# Follow the Norm: Accounting for Fine-Tuning and Prompt Effects on Model Rationales

arXiv: https://arxiv.org/abs/2608.13250 · HTML: https://mlsanity.com/paper/2608.13250/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 47 references that are in the map.
- First posted on arXiv 13 Aug 2026.

## Builds on (most cited first, 23 of 23)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Datasheets for Datasets](https://mlsanity.com/paper/1803.09010/) (2018)
- [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020)
- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Model Cards for Model Reporting](https://mlsanity.com/paper/1810.03993/) (2018)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Language (Technology) is Power: A Critical Survey of "Bias" in NLP](https://mlsanity.com/paper/2005.14050/) (2020)
- [Aligning AI With Shared Human Values](https://mlsanity.com/paper/2008.02275/) (2020)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://mlsanity.com/paper/2304.05335/) (2023)
- [The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity](https://mlsanity.com/paper/2506.06941/) (2025)
- [Risks from Learned Optimization in Advanced Machine Learning Systems](https://mlsanity.com/paper/1906.01820/) (2019)
- [Large Pre-trained Language Models Contain Human-like Biases of What is Right and Wrong to Do](https://mlsanity.com/paper/2103.11790/) (2021)
- [Towards Accountability for Machine Learning Datasets: Practices from Software Engineering and Infrastructure](https://mlsanity.com/paper/2010.13561/) (2020)
- [Accountability in an Algorithmic Society: Relationality, Responsibility, and Robustness in Machine Learning](https://mlsanity.com/paper/2202.05338/) (2022)
- [AI auditing: The Broken Bus on the Road to AI Accountability](https://mlsanity.com/paper/2401.14462/) (2024)
- [Normative Evaluation of Large Language Models with Everyday Moral Dilemmas](https://mlsanity.com/paper/2501.18081/) (2025)
- [Towards a Responsible AI Metrics Catalogue: A Collection of Metrics for AI Accountability](https://mlsanity.com/paper/2311.13158/) (2023)
- [Mapping Topics in 100,000 Real-life Moral Dilemmas](https://mlsanity.com/paper/2203.16762/) (2022)
- [LExT: Towards Evaluating Trustworthiness of Natural Language Explanations](https://mlsanity.com/paper/2504.06227/) (2025)
- [Unraveling the Nuances of AI Accountability: A Synthesis of Dimensions Across Disciplines](https://mlsanity.com/paper/2410.04247/) (2024)

## Abstract (arXiv)

Normative datasets are often used to train and align AI systems, but the norms they contain can function as action-guiding patterns rather than neutral moral knowledge. We propose treating the AI system as a proxy actor and test whether dataset-level norms can shift it away from its baseline safety behavior when it faces high-conflict dilemmas. We make three contributions. First, we demonstrate in controlled experiments that norm-breaking fine-tuning yields norm-divergent actions justified by self-interested rationales, suggesting a systematic shift in patterns of justification. Second, we establish a practical audit trail linking downstream justifications to upstream norms using mixed methods. Third, we show that system prompts can both suppress and elicit these patterns. We conducted experiments on three models (LLaMA-3.2-11B, Qwen-3.5-9B, and Pixtral-12B) using Low-Rank Adaptation (LoRA) fine-tuning on Social Chemistry 101 Fairness/Cheating (norm-following vs. norm-breaking) with prompt steering. Across all three models, we find that norm-breaking fine-tuning shifts the model's default rationale style from safety compliance to instrumental self-interest, whereas system prompts can override this behavior. Our results support a distributed view of alignment in which observed behavior depends jointly on training data, fine-tuning, and prompting, motivating norm-aware documentation and rationale logging for contestable oversight.