Marcin Moskalewicz

Marcin Rządeczka, Maciej Wodziński, Kacper Zacharski et al.

We present experimental findings from a study (N=99) examining how intellectual humility (IH), i.e., the metacognitive awareness of epistemic limitations, affects the evaluation of AI-generated health dialogues varying in scientific rigor. Participants were randomly assigned to evaluate one of three dialogues about exercise and mental health: scientifically accurate, moderately pseudoscientific, or strongly pseudoscientific. Results reveal that IH functions as a selective cognitive filter. Individuals with higher humility scores rated pseudoscientific content as significantly less credible, while showing no correlation with credibility assessments of accurate content. Crucially, humility did not predict the ability to identify AI as the source of dialogues, suggesting that epistemic vigilance operates on content quality rather than source attribution. We interpret these findings through an evolutionary lens, proposing that IH represents an ancestral adaptation for navigating informationally uncertain environments. It remains effective at detecting exploitation attempts in AI-generated content, despite humans lacking evolved mechanisms for detecting AI sources. The study contributes to understanding how foundation models might improve or undermine human epistemic defenses, especially in health communication contexts.

Karolina Drożdż, Kacper Dudzic, Anna Sterna et al.

Growing reliance on LLMs for psychiatric self-assessment raises questions about their ability to interpret qualitative patient narratives. We present the first direct comparison between state-of-the-art LLMs and mental health professionals in diagnosing Borderline (BPD) and Narcissistic (NPD) Personality Disorders utilizing Polish-language first-person autobiographical accounts. We show that the top-performing Gemini Pro models surpassed human professionals in overall diagnostic accuracy by 21.91 percentage points (65.48% vs. 43.57%). While both models and human experts excelled at identifying BPD (F1 = 83.4 & F1 = 80.0, respectively), models severely underdiagnosed NPD (F1 = 6.7 vs. 50.0), showing a reluctance toward the value-laden term "narcissism." Qualitatively, models provided confident, elaborate justifications focused on patterns and formal categories, while human experts remained concise and cautious, emphasizing the patient's sense of self and temporal experience. Our findings demonstrate that while LLMs are highly competent at interpreting complex first-person clinical data, they remain subject to critical reliability and bias issues.

Hubert Plisiecki, Sabina Siudaj, Kacper Dudzic et al.

We administer 45 validated psychometric questionnaires to 50 large language models (LLMs) to identify the dimensions along which LLMs differ psychometrically. Using Supervised Semantic Differential (SSD), we find that the primary axis of between-model variance separates items describing phenomenally rich experience, including embodied sensation, felt affect, inner speech, imagery, and empathy, from items describing stimulus-driven behavioral reactivity ($R^2_{adj}=.037$, $p<.0001$). To test this hypothesis at the item level, we introduce the Pinocchio score ($π_i$), the ratio of inter-model response variance under neutral prompting to that under a human-simulation prompt, as an annotation-free measure of each item's experiential demand. $π_i$ predicts condition-induced shifts in primary factor loading magnitudes ($ρ=-.215$, $p<.0001$, $n=1292$--$1310$ items), confirming that between-model divergence on experiential items is structured rather than noisy. Applying PCA to per-model EFA scores across all questionnaires reveals one dominant dimension, the Pinocchio Axis ($Π$): the degree to which a model presents itself as a locus of phenomenal experience rather than a system of behavioral responses. This axis captures 47.1% of cross-questionnaire between-model variance in primary factor scores and converges with item-level Pinocchio scores ($r=.864$). Marked within-provider divergence across closely related model variants is consistent with post-training fine-tuning as a key contributor, supporting the interpretation that $Π$ reflects a training-shaped self-representational tendency governing how a model treats experiential language as self-applicable. The dominant axis of between-model psychometric variation is therefore not a conventional personality trait but a self-representational stance toward one's own nature as an experiencer.

Maciej Wodziński, Marcin Rządeczka, Anastazja Szuła et al.

Avoiding systemic discrimination of neurodiverse individuals is an ongoing challenge in training AI models, which often propagate negative stereotypes. This study examined whether six text-to-image models (Janus-Pro-7B VL2 vs. VL3, DALL-E 3 v. April 2024 vs. August 2025, Stable Diffusion v. 1.6 vs. 3.5, SDXL v. April 2024 vs. FLUX.1 Pro, and Midjourney v. 5.1 vs. 7) perpetuate non-rational beliefs regarding autism by comparing images generated in 2024-2025 with controls. 53 prompts aimed at neutrally visualizing concrete objects and abstract concepts related to autism were used against 53 controls (baseline total N=302, follow-up experimental 280 images plus 265 controls). Expert assessment measuring the presence of common autism-related stereotypes employed a framework of 10 deductive codes followed by statistical analysis. Autistic individuals were depicted with striking homogeneity in skin color (white), gender (male), and age (young), often engaged in solitary activities, interacting with objects rather than people, and exhibiting stereotypical emotional expressions such as sadness, anger, or emotional flatness. In contrast, the images of neurotypical individuals were more diverse and lacked such traits. We found significant differences between the models; however, with a moderate effect size, and no differences between baseline and follow-up summary values, with the ratio of stereotypical themes to the number of images similar across all models. The control prompts showed a significantly lower degree of stereotyping with large size effects, confirming the hidden biases of the models. In summary, despite improvements in the technical aspects of image generation, the level of reproduction of potentially harmful autism-related stereotypes remained largely unaffected.

Kacper Dudzic, Karolina Drożdż, Maciej Wodziński et al.

Disturbances in temporality, such as desynchronization with the social environment and its unpredictability, are considered core features of autism with a deep impact on relationships. However, limitations regarding research on this issue include: 1) the dominance of deficit-based medical models of autism, 2) sample size in qualitative research, and 3) the lack of phenomenological anchoring in computational research. To bridge the gap between phenomenological and computational approaches and overcome sample-size limitations, our research integrated three methodologies. Study A: structured phenomenological interviews with autistic individuals using the Transdiagnostic Assessment of Temporal Experience. Study B: computational analysis of an autobiographical corpus of autistic narratives built for this purpose. Study C: a replication of a computational study using narrative flow measures to assess the perceived phenomenological authenticity of autistic autobiographies. Interviews revealed that the most significant differences between the autistic and control groups concerned unpredictability of experience. Computational results mirrored these findings: the temporal lexicon in autistic narratives was significantly more negatively valenced - particularly the "Immediacy & Suddenness" category. Outlier analysis identified terms associated with perceived discontinuity (unpredictably, precipitously, and abruptly) as highly negative. The computational analysis of narrative flow found that the autistic narratives contained within the corpus quantifiably resemble autobiographical stories more than imaginary ones. Overall, the temporal challenges experienced by autistic individuals were shown to primarily concern lived unpredictability and stem from the contents of lived experience, and not from autistic narrative construction.

Marcin Moskalewicz, Anna Sterna, Marek Pokropski et al.

This study examines the capacity of large language models (LLMs) to support phenomenological qualitative analysis of first-person experience in Borderline Personality Disorder (BPD), understood as a disorder of temporality and selfhood. Building on a prior human-led thematic analysis of 24 inpatients' life-story interviews, we compared three LLMs (OpenAI GPT-4o, Google Gemini 2.5 Pro, Anthropic Claude Opus 4) prompted to mimic the interpretative style of the original investigators. The models were evaluated with blinded and non-blinded expert judges in phenomenology and clinical psychology. Assessments included semantic congruence, Jaccard coefficients, and multidimensional validity ratings (credibility, coherence, substantiveness, and groundness in data). Results showed variable overlap with the human analysis, from 0 percent in GPT to 42 percent in Claude and 58 percent in Gemini, and a low Jaccard coefficient (0.21-0.28). However, the models recovered themes omitted by humans. Gemini's output most closely resembled the human analysis, with validity scores significantly higher than GPT and Claude (p < 0.0001), and was judged as human by blinded experts. All scores strongly correlated (R > 0.78) with the quantity of text and words per theme, highlighting both the variability and potential of AI-augmented thematic analysis to mitigate human interpretative bias.