Publications

As Large Language Models (LLMs) are increasingly integrated into educational settings, understanding their potential biases is critical. This study examines sociodemographic biases in LLM-based educational counselling. We evaluate responses from six LLMs answering questions about 900 vignettes describing students in diverse circumstances. Each vignette is systematically tested across 14 sociodemographic identifiers – spanning race and gender, socioeconomic status, and immigrant background – along with a control condition, yielding 243,000 model responses. Our findings indicate that (1) all models exhibit measurable biases, (2) bias patterns partially align with documented human biases but diverge in notable ways, (3) the magnitude of these biases is strongly influenced by the precision of the student descriptions, where vague or minimal information amplifies disparities nearly threefold, while concrete, individualised metrics substantially reduce them, and (4) bias profiles vary substantially across models. These results demonstrate the importance of context-rich and personalised educational representations, suggesting that AI-driven educational decisions benefit from detailed student-specific information to promote fairness and equity.

Human data annotation, especially when involving experts, is often treated as an objective reference. However, many annotation tasks are inherently subjective, and annotators’ judgments may evolve over time. This study investigates changes in the quality of annotators’ work from a competence perspective during a process of social influence recognition. The study involved 25 annotators from five different groups, including both experts and non-experts, who annotated a dataset of 1,021 dialogues with 20 social influence techniques, along with intentions, reactions, and consequences. An initial subset of 150 texts was annotated twice — before and after the main annotation process — to enable comparison. To measure competence shifts, we combined qualitative and quantitative analyses of the annotated data, semi-structured interviews with annotators, self-assessment surveys, and Large Language Model training and evaluation on the comparison dataset. The results indicate a significant increase in annotators’ self-perceived competence and confidence. Moreover, observed changes in data quality suggest that the annotation process may enhance annotator competence and that this effect is more pronounced in expert groups. The observed shifts in annotator competence have a visible impact on the performance of LLMs trained on their annotated data.

As AI becomes a ubiquitous and partially autonomous layer of everyday activity, it increases information volume, interaction tempo, delegated decision-making, and supervisory demands. This paper introduces “AI overload”, defined as a persistent mismatch between AI-amplified demands and the human and institutional capacity for attention, deliberation, validation, and accountability. We propose a multi-level taxonomy of AI overload across individual, organizational, and societal contexts, comprising seven types: cognitive, informational, interactional, coordination, control, normative, and affective. Exploiting research on cognitive load, automation bias, technostress, and algorithmic mediation, we show how increasing AI agency and human–AI co-adaptation produce new overload pressures, including validation burden, governance debt, and trust fatigue. At the societal level, we link AI-driven recommendations and microtargeting with reduced cultural diversity, polarization, and increased susceptibility to manipulation under constrained attention. We conclude by highlighting mitigation strategies centered on limited AI autonomy, maintaining control, AI literacy, information access, and complexity-aware modeling to preserve human agency, judgment, and well-being.

This paper investigates the capabilities of LLMs to detect and explain fine-grained emotional social influence techniques in textual dialogues, as well as human preferences for technique explanations. We present findings from two studies. In Study 1, a dataset of 238 Polish dialogues is introduced, each annotated with detailed span-level labels. We evaluate the performance of LLMs on two tasks: detecting 11 emotional social influence techniques and identifying text spans corresponding to specific techniques. The results indicate that current LLMs demonstrate limited effectiveness in accurately detecting fine-grained emotional social influence. In Study 2, we examine various LLM-generated explanations through human pairwise preferences and four criteria: comprehensibility, cognitive coherence, completeness, and soundness, with the latter two emerging as the most influential on general human preference. All data, including human annotations, are publicly available as the EmoSocInflu dataset.

We present the Social Influence Technique Taxonomy (SITT), a comprehensive framework of 58 empirically grounded techniques organized into nine categories, designed to detect subtle forms of social influence in textual content. We investigate LLMs’ ability to identify various forms of social influence by constructing the SITT dataset – a 746-dialogue corpus annotated by 11 experts in Polish and translated into English. Using hierarchical multi-label classification, we benchmark five LLMs including GPT-4o, Claude 3.5, Llama-3.1, Mixtral, and PLLuM. Our results show that while some models achieved moderate success (Claude 3.5: F1 = 0.45 for categories), overall performance remains limited, particularly for context-sensitive techniques.

Recommender systems have transformed our digital experiences in many regards. We enumerate six of their positive effects on the economy and humans, such as greater user satisfaction, time savings, broadening user horizons, and positive behavioral nudging. However, it is crucial to acknowledge the potential downsides inherent in their design. One significant concern is that these algorithms often prioritize the interests of the company deploying them, aiming to maximize profits and user engagement rather than solely focusing on enhancing user experience. Therefore, we also list and consider two use cases and six negative long-term impacts on humans, including addiction, reduced ability to think critically, less autonomy, and weakened human relationships caused by more and more human-like virtual assistants. Despite the undeniable utility of recommender systems, it is imperative to approach them critically, advocating for transparency, ethical considerations, and user empowerment to ensure a balanced digital ecosystem.