Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models

2026-03-28T19:27:47Z

Frontier LLM companies have repeatedly assured courts and regulators that their models do not store copies of training data. They further rely on safety alignment strategies via RLHF, system prompts, and output filters to block verbatim regurgitation of copyrighted works, and have cited the efficacy of these measures in their legal defenses against copyright infringement claims. We show that finetuning bypasses these protections: by training models to expand plot summaries into full text, a task naturally suited for commercial writing assistants, we cause GPT-4o, Gemini-2.5-Pro, and DeepSeek-V3.1 to reproduce up to 85-90% of held-out copyrighted books, with single verbatim spans exceeding 460 words, using only semantic descriptions as prompts and no actual book text. This extraction generalizes across authors: finetuning exclusively on Haruki Murakami's novels unlocks verbatim recall of copyrighted books from over 30 unrelated authors. The effect is not specific to any training author or corpus: random author pairs and public-domain finetuning data produce comparable extraction, while finetuning on synthetic text yields near-zero extraction, indicating that finetuning on individual authors' works reactivates latent memorization from pretraining. Three models from different providers memorize the same books in the same regions ($r \ge 0.90$), pointing to an industry-wide vulnerability. Our findings offer compelling evidence that model weights store copies of copyrighted works and that the security failures that manifest after finetuning on individual authors' works undermine a key premise of recent fair use rulings, where courts have conditioned favorable outcomes on the adequacy of measures preventing reproduction of protected expression.

Culturally Adaptive Explainable LLM Assessment for Multilingual Information Disorder: A Human-in-the-Loop Approach

2026-03-28T18:06:14Z

Recognizing information disorder is difficult because judgments about manipulation depend on cultural and linguistic context. Yet current Large Language Models (LLMs) often behave as monocultural, English-centric "black boxes," producing fluent rationales that overlook localized framing. Preliminary evidence from the multilingual Information Disorder (InDor) corpus suggests that existing models struggle to explain manipulated news consistently across communities. To address this gap, this ongoing study proposes a Hybrid Intelligence Loop, a human-in-the-loop (HITL) framework that grounds model assessment in human-written rationales from native-speaking annotators. The approach moves beyond static target-language few-shot prompting by pairing English task instructions with dynamically retrieved target-language exemplars drawn from filtered InDor annotations through In-Context Learning (ICL). In the initial pilot, the Exemplar Bank is seeded from these filtered annotations and used to compare static and adaptive prompting on Farsi and Italian news. The study evaluates span and severity prediction, the quality and cultural appropriateness of generated rationales, and model alignment across evaluator groups, providing a testbed for culturally grounded explainable AI.

From Influencers to Lecturers: Understanding Public Attitudes Toward Digital vs. Traditional Jobs

2026-03-28T15:01:02Z

The rapid expansion of high-speed internet has led to the emergence of new digital jobs, such as digital influencers, fitness models, and adult models who share content on subscription-based social media platforms. Across two experiments involving 1,002 participants, we combined theories from social psychology and information systems to investigate how digital jobs are perceived compared to matched established jobs, and predictors of attitudes toward those jobs (e.g., symbolic threat, contact, perceived usefulness). We found that individuals in digital professions were perceived as less favorably and less hard-working than those in matched established jobs. Digital jobs were also regarded as more threatening to societal values and less useful. The relation between job type and attitudes toward these jobs was partially mediated by contact with people working in these jobs, perceived usefulness, perception of hard work, and symbolic threat. These effects were consistent across both experiments, and various moderators: openness to new experiences, attitudes toward digitalization, political orientation, and age. Among the nine jobs examined, lecturers were perceived as most positive, while adult models were viewed as least positive. Overall, our findings demonstrate that integrating theories from social psychology and information systems can enhance our understanding of how attitudes are formed.

The First Issue Matters: Linking Task-Level Characteristics to Long-Term Newcomer Retention in OSS

2026-03-28T04:55:06Z

Sustaining newcomer participation is critical for the long-term health of open-source communities. Although prior research has explored various task recommendation approaches to help newcomers resolve their first-issue, these methods overlook how characteristics of first-issues may influence newcomers' long-term retention, limiting our understanding of whether initial success leads to sustained participation and hindering effective onboarding design. In this paper, we conduct a large-scale empirical study to examine how first-issue characteristics affect newcomer retention. We combine predictive analysis, interpretability techniques, and causal inference to estimate the causal effects of issue characteristics on retention outcomes. The prediction task supports the interpretation and shows that interaction-related characteristics exhibit stronger associations with retention than intrinsic issue attributes. The causal analysis further reveals that issues reported by moderately experienced contributors, accompanied by moderate discussion intensity and participation from project members, and neutral or slightly negative comment sentiment, have higher retention potential. These findings provide actionable insights for OSS maintainers on designing issue management practices that better support long-term newcomer retention.

Gender-Based Heterogeneity in Youth Privacy-Protective Behavior for Smart Voice Assistants: Evidence from Multigroup PLS-SEM

2026-03-28T04:11:05Z

This paper investigates how gender shapes privacy decision-making in youth smart voice assistant (SVA) ecosystems. Using survey data from 469 Canadian youths aged 16-24, we apply multigroup Partial Least Squares Structural Equation Modeling to compare males (N=241) and females (N=174) (total N = 415) across five privacy constructs: Perceived Privacy Risks (PPR), Perceived Privacy Benefits (PPBf), Algorithmic Transparency and Trust (ATT), Privacy Self-Efficacy (PSE), and Privacy Protective Behavior (PPB). Results provide exploratory evidence of gender heterogeneity in selected pathways. The direct effect of PPR on PPB is stronger for males (Male: \b{eta} = 0.424; Female: \b{eta} = 0.233; p < 0.1), while the indirect effect of ATT on PPB via PSE is stronger for females (Female: \b{eta} = 0.229; Male: \b{eta} = 0.132; p < 0.1). Descriptive analysis of non-binary (N=15) and prefer-not-to-say participants (N=39) shows lower trust and higher perceived risk than the binary groups, motivating future work with adequately powered gender-diverse samples. Overall, the findings provide exploratory evidence that gender may moderate key privacy pathways, supporting more responsive transparency and control interventions for youth SVA use.

Mind The Gap: How The Technical Mechanism Of Agentic AI Outpace Global Legal Frameworks

2026-03-28T01:35:58Z

This article presents the first systematic comparative survey of how public bodies, international organisations, national regulators, and the private sector define agentic artificial intelligence, identifying the technical inaccuracies pervading each definition. Analysing eleven regulatory instruments and industry frameworks -- including the EU AI Act, the OECD/G7 Principles, NIST, the UK ICO, and the European Commission -- alongside six leading developer architectures, this study demonstrates a persistent definitional gap: legal definitions consistently conflate model capability with agentic architecture, attribute cognitive deliberation to probabilistic token prediction, and treat autonomy as a scalar property rather than a structural shift from single-inference to iterative execution loops with tool integration. A consensus technical definition synthesised from developer documentation is proposed. The article examines the consequences of this gap, demonstrating that definitional imprecision produces regulatory instruments structurally incapable of governing the actual mechanisms -- system prompts, API permissions, sandboxing, and orchestration code -- that constitute agentic autonomy.

Voice-based debate with an AI adversary is associated with increased divergent ideation

2026-03-28T01:23:38Z

Concerns that interacting with generative AI homogenizes human cognition are largely based on evidence from text-based interactions, potentially conflating the effects of AI systems with those of written communication. This study examines whether these patterns depend on communication modality rather than on AI itself. Analyzing 957 open-ended debates between university students and a knowledgeable AI adversary, we show that modality corresponds to distinct structural patterns in discourse. Consistent with classic distinctions between orality and literacy, spoken interactions are significantly more verbose and exhibit greater repetition of words and phrases than text-based exchanges. This redundancy, however, is functional: voice users rely on recurrent phrasing to maintain coherence while exploring a wider range of ideas. In contrast, text-based interaction favors concision and refinement but constrains conceptual breadth. These findings suggest that perceived cognitive limitations attributed to generative AI partly reflect the medium through which it is accessed.

Persona-Based Simulation of Human Opinion at Population Scale

2026-03-28T00:10:57Z

What does it mean to model a person, not merely to predict isolated responses, preferences, or behaviors, but to simulate how an individual interprets events, forms opinions, makes judgments, and acts consistently across contexts? This question matters because social science requires not only observing and predicting human outcomes, but also simulating interventions and their consequences. Although large language models (LLMs) can generate human-like answers, most existing approaches remain predictive, relying on demographic correlations rather than representations of individuals themselves. We introduce SPIRIT (Semi-structured Persona Inference and Reasoning for Individualized Trajectories), a framework designed explicitly for simulation rather than prediction. SPIRIT infers psychologically grounded, semi-structured personas from public social media posts, integrating structured attributes (e.g., personality traits and world beliefs) with unstructured narrative text reflecting values and lived experience. These personas prompt LLM-based agents to act as specific individuals when answering survey questions or responding to events. Using the Ipsos KnowledgePanel, a nationally representative probability sample of U.S. adults, we show that SPIRIT-conditioned simulations recover self-reported responses more faithfully than demographic persona and reproduce human-like heterogeneity in response patterns. We further demonstrate that persona banks can function as virtual respondent panels for studying both stable attitudes and time-sensitive public opinion.

Multi-Level Barriers to Generative AI Adoption Across Disciplines and Professional Roles in Higher Education

2026-03-27T23:48:25Z

Generative Artificial Intelligence (GenAI) is rapidly reshaping higher education, yet barriers to its adoption across different disciplines and institutional roles remain underexplored. Existing literature frequently attributes adoption barriers to individual-level factors such as perceived usefulness and ease of use. This study instead investigates whether such barriers are structurally produced. Drawing on a multi-method survey analysis of 272 academic and professional services (PSs) staff at a Russell Group university, we examine how disciplinary contexts and institutional roles shape perceived barriers. By integrating multinomial logistic regression (MLR), structural equation modelling (SEM), and semantic clustering of open-ended responses, we move beyond descriptive accounts to provide a multi-level explanation of GenAI adoption. Our findings reveal clear, systematic differences: non-STEM academics primarily report ethical and cultural barriers related to academic integrity, whereas STEM and PSs staff disproportionately emphasize institutional, governance, and infrastructure constraints. We conclude that GenAI adoption barriers are deeply embedded in organizational ecosystems and epistemic norms, suggesting that universities must move beyond generalized training to develop role-specific governance and support frameworks.

The Last Fingerprint: How Markdown Training Shapes LLM Prose

2026-03-27T21:42:06Z

Large language models produce em dashes at varying rates, and the observation that some models "overuse" them has become one of the most widely discussed markers of AI-generated text. Yet no mechanistic account of this pattern exists, and the parallel observation that LLMs default to markdown-formatted output has never been connected to it. We propose that the em dash is markdown leaking into prose -- the smallest surviving unit of the structural orientation that LLMs acquire from markdown-saturated training corpora. We present a five-step genealogy connecting training data composition, structural internalization, the dual-register status of the em dash, and post-training amplification. We test this with a two-condition suppression experiment across twelve models from five providers (Anthropic, OpenAI, Meta, Google, DeepSeek): when models are instructed to avoid markdown formatting, overt features (headers, bullets, bold) are eliminated or nearly eliminated, but em dashes persist -- except in Meta's Llama models, which produce none at all. Em dash frequency and suppression resistance vary from 0.0 per 1,000 words (Llama) to 9.1 (GPT-4.1 under suppression), functioning as a signature of the specific fine-tuning procedure applied. A three-condition suppression gradient shows that even explicit em dash prohibition fails to eliminate the artifact in some models, and a base-vs-instruct comparison confirms that the latent tendency exists pre-RLHF. These findings connect two previously isolated online discourses and reframe em dash frequency as a diagnostic of fine-tuning methodology rather than a stylistic defect.

Using LLMs for Knowledge Component-level Correctness Labeling in Open-ended Coding Problems

2026-03-27T21:30:24Z

Fine-grained skill representations, commonly referred to as knowledge components (KCs), are fundamental to many approaches in student modeling and learning analytics. However, KC-level correctness labels are rarely available in real-world datasets, especially for open-ended programming tasks where solutions typically involve multiple KCs simultaneously. Simply propagating problem-level correctness to all associated KCs obscures partial mastery and often leads to poorly fitted learning curves. To address this challenge, we propose an automated framework that leverages large language models (LLMs) to label KC-level correctness directly from student-written code. Our method assesses whether each KC is correctly applied and further introduces a temporal context-aware Code-KC mapping mechanism to better align KCs with individual student code. We evaluate the resulting KC-level correctness labels in terms of learning curve fit and predictive performance using the power law of practice and the Additive Factors Model. Experimental results show that our framework leads to learning curves that are more consistent with cognitive theory and improves predictive performance, compared to baselines. Human evaluation further demonstrates substantial agreement between LLM and expert annotations.

Persona Alchemy: Designing, Evaluating, and Implementing Psychologically-Grounded LLM Agents for Diverse Stakeholder Representation

2026-03-27T21:05:03Z

Despite advances in designing personas for Large Language Models (LLM), challenges remain in aligning them with human cognitive processes and representing diverse stakeholder perspectives. We introduce a Social Cognitive Theory (SCT) agent design framework for designing, evaluating, and implementing psychologically grounded LLMs with consistent behavior. Our framework operationalizes SCT through four personal factors (cognitive, motivational, biological, and affective) for designing, six quantifiable constructs for evaluating, and a graph database-backed architecture for implementing stakeholder personas. Experiments tested agents' responses to contradicting information of varying reliability. In the highly polarized renewable energy transition discourse, we design five diverse agents with distinct ideologies, roles, and stakes to examine stakeholder representation. The evaluation of these agents in contradictory scenarios occurs through comprehensive processes that implement the SCT. Results show consistent response patterns ($R^2$ range: $0.58-0.61$) and systematic temporal development of SCT construct effects. Principal component analysis identifies two dimensions explaining $73$% of variance, validating the theoretical structure. Our framework offers improved explainability and reproducibility compared to black-box approaches. This work contributes to ongoing efforts to improve diverse stakeholder representation while maintaining psychological consistency in LLM personas.

Transparency as Architecture: Structural Compliance Gaps in EU AI Act Article 50 II

2026-03-27T20:50:42Z

Art. 50 II of the EU Artificial Intelligence Act mandates dual transparency for AI-generated content: outputs must be labeled in both human-understandable and machine-readable form for automated verification. This requirement, entering into force in August 2026, collides with fundamental constraints of current generative AI systems. Using synthetic data generation and automated fact-checking as diagnostic use cases, we show that compliance cannot be reduced to post-hoc labeling. In fact-checking pipelines, provenance tracking is not feasible under iterative editorial workflows and non-deterministic LLM outputs; moreover, the assistive-function exemption does not apply, as such systems actively assign truth values rather than supporting editorial presentation. In synthetic data generation, persistent dual-mode marking is paradoxical: watermarks surviving human inspection risk being learned as spurious features during training, while marks suited for machine verification are fragile under standard data processing. Across both domains, three structural gaps obstruct compliance: (a) absent cross-platform marking formats for interleaved human-AI outputs; (b) misalignment between the regulation's 'reliability' criterion and probabilistic model behavior; and (c) missing guidance for adapting disclosures to heterogeneous user expertise. Closing these gaps requires transparency to be treated as an architectural design requirement, demanding interdisciplinary research across legal semantics, AI engineering, and human-centered desi

In your own words: computationally identifying interpretable themes in free-text survey data

2026-03-27T19:12:53Z

Free-text survey responses can provide nuance often missed by structured questions, but remain difficult to statistically analyze. To address this, we introduce In Your Own Words, a computational framework for exploratory analyses of free-text survey data that identifies structured, interpretable themes in free-text responses more precisely than previous computational approaches, facilitating systematic analysis. To illustrate the benefits of this approach, we apply it to a new dataset of free-text descriptions of race, gender, and sexual orientation from 1,004 U.S. participants. The themes our approach learns have three practical applications in survey research. First, the themes can suggest structured questions to add to future surveys by surfacing salient constructs -- such as belonging and identity fluidity -- that existing surveys do not capture. Second, the themes reveal heterogeneity within standardized categories, explaining additional variation in health, well-being, and identity importance. Third, the themes illuminate systematic discordance between self-identified and perceived identities, highlighting mechanisms of misrecognition that existing measures do not reflect. More broadly, our framework can be deployed in a wide range of survey settings to identify interpretable themes from free text, complementing existing qualitative methods.

Archetypes and gender in fiction: A data-driven mapping of gender stereotypes in stories

2026-03-27T19:09:53Z

Fictional character representations reflect social norms and biases. For example, women are relatively underrepresented in television and film, irrespective of genre, and are frequently stereotyped in these media. Here, we draw on a data-driven operationalization of archetypes -- archetypometrics -- to explore the characterization of 2,000 canonically male and female characters. From an overall space of six pairs of base archetypes, we find that canonically female characters tend more toward Hero, Adventurer, Diva, and Sophisticate archetypes, while male characters, tend toward Fool, Traditionalist, Outcast, Brute and Outcast types. However, overarching patterns by gender nevertheless sustain traditional stereotypes: The seemingly positive heroic bias toward females is undercut by heroic female characters being more masculine than other female characters. We discuss the societal implications of skewed archetype representation by character gender.