Our Work

This repository contains the replication package for the paper "Human Perception of Ethics in Self-Driving Cars: An Empirical Study." The replication package includes the materials needed to repr…

This repository contains the replication package for the paper "Human Perception of Ethics in Self-Driving Cars: An Empirical Study." The replication package includes the materials needed to reproduce and inspect the results reported in the paper, such as anonymized survey data, CARLA simulation scripts, screencasts, data-analysis notebooks, the complete survey questionnaire, and the qualitative coding of participants' open-ended responses. These artifacts enable researchers to reproduce the quantitative analyses, inspect the test-case generation and execution process, and trace the reported qualitative findings to the original anonymized responses and assigned codes. Paper Abstract Ethical considerations in self-driving cars (SDCs) pose challenges beyond traditional functional or safety metrics. Automated testing frameworks rely on pass–fail as quality criteria for the SDC behavior, but it remains unclear how these outcomes align with human judgments of ethical behavior. To address this gap, we conduct an empirical study with 60 participants to investigate how humans perceive SDC behavior. The study employs test cases in which a pedestrian crosses the road, and the SDC exhibits behaviors varying in ethicality levels (ethical vs. unethical) and functional compliance (pass vs. fail). Participants evaluate each test case as ethical, neutral, or unethical and provide textual justifications for their assessments. The study also examines participants' perceptions of the relationship between ethical behavior and safety. Our results show that participants reliably identify unethical behavior, while ethical test cases elicited mixed responses influenced by pedestrian intentions, driving norms, and precautionary actions. Most participants distinguish ethics from safety, indicating that functional pass–fail alone does not capture ethical acceptability. These results highlight the need to integrate human-centered ethical considerations into SDC testing and design, complementing technical safety criteria to better align vehicle behavior with the expectations of technically literate observers.

This work presents proposing two artificial intelligence methods including Least squares support vector machine (LSSVM) and Adaptive neuro fuzzy inference system (ANFIS) for the prediction of caustic …

This work presents proposing two artificial intelligence methods including Least squares support vector machine (LSSVM) and Adaptive neuro fuzzy inference system (ANFIS) for the prediction of caustic current efficiency (CCE) and cell voltage as a function of pH, current density, brine concentration, electrolyte velocity, operating temperature, and run time. The predictions of LSSVM and ANFIS models were evaluated by the experimental values of this process graphically and statistically. The overall R-squared values of LSSVM and ANFIS for prediction of CCE were 0.999 and 0.972, respectively. On the other hand, these values for cell voltage prediction were 1 and 0.998. According to the CCE and cell voltage predictions results, LSSVM algorithm has great performance in prediction of chlor-alkali membrane cell processes. Furthermore, artificial intelligence methods can have wide use in electrolytic processes to enhance power consumption.

AI Behavioral Assurance establishes a process-level methodology for governing the behavioral outputs of Large Language Models and agentic AI systems in consequential decision environments. Its structu…

AI Behavioral Assurance establishes a process-level methodology for governing the behavioral outputs of Large Language Models and agentic AI systems in consequential decision environments. Its structural claim is that LLM and agentic systems satisfy the necessary and sufficient conditions to be treated as governable stochastic processes under Statistical Process Control, and that the Lean Six Sigma toolkit developed for manufacturing can be systematically translated to AI governance. The work adapts the five-phase DMAIC cycle (Define, Measure, Analyze, Improve, Control) to non-normal, non-stationary, unbounded, and opaque AI processes, using an Adopt, Adapt, Innovate taxonomy that classifies 40 tools and constructs by their applicability. It introduces a purpose-built measurement system, the BME (Behavioral Measurement of Entropy) Metric Suite, and a governance architecture, ALAGF (Adaptive Lifecycle Agentic Governance Framework), that embeds DMAIC as a repeatable lifecycle process. Behavioral entropy serves as the governing process variable. The methodology is mapped to ISO/IEC 42001:2023, the NIST AI Risk Management Framework 1.0, IEEE standards, and the EU AI Act, and is demonstrated through worked clinical, insurance-triage, and agentic use cases. The book targets standards bodies, governance researchers, and practitioners deploying AI in regulated environments.

The Collaborative Online International Learning (COIL) design for a community of practice, as outlined in WP4.3 of the ELCRA project, is titled: Environmental Literacy and Collective Action in Europe:…

The Collaborative Online International Learning (COIL) design for a community of practice, as outlined in WP4.3 of the ELCRA project, is titled: Environmental Literacy and Collective Action in Europe: A Collaborative Online International Learning Experience.  The COIL, as part of the work in the development of the ELCRA project, is structured according to the following parameters: Five participating universities (Complutense University of Madrid, Jagiellonian University, University of Coimbra, Klaipėda University and University of Siena); Fifty students (ten from each university participating in the project); Five modules (one per university); Six hours of dedication per module and Final collaborative product in Genially.

This deposit contains the frozen analysis chain and verification package for the manuscript "Municipal Fire Incidents as a Source of Regional Emission Burden: A Screening-Level Assessment Using Routin…

This deposit contains the frozen analysis chain and verification package for the manuscript "Municipal Fire Incidents as a Source of Regional Emission Burden: A Screening-Level Assessment Using Routine Fire Service Records (Kocaeli, Türkiye)". Contents: the frozen damage-severity parser (specification v1.3, rule dictionary and executable code), the burned-area parser v1.1 with its synthetic test suite, the pre-registered adjudication analysis plan and its binding stopping rule, the blind adjudication outputs, category allocation anchors and parameters, the asymmetric emission-factor parameterisation with a machine-readable provenance record, the residual-dependence scenario results, a data dictionary, an access statement, pinned environment specifications and SHA-256 checksums for every content file. Verification: running "python code/reproduce_all.py" from the archive root recomputes 47 published quantities from the deposited derived data, compares each against its published value, and writes verification_log.txt. The archived log records 47 checks with 0 failures. All stochastic analyses use a fixed random seed of 2026. Exact package versions are pinned in requirements.txt. Not included: the raw incident narratives from the municipal operational records, which contain identifiable personal details and are therefore withheld. The reasons and the controlled-access route are set out in ACCESS_STATEMENT.md. The damage-severity parser test suite depends on those narratives and is published as expectations plus input digests only. Licensing note: the data, documentation and derived files are released under CC BY 4.0; the code files in code/ may be reused under the MIT License.

Effective fungal control is essential to prevent spoilage, contamination, and infections in the realms of food, healthcare, and industry. This study explores the antifungal activity of plasma-activate…

Effective fungal control is essential to prevent spoilage, contamination, and infections in the realms of food, healthcare, and industry. This study explores the antifungal activity of plasma-activated water (PAW) generated by transient spark (TS) electrical discharge combined with natural phenolic bioactive compounds on the planktonic growth, biofilm formation, and surface adhesion of yeast Saccharomyces cerevisiae and Wickerhamomyces anomalous. The phenolics were assayed at 1 or 2 mg/mL for planktonic growth after determining their minimum inhibitory concentration. The biofilms were grown on sterile microscope slides for 48 h, stained with acridine orange or calcofluor stain, and observed under a fluorescence microscope. PAW was used immediately after plasma treatment of tap water to create PAW-phenolic solutions. The combination of PAW with cinnamic acid and vanillin showed the most significant antiyeast activity against planktonic and biofilm growth, as well as surface adhesion. The biofilms formed with the PAW-phenolic combination were notably fragmented, and residual cells were structurally damaged. A complete inhibition of biofilm formation was observed when 500 μg/mL of cinnamic acid was used in combination with PAW. The findings indicate a strong potential of combining PAW with phenolic compounds that represents an effective novel strategy for preventing fungal growth.

First Edition