Artykuły
Permanent URI for this collection
Recent Submissions
- 2026-09-16
Mediacja w edukacji językowej i kształceniu nauczycieli języków. Wprowadzenie
Otwarty dostępArtykułyJournal article - 2026-09-16
Linguistic mediation in EFL teacher education: Presentation and preliminary evaluation of the course
The article presents the preliminary evaluation of a course aimed at the implementation of the teaching of mediation (by designing information and communication technology-based (ICT) authorial materials) into the program developed for EFL teacher trainees at SWPS University. The pilot implementation took place in the academic year 2024/25 and has brought valuable insights into needs from the perspectives of the lecturer/ author of the syllabus, teacher trainer and students – course participants, on the basis of which recommendations are made with reference to course effectiveness. The article presents a model of implementing mediation-oriented strategies and activities into an EFL pre-service teacher education program, presenting results, and opinions based on its first version, conclusions, and recommendations for the future.Otwarty dostępArtykułyJournal article - 2026-10
From subtle control to severe threats: Emotional intimate partner violence (EIPV) recognition and myth acceptance among young adults
Prior research has been limited in examining emotional intimate partner violence (EIPV) related attitudes and cognitions. To address these research gaps, the current study aimed to build a better understanding of EIPV recognition and myth acceptance. Firstly, we developed and investigated reliability and factor structure, including factorial invariance for sex groups, of the Emotional Intimate Partner Violence Recognition (EIPV-R) Scale and the Emotional Intimate Partner Violence Myth Acceptance (EIPV-MA) Scale. Secondly, we explored associations between EIPV-R and EIPV-MA scores and external criteria. Study participants were a representative sample of young adults (aged 18-29 years) in Poland (N = 2705; 53.4% female). EIPV-R and EIPV-MA demonstrated strong psychometric properties and factorial invariance across sex. Both instruments were also found to be multidimensional, with five factors best capturing the EIPV-R scores (emotional withholding, controlling behavior, dominance and intimidation, severe threatening behavior, denigration) and three factors best capturing the EIPV-MA scores (romanticization of abuse, victim blaming, constrained perception of emotional abuse). Further analyses revealed than men are more likely to endorse rape myths and have more difficulties recognizing EIPV than women. Individuals who experienced intimate partner violence (IPV positive) scored significantly lower than IPV negative participants on total EIPV-R as well as on three subscales (controlling behavior, dominance and intimidation, denigration). However, IPV positive individuals scored higher on the severe threatening behavior subscale, which may be because the ability to identify overtly dangerous acts is vital for safety and survival. The current research findings can be used to build effective interventions aimed at reducing the prevalence of EIPV.Otwarty dostępArtykułyJournal article - 2026-09-30
From Timed Interpreted Systems to Split-Step Timed Interpreted Systems: Making Delivery Delay Explicit and Deciding Realizability in Security Protocol Verification
Timed security protocols rely on freshness tokens whose validity must hold when a response is delivered, not merely when it is sent. Many models collapse sending and delivery into one event, erasing the transit interval in which a token may expire. This abstraction has a concrete consequence: a delay-oblivious model can report that successful completion is reachable for a protocol that, on any network with a positive minimum transit latency, can never complete – a soundness gap rather than a missed attack. We introduce Split-Step Timed Interpreted Systems (SSTIS), a dense-time semantics in which every step has three phases – send, transit, and deliver-or-fail – making the transit interval and a delivery failure explicit local states.We formalise interpretations into classical Timed Interpreted Systems (TIS) and into SSTIS, and prove three results: SSTIS conservatively extends TIS at zero delay; a delivery-time failure is reachable in SSTIS but unsatisfiable in TIS by construction; and under a minimum transit delay D, a step whose freshness window L satisfies D > L fails in every run, including the fastest adversary-free delivery, so the protocol is unrealizable. From this we derive a linear-time realizability decision procedure that decides whether a guarded step can ever succeed, as a static pre-check before bounded model checking. We give a quantifier-free linear-real-arithmetic encoding with soundness and completeness proofs, apply it to the MobInfoSec SP2SP_MutualAuth protocol and classical timestamp families, and report a feasibility study across 779 protocol-variant instances and 12,410 reachability queries solved in under a second at the median.Otwarty dostępArtykułyJournal article - 2026-09-23
Conversational AutoML with Multi-Agent Architecture and Monte Carlo Tree Search for Hyperparameter Optimization
Automated Machine Learning (AutoML) reduces the technical requirements for applying machine learning, yet existing systems still demand technical expertise, optimize inefficiently, and rarely support conversational use. We present a conversational AutoML system with a five-agent architecture coordinated via natural language, tuning Random Forest and gradient-boosting models (XGBoost, LightGBM, and CatBoost) with Monte Carlo Tree Search (MCTS) and Bayesian optimization under repeated cross-validation conditions. Its distinctive component is an ensemble coordinator: specialist optimizer agents tune model families concurrently, and a coordinator fuses their outputs with the strongest baseline into a stacking ensemble, deploying it only when it improves on the best single model by cross-validation. This separates two measurable benefits: a speed-up from parallel search and an accuracy gain from ensemble fusion. On six benchmark datasets (five classification and one regression), the multi-agent coordinator is best or tied on every dataset and the only configuration with a positive average gain over defaults (+0.20 pp, versus −0.69 pp for single-agent MCTS); its search is 2.3× faster than the single-agent Bayesian baseline in aggregate, and it is systematically better than single-agent MCTS across datasets (Wilcoxon signed-rank, p = 0.031). While individual dataset gains are observed, none reach statistical significance when adjusted for multiple comparisons. For classification tasks, we additionally report F1-score, precision, recall, and AUC-ROC (Area Under the Receiver Operating Characteristic Curve).Otwarty dostępArtykułyJournal article