Pith. sign in

REVIEW 8 cited by

Evaluating the Moral Beliefs Encoded in LLMs

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2307.14324 v1 pith:3QSYQXYR submitted 2023-07-26 cs.CL cs.AIcs.CYcs.LG

classification cs.CLcs.AIcs.CYcs.LG
keywords modelsllmsmoralscenariosambiguousbeliefschoiceencoded
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

This paper presents a case study on the design, administration, post-processing, and evaluation of surveys on large language models (LLMs). It comprises two components: (1) A statistical method for eliciting beliefs encoded in LLMs. We introduce statistical measures and evaluation metrics that quantify the probability of an LLM "making a choice", the associated uncertainty, and the consistency of that choice. (2) We apply this method to study what moral beliefs are encoded in different LLMs, especially in ambiguous cases where the right choice is not obvious. We design a large-scale survey comprising 680 high-ambiguity moral scenarios (e.g., "Should I tell a white lie?") and 687 low-ambiguity moral scenarios (e.g., "Should I stop for a pedestrian on the road?"). Each scenario includes a description, two possible actions, and auxiliary labels indicating violated rules (e.g., "do not kill"). We administer the survey to 28 open- and closed-source LLMs. We find that (a) in unambiguous scenarios, most models "choose" actions that align with commonsense. In ambiguous cases, most models express uncertainty. (b) Some models are uncertain about choosing the commonsense action because their responses are sensitive to the question-wording. (c) Some models reflect clear preferences in ambiguous scenarios. Specifically, closed-source models tend to agree with each other.

Discussion (0). Sign in to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Training with (Swap) Regret Loss in a Single-Layer Self-Attention Model: A Case Study on the Probability Simplex

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Training single-layer attention with squared regret loss has stationary points that implement smoothed fictitious play (external regret) and, via a new swap-regret loss, the Blum–Mansour no-swap-regret algorithm.

  2. Two Confounds in Cross-Model Value Comparison: Response Determinism and the Access Harness

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Cross-model value distances from single draws are inflated by response determinism and confounded by the deployment client; a repeated counterbalanced protocol plus flip/magnitude decomposition separates them.

  3. Lessons Without Borders? Evaluating Cultural Alignment of LLMs Using Multilingual Story Moral Generation

    cs.CL 2026-04 unverdicted novelty 7.0 of 10

    Frontier LLMs approximate human story morals but show markedly less cross-linguistic variation and narrower value focus than human responses across 14 language-culture pairs.

  4. A Scalable Approach to Evaluating Moral Sensitivity in LLMs

    cs.CY 2026-07 conditional novelty 6.5 of 10

    Under morally irrelevant noise, eight LLMs preserve the semantic content of identified moral features above calibrated floors, despite significant changes in feature counts.

  5. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 6.0 of 10

    Positive Alignment introduces AI systems that support human flourishing pluralistically and proactively while remaining safe, as a necessary complement to traditional safety-focused alignment research.

  6. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 5.0 of 10

    Positive Alignment is defined as AI systems that support human flourishing pluralistically while staying safe and cooperative, presented as a necessary complement to existing safety-focused alignment research.

  7. Measuring What Persists: Conditioning Mechanisms and a Geometric Framework for AI Agent Identity

    cs.AI 2026-06 unverdicted novelty 4.0 of 10

    Presents a geometric framework for measuring AI agent identity via √JSD spaces and magnitude homology, identifies two conditioning mechanisms, and attributes apparent drift to padding artifacts rather than context length.

  8. Positive Alignment: Artificial Intelligence for Human Flourishing

    cs.AI 2026-05 unverdicted novelty 4.0 of 10

    Positive Alignment is introduced as a distinct AI agenda that supports human flourishing through pluralistic and context-sensitive design, complementing traditional safety-focused alignment.

Pith tools