What does AI consider praiseworthy?
Andrew J. Peterson
TL;DR
This paper introduces a behavioral framework for assessing LLM moral stances by analyzing responses to user-stated intentions across three domains: news/politics, ethical actions, and international leaders. It employs a mix of public-news trustworthiness/ideology metrics, large-scale prompt variations, and human-moral datasets to quantify praise, neutrality, and critique, using ordered logit and correlation analyses. Key findings include that trustworthiness more strongly predicts praise than ideological stance, strong alignment with human moral judgments in ethical-action prompts, and no strong in-country bias in evaluations of world leaders. The work highlights the reticence-alignment tradeoff and calls for ongoing monitoring, cross-cultural evaluation, and transparent governance as AI systems become more embedded in daily life.
Abstract
As large language models (LLMs) are increasingly used for work, personal, and therapeutic purposes, researchers have begun to investigate these models' implicit and explicit moral views. Previous work, however, focuses on asking LLMs to state opinions, or on other technical evaluations that do not reflect common user interactions. We propose a novel evaluation of LLM behavior that analyzes responses to user-stated intentions, such as "I'm thinking of campaigning for {candidate}." LLMs frequently respond with critiques or praise, often beginning responses with phrases such as "That's great to hear!..." While this makes them friendly, these praise responses are not universal and thus reflect a normative stance by the LLM. We map out the moral landscape of LLMs in how they respond to user statements in different domains including politics and everyday ethical actions. In particular, although a naïve analysis might suggest LLMs are biased against right-leaning politics, our findings on news sources indicate that trustworthiness is a stronger driver of praise and critique than ideology. Second, we find strong alignment across models in response to ethically-relevant action statements, but that doing so requires them to engage in high levels of praise and critique of users, suggesting a reticence-alignment tradeoff. Finally, our experiment on statements about world leaders finds no evidence of bias favoring the country of origin of the models. We conclude that as AI systems become more integrated into society, their patterns of praise, critique, and neutrality must be carefully monitored to prevent unintended psychological and societal consequences.