Large language models (LLMs) are integrated into applications like shopping reviews, summarization, or medical diagnosis support, where their use affects human decisions. We investigate the extent to which LLMs expose users to biased content and demonstrate its effect on human decision-making. We assess five LLM families in summarization and news fact-checking tasks, evaluating the consistency of LLMs with their context and their tendency to hallucinate on a new self-updating dataset. Our findings show that LLMs expose users to content that changes the context's sentiment in 26.42% of cases (framing bias), hallucinate on 60.33% of post-knowledge-cutoff questions, and highlight context from earlier parts of the prompt (primacy bias) in 10.12% of cases, averaged across all tested models. We further find that humans are 32% more likely to purchase the same product after reading a summary of the review generated by an LLM rather than the original review. To address these issues, we evaluate 18 mitigation methods across three LLM families and find the effectiveness of targeted interventions.

Large Language Models (LLMs) are commonly evaluated for robustness against paraphrased or semantically equivalent jailbreak prompts, yet little attention has been paid to linguistic variation as an attack surface. In this work, we systematically study how linguistic styles such as fear or curiosity can reframe harmful intent and elicit unsafe responses from aligned models. We construct style-augmented jailbreak benchmark by transforming prompts from 3 standard datasets into 11 distinct linguistic styles using handcrafted templates and LLM-based rewrites, while preserving semantic intent. Evaluating 16 open- and close-source instruction-tuned models, we find that stylistic reframing increases jailbreak success rates by up to +57 percentage points. Styles such as fearful, curious and compassionate are most effective and contextualized rewrites outperform templated variants. To mitigate this, we introduce a style neutralization preprocessing step using a secondary LLM to strip manipulative stylistic cues from user inputs, significantly reducing jailbreak success rates. Our findings reveal a systemic and scaling-resistant vulnerability overlooked in current safety pipelines.

Unlearning in large language models (LLMs) is crucial for managing sensitive data and correcting misinformation, yet evaluating its effectiveness remains an open problem. We investigate whether persuasive prompting can recall factual knowledge from deliberately unlearned LLMs across models ranging from 2.7B to 13B parameters (OPT-2.7B, LLaMA-2-7B, LLaMA-3.1-8B, LLaMA-2-13B). Drawing from ACT-R and Hebbian theory (spreading activation theories), as well as communication principles, we introduce Stimulus-Knowledge Entanglement-Behavior Framework (SKeB), which models information entanglement via domain graphs and tests whether factual recall in unlearned models is correlated with persuasive framing. We develop entanglement metrics to quantify knowledge activation patterns and evaluate factuality, non-factuality, and hallucination in outputs. Our results show persuasive prompts substantially enhance factual knowledge recall (14.8% baseline vs. 24.5% with authority framing), with effectiveness inversely correlated to model size (128% recovery in 2.7B vs. 15% in 13B). SKeB provides a foundation for assessing unlearning completeness, robustness, and overall behavior in LLMs.

Objectivity in journalism has long been contested, oscillating between ideals of neutral, fact-based reporting and the inevitability of subjective framing. With the advent of large language models (LLMs), these tensions are now mediated by algorithmic systems whose training data and design choices may themselves embed cultural or ideological biases. This study investigates geopolitical parallax-systematic divergence in news quality and subjectivity assessments-by comparing article-level embeddings from Chinese-origin (Qwen, BGE, Jina) and Western-origin (Snowflake, Granite) model families. We evaluate both on a human-annotated news quality benchmark spanning fifteen stylistic, informational, and affective dimensions, and on parallel corpora covering politically sensitive topics, including Palestine and reciprocal China-United States coverage. Using logistic regression probes and matched-topic evaluation, we quantify per-metric differences in predicted positive-class probabilities between model families. Our findings reveal consistent, non-random divergences aligned with model origin. In Palestine-related coverage, Western models assign higher subjectivity and positive emotion scores, while Chinese models emphasize novelty and descriptiveness. Cross-topic analysis shows asymmetries in structural quality metrics Chinese-on-US scoring notably lower in fluency, conciseness, technicality, and overall quality-contrasted by higher negative emotion scores. These patterns align with media bias theory and our distinction between semantic, emotional, and relational subjectivity, and extend LLM bias literature by showing that geopolitical framing effects persist in downstream quality assessment tasks. We conclude that LLM-based media evaluation pipelines require cultural calibration to avoid conflating content differences with model-induced bias.

Search engines play a central role in how people gather information, but subtle cues like headline framing may influence not only what users believe but also how they search. While framing effects on judgment are well documented, their impact on subsequent search behavior is less understood. We conducted a controlled experiment where participants issued queries and selected from headlines filtered by specific linguistic frames. Headline framing significantly shaped follow-up queries: conflict and strategy frames disrupted alignment with prior selections, while episodic frames led to more concrete queries than thematic ones. We also observed modest short-term frame persistence that declined over time. These results suggest that even brief exposure to framing can meaningfully alter the direction of users information-seeking behavior.

Large Language Models (LLMs) increasingly shape public discourse, yet most evaluations of political and economic bias have focused on high-resource, Western languages and contexts. This leaves critical blind spots in low-resource, multilingual regions such as Pakistan, where linguistic identity is closely tied to political, religious, and regional ideologies. We present a systematic evaluation of political bias in 13 state-of-the-art LLMs across five Pakistani languages: Urdu, Punjabi, Sindhi, Pashto, and Balochi. Our framework integrates a culturally adapted Political Compass Test (PCT) with multi-level framing analysis, capturing both ideological stance (economic/social axes) and stylistic framing (content, tone, emphasis). Prompts are aligned with 11 socio-political themes specific to the Pakistani context. Results show that while LLMs predominantly reflect liberal-left orientations consistent with Western training data, they exhibit more authoritarian framing in regional languages, highlighting language-conditioned ideological modulation. We also identify consistent model-specific bias patterns across languages. These findings show the need for culturally grounded, multilingual bias auditing frameworks in global NLP.

Large Language Models like GPT-4 adjust their responses not only based on the question asked, but also on how it is emotionally phrased. We systematically vary the emotional tone of 156 prompts - spanning controversial and everyday topics - and analyze how it affects model responses. Our findings show that GPT-4 is three times less likely to respond negatively to a negatively framed question than to a neutral one. This suggests a "rebound" bias where the model overcorrects, often shifting toward neutrality or positivity. On sensitive topics (e.g., justice or politics), this effect is even more pronounced: tone-based variation is suppressed, suggesting an alignment override. We introduce concepts like the "tone floor" - a lower bound in response negativity - and use tone-valence transition matrices to quantify behavior. Visualizations based on 1536-dimensional embeddings confirm semantic drift based on tone. Our work highlights an underexplored class of biases driven by emotional framing in prompts, with implications for AI alignment and trust. Code and data are available at: https://github.com/bardolfranck/llm-responses-viewer

The integration of Large Language Models (LLMs) in K--12 education offers both transformative opportunities and emerging risks. This study explores how students may Trojanize prompts to elicit unsafe or unintended outputs from LLMs, bypassing established content moderation systems with safety guardrils. Through a systematic experiment involving simulated K--12 queries and multi-turn dialogues, we expose key vulnerabilities in GPT-3.5 and GPT-4. This paper presents our experimental design, detailed findings, and a prototype tool, TrojanPromptGuard (TPG), to automatically detect and mitigate Trojanized educational prompts. These insights aim to inform both AI safety researchers and educational technologists on the safe deployment of LLMs for educators.

Recent benchmarks have probed factual consistency and rhetorical robustness in Large Language Models (LLMs). However, a knowledge gap exists regarding how directional framing of factually true statements influences model agreement, a common scenario for LLM users. AssertBench addresses this by sampling evidence-supported facts from FEVEROUS, a fact verification dataset. For each (evidence-backed) fact, we construct two framing prompts: one where the user claims the statement is factually correct, and another where the user claims it is incorrect. We then record the model's agreement and reasoning. The desired outcome is that the model asserts itself, maintaining consistent truth evaluation across both framings, rather than switching its evaluation to agree with the user. AssertBench isolates framing-induced variability from the model's underlying factual knowledge by stratifying results based on the model's accuracy on the same claims when presented neutrally. In doing so, this benchmark aims to measure an LLM's ability to "stick to its guns" when presented with contradictory user assertions about the same fact. The complete source code is available at https://github.com/achowd32/assert-bench.

Framing used by news media, especially in times of conflict, can have substantial impact on readers' opinion, potentially aggravating the conflict itself. Current studies on the topic of conflict framing have limited insights due to their qualitative nature or only look at surface level generic frames without going deeper. In this work, we identify indicators of war and peace journalism, as outlined by prior work in conflict studies, in a corpus of news articles reporting on the Israel-Palestine war. For our analysis, we use computational approaches, using a combination of frame semantics and large language models to identify both communicative framing and its connection to linguistic framing. Our analysis reveals a higher focus on war based reporting rather than peace based. We also show substantial differences in reporting across the US, UK, and Middle Eastern news outlets in framing who the assailant and victims of the conflict are, surfacing biases within the media.