moral Value Alignment of LLMs Depend on the Language we Prompt them incited by 45
- 2024Learning to Trust Your Feelings: Leveraging Self-awareness in LLMs for Hallucination Mitigationcited by 43
- 2024When to Trust LLMs: Aligning Confidence with Response Qualitycited by 42
- 2024RLHF Can Speak Many Languages: Unlocking Multilingual Preference Optimization for LLMscited by 42
- 2024Distract Large Language Models for Automatic Jailbreak Attackcited by 40
- 2025A Survey on Personalized Alignment -- The Missing Piece for Large Language Models in Real-World Applicationscited by 40
- 2025Know You First and Be You Better: Modeling Human-Like User Simulators via Implicit Profilescited by 38
- 2023Stealthy and Persistent Unalignment on Large Language Models via Backdoor Injectionscited by 37
- 2023Flames: Benchmarking Value Alignment of LLMs in Chinesecited by 35
- 2021The Challenge of Value Alignment: from Fairer Algorithms to AI Safetycited by 34
- 2024Can Language Models Reason about Individualistic Human Values and Preferences?cited by 34
- 2024Gradient-Based Language Model Red Teamingcited by 33
- 2024KorNAT: LLM Alignment Benchmark for Korean Social Values and Common Knowledgecited by 33
- 2022Aligning to Social Norms and Values in Interactive Narrativescited by 32
- 2024Evidence-Driven Retrieval Augmented Response Generation for Online Misinformationcited by 32
50 most cited of 181 All 181 (JSON)
Built on by
Citing papers per year: 2026: 2.
2 papers.