Transparency, trust, and accountability are among the issues raised by artificial intelligence's (AI) growing reliance on black-box models, especially in high-stakes industries like healthcare, finance, and criminal justice. These models, which are frequently distinguished by their intricacy and opacity, are capable of producing extremely accurate forecasts, but users and decision-makers are still unable to fully understand how they operate. In response to this challenge, the field of Explainable AI (XAI) has emerged with the goal of demystifying these models by offering insights into their decision-making processes. Our ability to interpret model behavior has greatly improved with recent developments in XAI techniques, such as SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-agnostic Explanations), and counterfactual explanations. These instruments make it easier to recognize bias, promote trust, and guarantee adherence to moral principles and laws like the GDPR and the AI Act. Modern XAI techniques are reviewed in this research along with how they are used in moral decision-making. It looks at how explainability can improve fairness, reduce the risks of AI bias and discrimination, and assist well-informed decision-making in a variety of industries. It also examines the trade-offs between performance and interpretability of models, as well as the growing trends toward user-centric explainability techniques. In order to ensure responsible AI development and deployment, XAI's role in fostering accountability and transparency will become increasingly important as AI becomes more integrated into critical systems.

Autonomous vehicles have gained popularity in recent years, but they are still not compatible with other vulnerable components of the traffic system, including pedestrians, bicyclists, motorcyclists, and occupants of smaller vehicles such as passenger cars. This incompatibility leads to reduced system performance and undermines traffic safety and comfort. To address this issue, the authors considered pedestrian crosswalks where vehicles, pedestrians, and micro-mobility vehicles collide at right angles in an urban road network. These road sections are areas where vulnerable people encounter vehicles perpendicularly. In order to prevent accidents in these areas, it is planned to introduce a warning system for vehicles and pedestrians. This procedure consists of multi-stage activities by sending warnings to drivers, disabled individuals, and pedestrians with phone addiction simultaneously. This collective autonomy is expected to reduce the number of accidents drastically. The aim of this paper is the automatic detection of a pedestrian crosswalk in an urban road network, designed from both pedestrian and vehicle perspectives. Faster R-CNN (R101-FPN and X101-FPN) and YOLOv7 network models were used in the analytical process of a dataset collected by the authors. Based on the detection performance comparison between both models, YOLOv7 accuracy was 98.6%, while the accuracy for Faster R-CNN was 98.29%. For the detection of different types of pedestrian crossings, YOLOv7 gave better prediction results than Faster R-CNN, although quite similar results were obtained.

One of the solutions to reduce environmental emissions is related to the deployment of electric vehicles (EVs) with sustainable energy. In order to be able to increase the number of electric vehicles in circulation, it is important to implement optimal planning and design of the infrastructure, with particular reference to areas equipped with charging stations. The suitable analysis of the location of current electric vehicle charging stations (EVCSs) is the central theme of this document. The research focused on the actual location of the charging stations of five major EVCS companies in the province by selecting Istanbul as the study area. The study was conducted through a three-step approach and specifically (i) the application of the analytical hierarchy process (AHP) method for creating the weights of the 6 main and 18 secondary criteria that influence the location of EVCSs; (ii) a geospatial analysis using GIS considering each criterion and developing the suitability map for the locations of EVCSs, and (iii) application of the technique for order preference by similarity to ideal solution (TOPSIS) to evaluate the location performance of current EVCSs. The results show that the ratio between the most suitable and unsuitable areas for the location of EVCSs in Istanbul and the study area is about 5% and 4%, respectively. The results achieved means of improving sustainable urban planning and laying the basis for an assessment of other areas where EVCSs could be placed.

Using artificial intelligence (AI), machine learning (ML), deep learning (DL), and conversational models like ChatGPT, smart farming is revolutionizing the agricultural industry by increasing productivity, cutting down on resource usage, and improving decision-making. Critical agricultural problems including crop monitoring, pest identification, weather forecasting, and soil analysis can be resolved with the help of these technologies. Predictive analytics is made possible by AI and ML algorithms, which enhance crop yield by foreseeing disease outbreaks and maximizing planting schedules. With sophisticated image processing, deep learning models (DL models) enable real-time monitoring of livestock and crops, providing detailed information for precision farming. Smart farming is being further enhanced by ChatGPT and other AI-driven conversational agents. These agents offer real-time advisory services, make it possible for farmers to communicate with AI tools using natural language, and streamline difficult tasks like supply chain management, market analysis, and crop selection. Future developments in smart farming include the integration of AI with IoT devices, blockchain technology for traceability, and improved edge computing capabilities to facilitate localized, real-time decision-making.

In the age of massive datasets and real-time applications, scalable and adaptive deep learning algorithms are critical to meeting the ever-increasing demands of large-scale machine learning (ML) systems. The state-of-the-art developments in scalable deep learning methods are examined in this research, with particular attention paid to architectural breakthroughs that facilitate effective model training, adaptive learning, and inference across distributed systems. It is emphasized that contemporary algorithms—like distributed gradient descent optimization, model parallelism, and sophisticated reinforcement learning techniques—are essential for controlling the complexity of big datasets without compromising performance. The research also explores how resource optimization and auto-scaling mechanisms work together, which is crucial for reducing computational overhead in cloud-based machine learning systems. It is highlighted that adaptive models—which can modify their architecture in response to patterns in input data and changes in the surrounding environment—are essential for maintaining robustness and flexibility. High-dimensional data, dynamic workload allocation, and latency minimization in real-time learning tasks are among the scalability challenges tackled. A closer look at more recent frameworks like Federated Learning, which makes it easier for decentralized model training across edge devices, shows how promising these scalable methods can be for privacy-preserving applications. The areas include automated machine learning (AutoML), hyperparameter tuning, and self-supervised learning.