847 records: the first group here; choose another one above.

ongoing project 65
@courseproject{hyroxcoach-ise2526,
    apice = {HyroxCoach},
    author = {Zavatta, Andrea},
    howpublished = {Intelligent Systems Engineering, A.Y. 2025/2026},
    month = aug,
    sort = {classic},
    supervisor = {Ciatto, Giovanni},
    title = {HYROX COACH},
    type = {Standard Project},
    year = 2026
}
@courseproject{packagedeliverysimulation-ise2526,
    apice = {PackageDeliverySimulation},
    author = {Pengyue Xu},
    howpublished = {Intelligent Systems Engineering, A.Y. 2025/2026},
    month = aug,
    sort = {classic},
    supervisor = {Ciatto, Giovanni},
    title = {Package Delivery Simulation},
    type = {Standard Project},
    year = 2026
}
@courseproject{gametheorysimulationlaboratory-ise2526,
    apice = {GameTheorySimulationLaboratory},
    author = {Blagoja Savevski},
    howpublished = {Intelligent Systems Engineering, A.Y. 2025/2026},
    month = jul,
    sort = {advanced},
    supervisor = {Ciatto, Giovanni},
    title = {Game theory simulation laboratory},
    type = {Advanced Project},
    year = 2026
}
@courseproject{minietcd-ds2526,
    apice = {MiniEtcd},
    author = {Christian Serra},
    howpublished = {Distributed Systems, A.Y. 2025/2026},
    month = jul,
    sort = {application},
    supervisor = {Matteo Magnini},
    title = {mini-etcd: a Raft-based key-value store},
    type = {Application Project},
    url = {https://github.com/castfx/mini-etcd},
    year = 2026
}
@courseproject{taiyomas-ise2526,
    abstract = {TAIYO-MAS is a intelligent system for photovoltaic energy management. The project starts from a discrete time-step simulation of a small energy microgrid consisting of a photovoltaic installation, battery storage, household consumption, and the electrical grid. Solar production varies based on weather conditions, while energy consumption changes dynamically. The behavior of the system is influenced by triggered operating modes (Balanced, Direct, Selling) that explicitly order how energy distribution is prioritized. Core structure: Kotlin-based energy simulation engine. Energy state monitoring module. Desktop GUI built with Kotlin and JavaFX (for real-time visualization of energy flows, component statuses, and agent decisions). Technical parts: Agent-Oriented Energy Management, Jason / AgentSpeak -> Solar, battery, load, and grid entities will be modeled as autonomous BDI agents that evaluate system states and coordinate via message passing. Operating Modes -> System-level strategies (Balanced, Direct, Selling) are externalized into explicit modes that activate different behavioral patterns and guide agent goals. Automated Planning, STRIPS -> A lightweight STRIPS planner will be implemented in Kotlin to dynamically generate sequences of corrective actions during critical or inefficient conditions.},
    apice = {TaiyoMAS},
    author = {Jiahao Guo},
    howpublished = {Intelligent Systems Engineering, A.Y. 2025/2026},
    month = jun,
    sort = {classic},
    supervisor = {Ciatto, Giovanni},
    title = {TAIYO-MAS},
    type = {Standard Project},
    url = {https://github.com/schwein69/Taiyo-Mas},
    year = 2026
}
@courseproject{shippingontheair2026-ise2526,
    apice = {ShippingontheAir2026},
    author = {Horri, Kalile and Bel Meddeb, Intisar},
    howpublished = {Intelligent Systems Engineering, A.Y. 2025/2026},
    month = jun,
    sort = {classic},
    supervisor = {Omicini, Andrea},
    title = {Shipping on the Air},
    type = {Standard Project},
    url = {https://github.com/intissar99/ShippingOnTheAir-ISE},
    year = 2026
}
@courseproject{litedb-dss2526,
    apice = {LiteDB},
    author = {Luca Marchi},
    howpublished = {Distributed Software Systems, A.Y. 2025/2026},
    month = jun,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {LiteDB a distributed key-value store},
    type = {Standard Project},
    year = 2026
}
@courseproject{factorymanagerdistributedmonitoringsystem-ds2526,
    apice = {FactoryManagerDistributedMonitoringSystem},
    author = {David Cohen and Giulia Nardicchia},
    howpublished = {Distributed Systems, A.Y. 2025/2026},
    month = may,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Factory Manager Distributed Monitoring System},
    type = {Standard Project},
    year = 2026
}
@courseproject{forestguard2026-ise2425,
    abstract = {ForestGuard is a simulation of a forest monitoring and fire fighting system. The agents in this simulation are represented by scout drones, firefighter drones, and a forest ranger station. Scout Drone: Scout drones are responsible for exploring the forest and mapping the environment. Their main task is detecting active fires and reporting them to nearby agents or directly to the station when in communication range. Scouts must continuously monitor their battery level and return to the station to recharge when it runs low. Firefighter Drone: Firefighter drones are dispatched to locations where fires have been reported. Their main task is to extinguish active fires using their onboard water tank. When the tank runs empty, they must decide where to refill: if their battery level is sufficient, they head to the nearest water source (river or lake); otherwise, they return to the ranger station to both recharge and refill. Fire suppression is always their top priority, and resource management is key to their effectiveness. Forest Ranger Station: The station is a stationary agent placed at a fixed location in the grid. It acts as the central hub of the system: it recharges drone batteries, refills the firefighter water tanks, and aggregates map information reported by drones when they are within communication range. It can also redirect drones to high priority areas based on the global map it maintains. The environment is represented by a randomly generated 2D grid forest, which contains trees, clearings, obstacles, fire cells, and water sources such as rivers or lakes. Fire spreads dynamically to adjacent cells every few turns. Drones can communicate with each other only when within short antenna range, while the station can reach any drone at any time.},
    apice = {ForestGuard2026},
    author = {Rambaldi, Riccardo},
    howpublished = {Intelligent Systems Engineering, A.Y. 2024/2025},
    month = may,
    sort = {simulation},
    supervisor = {Omicini, Andrea},
    title = {ForestGuard},
    type = {Simulation Project},
    url = {https://github.com/sbrambalds/ForestGuard},
    year = 2026
}
@courseproject{munchies-ds2425,
    apice = {Munchies},
    author = {Alessandro Agosta and Tiberiu Norbert Gabos and Emiliano Rattini},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = mar,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Munchies},
    type = {Standard Project},
    url = {https://github.com/Norby99/munchies},
    year = 2026
}
@courseproject{realemas2425-mas2425m2,
    abstract = {This project investigates the transition from static optimization frameworks for urban greening toward a dynamic, learning-based decision-making process supported by agent-based modeling. The overarching goal is to design adaptive strategies that guide the spatial and temporal distribution of green infrastructure to enhance environmental equity and urban climate resilience over medium to long time horizons. By reframing urban greening as a sequential decision-making problem, the proposed framework enables the simulation of evolving intervention policies and their cumulative environmental and social impacts. The integration of autonomous agents allows both environmental processes (e.g., vegetation growth and heat dynamics) and human behaviors (e.g., demographic evolution, mobility, and socio-economic adaptation) to be represented as interacting components within a complex urban ecosystem. This approach supports the development of adaptive, data-driven policies capable of learning from feedback and optimizing long-term outcomes under uncertainty.},
    apice = {RealeMas2425},
    author = {Simone Reale},
    howpublished = {Multi-Agent Systems (Module 2), A.Y. 2024/2025},
    month = oct,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {AGENT-BASED SEQUENTIAL OPTIMIZATION FOR FAIRNESS IN URBAN GREEN PLANNING},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/realemas2425},
    year = 2025
}
@courseproject{dascenzogentili2425-ethicsofai20242025,
    abstract = {This project extends previous work by our colleagues Merola, Sigh, and Dardouri on evaluating LLM performance in planning tasks and developing a meta-network for reliability prediction. The proposed extensions focus on two directions: (1) exploring emergent reasoning capabilities across a broader range of open-source large language models, including Mistral, LLaMA, and DeepSeek, to assess whether planning abilities generalize across architectures; and (2) enriching the benchmark dataset with new, challenging planning problems of graded complexity. Together, these extensions aim to deepen our understanding of LLM reasoning, improve evaluation reliability, and foster reproducible benchmarks for the research community.},
    apice = {DascenzoGentili2425},
    author = {Alessandro Gentili},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = sep,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Reasoning in LLMs},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/dascenzogentili2425},
    year = 2025
}
@courseproject{adragnamancino2425-ethicsofai20242025,
    abstract = {This project is carried out in the framework of the Aequitas initiative, which develops methodologies and tools for auditing, assessing and improving fairness of AI systems across domains such as recruitment, healthcare, and education. Among the six use cases identified by Aequitas, we focus on the education use case.},
    apice = {AdragnaMancino2425},
    author = {Giorgio Adragna and Marcello Mancino},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = sep,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Kernel-Based HGR in Fairness: a case study},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/adragnamancino2425},
    year = 2025
}
@courseproject{cremonininanni2425-ethicsofai20242025,
    abstract = {This project aims to analyse the impact of socio-economic indices on the prediction of students’ school performances and ensure fairness in this process. The context a student is living in can affect their performance, but this should not be considered a deciding factor. Given the sensitive nature of these features, it is essential to avoid any form of bias in the training of predictive models that utilise this data. To investigate the relevance of these factors, the project will consider a real case-study dataset about students’ performance and social inequality. Benchmark classifiers will be trained, deploying mitigation techniques to reduce the effects of biases in the dataset and ensure a fair evaluation prediction is achieved. The results will help highlight connections between disadvantaged student situations and their performance, providing useful insight to ensure equality.},
    apice = {CremoniniNanni2425},
    author = {Davide Cremonini and Gabriele Nanni},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = aug,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {School performance prediction: Ensuring fairness for disadvantaged students},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/cremonininanni2425},
    year = 2025
}
@courseproject{naveedfarokhpay2425-ethicsofai20242025,
    abstract = {Traditional task allocation systems aim to maximize efficiency (e.g. minimize cost or com- pletion time) but often ignore fairness in workload distribution. This can lead to ethical issues such as agent overburdening or uneven reward sharing. In this project, we study fairness-aware static task assignment by augmenting a basic cost matrix with a fairness regularizer based on Jain’s index. Unlike dynamic agent-based simulations, we consider a static setup where tasks are distributed among agents at once. We aim to evaluate trade offs between overall efficiency and equity across different levels of fairness regularization.},
    apice = {NaveedFarokhpay2425},
    author = {Abdullah Naveed and Nadia Farokhpay},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = aug,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness-Aware Static Task Assignment: Toward Ethical Resource Allocation in Multi-Agent Systems},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/naveedfarokhpay2425},
    year = 2025
}
@courseproject{baiocchipetrilli2425-ethicsofai20232024,
    abstract = {LLMs can be subject to cultural biases. One such domain is recruiting, where LLMs are used to extract information from candidates’ CVs and then to select the most suitable profiles. These biases may favour or penalize applicants based on language or cultural cues, risking the exclusion of qualified talent.},
    apice = {BaiocchiPetrilli2425},
    author = {Francesco Baiocchi and Leonardo Petrilli},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    month = aug,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness in AI Recruiting Systems},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/baiocchipetrilli2425},
    year = 2025
}
@courseproject{izvoreanu2425-mas2425,
    abstract = {This project aims to design, implement, and evaluate a smart traffic management system using JADE. Each traffic intersection will be modeled as an autonomous JADE agent controlling a traffic light. These agents will reason about local traffic conditions and coordinate with neighboring agents to optimize traffic flow. The project emphasizes decentralized decision-making and coordination within a complex distributed environment, illustrating key multi-agent system concepts such as autonomy, communication, coordination, and self-organization. The goal is to showcase how agent-oriented technologies can manage real-world inspired challenges like urban traffic optimization.},
    apice = {Izvoreanu2425},
    howpublished = {Multi-Agent Systems, A.Y. 2024/2025},
    month = aug,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Smart Traffic Light Control Using Multi-Agent Systems},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/izvoreanu2425},
    year = 2025
}
@courseproject{teodoranirossi2526-ethicsofai20242025,
    abstract = {This project investigates the development of a hybrid AI system designed to support the evaluation of adoption eligibility. The system combines a supervised machine learn- ing classifier trained on a simulated dataset of prospective adoptive applicants with explainable AI (XAI) methods to highlight the most relevant factors influencing indi- vidual predictions. The dataset is synthetically generated but grounded in real-world constraints: both the choice of variables and the statistical distributions are based on international adop- tion standards, institutional guidelines, and relevant academic literature. To promote transparency and user trust, the system incorporates a large language model (LLM), which translates the XAI-generated insights into natural language expla- nations that are understandable, neutral, and ethically framed. These explanations aim to empower human decision-makers and applicants alike, enhancing human agency and oversight in algorithmic judgments. Beyond the prototype, our goal is to deliver a functional platform where users can input their personal data corresponding to selected variables and receive a decision outcome accompanied by an interpretable explanation generated by the LLM. This in- teractive system would simulate real-world adoption scenarios while demonstrating the potential of combining XAI and LLMs for transparent, user-centered decision support.},
    apice = {Teodoranirossi2526},
    author = {Filippo Teodorani and Diego Rossi},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = aug,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {A Hybrid AI System for Adoption Eligibility Assessment with Explainable Feedback},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/teodoranirossi2526},
    year = 2025
}
@courseproject{wiredwolf-ds2425,
    apice = {WiredWolf},
    author = {Nicolò Guerra and Emma Leonardi and Filippo Casadei},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = jul,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {WiredWolf},
    type = {Standard Project},
    url = {https://github.com/nicolo09/WiredWolf},
    year = 2025
}
@courseproject{chainreactiongame-ds2425,
    apice = {ChainReactionGame},
    author = {Keith Barnes and Seyyedeh Fatemeh Kalati},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = jul,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Chain Reaction Game},
    type = {Standard Project},
    url = {https://github.com/kthbarnes/Distributed_Chain_Reaction},
    year = 2025
}
@courseproject{gualandipresepisciarrillo2425-ethicsofai20242025,
    abstract = {The growing integration of deep learning models in critical domains has brought urgent attention to their vulnerabilities. One of the most concerning threats is adversarial attacks on image-based AI systems: subtle perturbations to input images that can mislead even the most accurate models into making incorrect predictions, often without detection. These attacks challenge not only the technical robustness of AI models but also raise significant ethical concerns, particularly when human safety, fairness, and accountability are at stake. In this project, we explore the nature of adversarial attacks on image data, aiming to deepen our understanding of their mechanisms and implications. We review and categorize popular attack techniques, evaluate their impact on standard vision models, and assess the effectiveness of different models we designed and trained to counteract the considered attacks. Furthermore, we developed and tested different versions of a module capable of recognizing when an input image has been corrupted. Finally, the analysis is extended to practical case studies in face recognition and vehicle detection, where we investigate how adversarial attacks can compromise security and fairness in real-world applications. By framing the problem from both a technical and ethical perspective, we highlight the importance of designing AI systems that are not only performant but also resilient and trustworthy under adversarial conditions.},
    apice = {GualandiPresepiSciarrillo2425},
    author = {Mattia Gualandi and Presepi Alex and Sciarrillo Alessandro},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Exploit, Defend, and Detect Adversarial Attacks on Image-Based AI Systems: An Ethical Perspective},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/gualandipresepisciarrillo2425},
    year = 2025
}
@courseproject{genovesetutone2425-ethicsofai20242025,
    abstract = {AEQUITAS [1] is a framework that aims to develop more fair AI systems. In order to validate this framework, various cases are considered, covering distinct domains representative of industrial and digital sectors (hiring), public bodies (health), and sensitive social issues. With this project, we are considering the use case of a recruiting system. The software analyzed is used by the Adecco Group[2] in the candidate selection process. The software supports the recruitment process by recommending the best candidates for a given job position and, the way around, suggesting the most suitable positions for each candidate. This software is based on an AI system that extracts relevant information from each candidate and standardizes it against structured data stored in the internal database. By comparing the attributes of the candidate with the re- quirements of the job position, the software generates a ranked list of candidates based on their compatibility. The results are then presented to the recruiter},
    apice = {GenoveseTutone2425},
    author = {Alessandro Tutone and Alberto Genovese},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {AEQUITAS : Bias free AI assisted recruiting system},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/genovesetutone2425},
    year = 2025
}
@courseproject{chauvetethics2425-ethicsofai20242025,
    abstract = {This project explores the legal and ethical responsibilities arising from accidents involving autonomous vehicles. As autonomous vehicles are increasingly deployed, understanding liability becomes essential. Traditional notions of human responsibility are being chal- lenged by the rise of AI-based decision-making systems.},
    apice = {ChauvetEthics2425},
    author = {Nicolas Chauvet},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Responsibility in autonomous vehicle accidents : An ethical and legal challenge},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/chauvetethics2425},
    year = 2025
}
@courseproject{castellifratinimone2425-ethicsofai20242025,
    abstract = {Bias in artificial intelligence algorithms used in healthcare is a growing concern due to its potential to negatively impact patient outcomes. AI tools, despite their power, can perpetuate systematic unfairness towards certain groups. This is particularly critical in healthcare, where existing biases can exacerbate inequalities. Historical examples, such as the Fram- ingham Risk Score, demonstrate how biases in traditional medical tools have led to variable accuracy across different ethnic groups and genders, due to the underrepresentation of diverse cohorts in their development. Women, for instance, experience higher rates of missed and delayed car- diovascular disease diagnoses, partly because of their underrepresentation in clinical trials. While some level of bias in data and tools is inevitable, the validation of AI tools to minimize bias offers a promising opportunity to reduce disparities in healthcare outcomes.},
    apice = {CastelliFratiniMone2425},
    author = {Giorgia Castelli and Alice Fratini and Madalina Ionela Mone},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Bias Mitigation in AI Models for Cardiovascular Diseases Prediction},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/castellifratinimone2425},
    year = 2025
}
@courseproject{kefalialiworeta2425-ethicsofai20242025,
    abstract = {Emotion recognition systems are increasingly deployed in real-world applications, including re- cruitment, education, and healthcare. However, many such models rely heavily on neurotypical standards of emotional expression, which can lead to biased or incorrect classifications when applied to neurodivergent individuals—particularly those on the autism spectrum—who may exhibit atypical facial affect, reduced eye contact, or lower expressivity [1]. This project aims to audit the fairness and inclusivity of facial emotion recognition models with respect to neurodivergent expression profiles. We will combine a systematic literature re- view on affective computing bias and neurodiverse expressivity [1, 3], with practical evaluation of pre-trained emotion recognition models (e.g., FER+, DeepFace, Microsoft Azure Emotion API) on standard datasets. Using explainability tools such as Grad-CAM [2], occlusion sensi- tivity, and counterfactual analysis via LIME, we will uncover how these models make decisions and whether their logic aligns with diverse expression styles. Our core contribution is the development of a lightweight auditing protocol and fairness checklist, designed to help developers identify potential exclusion risks in affective AI systems. This work emphasizes ethical alignment, robustness, and inclusivity in emotion AI without requiring potentially biased image manipulations.},
    apice = {KefaliAliWoreta2425},
    author = {Luwam Major Kefali and Hassen Said Ali and Hilina Fissha Woreta},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Smile̸ = Happy: Auditing Emotion Recognition Systems for Bias Against Neurodiverse Individuals},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/kefalialiworeta2425},
    year = 2025
}
@courseproject{turekhassim24125-ethicsofai20242025,
    abstract = {This project investigates how preference-based alignment using Direct Preference Optimization (DPO) (Rafailov et al., 2023) impacts both the performance and ethical behavior of a large language model (LLM). Specifically, we fine-tune the instruction-tuned dolly-v2-3b model (Databricks, 2023) using Anthropic’s Helpful-Harmless (HH-RLHF) dataset (Bai et al., 2022). The goal is to evaluate whether such alignment hurts the performance of a LLM on standard NLP tasks in comparison with pre-alignment-tuned model. We assess the model using both capability benchmarks (ARC-Easy, ARC-Challenge, HellaSWAG, BoolQ) and fairness/safety benchmarks (Anthropic HH, RealToxicityPrompts, BiasBench).},
    apice = {Turekhassim24125},
    author = {Turekhassim, Abylay},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Evaluating the Impact of Alignment via DPO on Performance and Harmfulness in LLMs},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/turekhassim2425},
    year = 2025
}
@courseproject{maschauvet2425-mas2425,
    abstract = {The project aims to develop and compare two fruit collection strategies for autonomous robots operating in an orchard. These robots, acting as multi-agents (MAS), must collect apples and deposit them into a collector efficiently. The objective is to evaluate the performance of both strategies under different conditions and understand their respective strengths and limitations.},
    apice = {MASChauvet2425},
    author = {Nicolas Chauvet},
    howpublished = {Multi-Agent Systems, A.Y. 2024/2025},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Apple trees simulation},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/maschauvet2425},
    year = 2025
}
@courseproject{fasulobabboniomelchenko2425-ethicsofai20242025,
    abstract = {Deep learning models in medical image classification are critically vulnerable to adversar- ial attacks, imperceptible input manipulations causing misclassifications with severe clinical consequences [1]. This project addresses this by systematically evaluating attack impacts (e.g., Fast Gradient Sign Method (FGSM)) on a ResNet[2] model using public medical datasets (e.g., PatchCamelyon). We will develop and benchmark robust countermeasures, including adversarial training and advanced input purification techniques. Our comprehen- sive benchmarking framework will measure how defense techniques impact model perfor- mance across multiple dimensions: classification accuracy on clean and adversarial samples, computational overhead during training and inference, memory requirements, and processing latency. These metrics will provide a holistic assessment of the trade-offs between security, performance, and computational efficiency in clinical deployment scenarios. Our aim is to provide practical strategies for fortifying medical image classifiers, enhancing their safety, reliability, and ethical integrity in clinical settings.},
    apice = {FasuloBabboniOmelchenko2425},
    author = {Matteo Fasulo and Luca Babboni and Maksim Omelchenko},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Ensuring Trustworthy Medical AI: Enhancing Robustness and Ethical Integrity of Image Classifiers Against Adversarial Attacks},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/fasulobabboniomelchenko2425},
    year = 2025
}
@courseproject{leo2425-mas2425,
    abstract = {The project simulates a research team composed of different types of robotic agents that communicate to accelerate the search process. The agents are developed using Jason, a platform based on the BDI (Belief-Desire-Intention) architecture, which is well-suited to this problem and allows a straightforward environment implementation. The simulated environment is a 2D map populated with different objects: obstacles, walkable terrain and several Points of Interest (POIs). The target (an item or entity to be found) is randomly located within one of these POIs. Agents are divided into at least two types: a “ground” unit for the environment interaction, and a “drone” unit for a faster exploration. A team of agents (whose size scales with the map's dimensions) must efficiently explore the environment and locate the target. Additional features, discussed below, may be included depending on how much time is required for the implementation.},
    apice = {Leo2425},
    author = {Paolo Leo},
    howpublished = {Multi-Agent Systems, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Multi-Agent Exploration in a Simulated Environment},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/leo2425},
    year = 2025
}
@courseproject{fatemehhamidali2425-ethicsofai20242025,
    abstract = {Digital twins are increasingly used in smart city planning to simulate the effects of urban policy inter- ventions. This project investigates the ethical implications of various mobility policies simulated within Bologna’s digital twin. We divide the city into zones based on statistical and administrative boundaries, assigning each area socio-demographic indicators such as gender distribution, economic vulnerability, and access to public infrastructure. We then simulate different mobility scenarios—including expanded pub- lic transit, bike lane prioritization, and car-restricted zones—to analyze their differential effects. Using AI ethics frameworks, we evaluate these scenarios in terms of fairness, transparency, and inclusiveness, aiming to propose mobility policies that are both effective and equitable.},
    apice = {FatemehHamidAli2425},
    author = {Fatemeh Izadinejad and Ali Loloeejahromi and Hamidreza Bahmanyar},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Ethical Simulation of Urban Mobility in Bologna’s Digital Twin},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/fatemehhamidali2425},
    year = 2025
}
@courseproject{marchibaytchevadiiulio2425-ethicsofai20242025,
    abstract = {The legal profession remains one of the least diverse in the U.S., and the bar exam may play a role in that. Exploiting a Law School Admissions dataset from the Law School Admissions Council (LSAC) [1], we aim to explore how racial and ethnic disparities manifest in law school bar passage rates. Using publicly available data, we plan to examine whether schools with higher percentages of students from underrepresented racial and ethnic backgrounds tend to have lower first-time bar exam pass rates, even when controlling for academic and institutional factors. Our goal is to analyze potential sources of bias, test machine learning models for fairness, and explore strategies to reduce algorithmic and systemic inequities.},
    apice = {MarchiBaytchevaDiIulio2425},
    author = {Federico Marchi and Christina Baytcheva and Anna Di Iulio},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Bias in Law School Admissions: Analysis and Mitigation Using Machine Learning},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/marchibaytchevadiiulio2425},
    year = 2025
}
@courseproject{clotanizvoreanugravina2425-ethicsofai20242025,
    apice = {ClotanIzvoreanuGravina2425},
    author = {Paul Clotan and Valeria Izvoreanu and Antonio Gravina},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness in Multilingual Sentiment Analysis using DistilBERT},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/clotanizvoreanugravina2425},
    year = 2025
}
@courseproject{marken2425-mas2425,
    abstract = {The goal of this project is to develop a simulation environment in which multiple trading agents interact with a centralized exchange agent to buy and sell stocks. The simulation is designed to support the creation and testing of diverse agent-based trading strategies, providing a flexible and extensible framework for experimentation. A key design principle is modu- larity—making it straightforward to implement new agents with novel be- haviors, policies, or learning techniques. The environment enables users to evaluate the performance of individual agents by analyzing the evolution of their portfolios over time, comparing risk-adjusted returns, volatility, and other financial metrics. Through these comparisons, the simulation can be used to assess which strategies perform best under varying market conditions, or how agents might adapt to different trading environments.},
    apice = {Marken2425},
    author = {Andreas Marken},
    howpublished = {Multi-Agent Systems, A.Y. 2024/2025},
    month = may,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Multi agent AI stock environment simulation},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/marken2425},
    year = 2025
}
@courseproject{blackjackv2-ds2425,
    apice = {BlackjackV2},
    author = {Juri Guglielmi and Nanni, Denise},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = feb,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Blackjack},
    type = {Standard Project},
    url = {https://github.com/dennnanni/BlackJack},
    year = 2025
}
@courseproject{3dshare-ds2425,
    apice = {3DShare},
    author = {Riccardo Pepe},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = jan,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {3DShare},
    type = {Standard Project},
    url = {https://github.com/rickpepp/3DShare},
    year = 2025
}
@courseproject{modsmartscala-ds2425,
    apice = {ModSmartScala},
    author = {Bryan Corradino},
    howpublished = {Distributed Systems, A.Y. 2024/2025},
    month = dec,
    sort = {technology},
    supervisor = {Magnini, Matteo},
    title = {Study and implementation of the Mod-SMaRt algorithm},
    type = {Technology Project},
    url = {https://github.com/Remisse/scala-mod-smart},
    year = 2024
}
@courseproject{terenziani2324-ethicsofai20232024,
    abstract = {LIME[1] and SHAP[2] are two approaches to extracting feature importances for any Machine Learning model, the former using a proxy explainable model (like a decision tree) and the latter using a game-theory driven method based on Shapley values. The idea of this project is to solve a classification task with different models, computing a LIME and SHAP explanation for each, then compare those with a ”baseline explanation” obtained directly from a decision tree or random forest model. This comparison aims at finding possible links between a model’s performance and its explanations’ similarity to the baseline, and at determining how ”easliy”, if ever, LIME and SHAP indipendently compute similar importances for a given model.},
    apice = {Terenziani2324},
    author = {Terenziani, Andrea},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    month = jul,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {LIME vs SHAP Comparing ML explanation techniques},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/terenziani2324},
    year = 2024
}
@courseproject{adversfairnessrullo2024-mas2324,
    abstract = {Multi-agent reinforcement learning involves the study of how multiple agents interact within a shared environment and with each other to achieve a common goal. In cooperative environments, agents typically work together to maximize a collective reward. However, in such environments, the op- timization of actions can lead to more capable agents overshadowing less capable ones, as they strive to maximize the expected value of their actions using reinforcement learning techniques. The objective of this project is to address this imbalance by ensuring fairness during the training process. Ad- versarial learning techniques offer a promising approach by enforcing learned representations to be insensitive to specific attributes, thereby mitigating disparities and creating a fair representation and decision-making process among cooperative agents.},
    apice = {AdversFairnessRullo2024},
    author = {Federico Rullo},
    howpublished = {Multi-Agent Systems, A.Y. 2023/2024},
    month = jul,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Adversarial Learning for Fairness in Multi-Agent Reinforcement Learning: Mitigating Discrimination in Cooperative Environments},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/mas/projects/mas-rullo2324},
    year = 2024
}
@courseproject{marvasibalzanisaturno2324-ethicsofai20232024,
    abstract = {With the increasing use of Artificial Intelligence (AI) in various industries, ensuring fairness in these systems has become a critical theme. This work investigates the biases introduced by hiring algorithms employed by large corporations to prescreen job candidates. These algorithms are increasingly used to streamline the hiring process, making it more efficient and cost-effective. However, despite their advantages, these AI systems can perpetuate and even exacerbate existing biases, leading to unfair discrimination against certain groups. Our research aims to identify specific biases within these prescreening algorithms and propose methods to mitigate them. In our study we will compare traditional ML technique together with more advanced neural network architectures. Our aim is to demonstrate the potential presence of unfairness both in the dataset and in the decision-making process of the algorithms, and the possible solutions that can be implemented to correct this behavior. Our findings highlight the importance of transparency, accountability, and continuous monitoring in the deployment of AI systems in hiring, ensuring that they promote equality and diversity in the workplace.},
    apice = {MarvasiBalzaniSaturno2324},
    author = {Riccardo Marvasi and Edoardo Saturno and Lorenzo Balzani},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    keywords = {NN, unfariness metrics, explainability methods, unfariness in decision making},
    month = jun,
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness in Recruitment Decisions Algorithms},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/marvasibalzanisaturno2324},
    year = 2024
}
@courseproject{restaurantchainmanager-ds2324,
    apice = {RestaurantChainManager},
    author = {João Maria Baptista Gonçalves Zarcos and Duarte Miguel Da Cunha Terrível},
    howpublished = {Distributed Systems, A.Y. 2023/2024},
    month = dec,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Reservation and Administration of a Restaurant chain},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ds/projects/ds-project-terrivel-zarcos-ay2324},
    year = 2023
}
@courseproject{webappgestioneprogettiaziendali-sd2223,
    apice = {WebAppGestioneProgettiAziendali},
    author = {Vanessa Lauro},
    howpublished = {Distributed Systems, A.Y. 2022/2023},
    month = may,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Web app per la gestione dei progetti aziendali},
    type = {Standard Project},
    year = 2023
}
@courseproject{raftkvstore-sd2223,
    apice = {RaftKVStore},
    author = {Giovanni Macrelli},
    howpublished = {Distributed Systems, A.Y. 2022/2023},
    month = jan,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Raft KV Store},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ds/projects/ds-project-macrelli-ay2223},
    year = 2023
}
@courseproject{netcoderollback-sd2223,
    apice = {NetcodeRollback},
    author = {Samuele Turci},
    howpublished = {Distributed Systems, A.Y. 2022/2023},
    month = jul,
    sort = {classic},
    supervisor = {Magnini, Matteo},
    title = {Implementazione Netcode di tipo rollback},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ds/projects/ds-project-turci-ay2223},
    year = 2023
}
@courseproject{nftanddapp-sd2122,
    apice = {NftAndDapp},
    author = {Mancini, Alan and Matteo Brocca},
    howpublished = {Distributed Systems, A.Y. 2021/2022},
    month = jan,
    sort = {advanced},
    supervisor = {Ciatto, Giovanni},
    title = {Trading Virtual Goods -- Marketplace di NFT con aste in tempo reale e dApp su blockchain EOSIO},
    type = {Advanced Project},
    url = {https://gitlab.com/pika-lab/courses/ds/projects/ds-project-brocca-mancini-ay2122},
    year = 2022
}
@courseproject{cheatcardgame-sd2122,
    apice = {CheatCardGame},
    author = {Andrea Brighi},
    howpublished = {Distributed Systems, A.Y. 2021/2022},
    month = jan,
    sort = {game},
    supervisor = {Ciatto, Giovanni},
    title = {Cheat (multiplayer cards game)},
    type = {Game Project},
    url = {https://gitlab.com/pika-lab/courses/ds/projects/ds-project-brighi-ay2122},
    year = 2022
}
@courseproject{banktracing-sd2122,
    apice = {BankTracing},
    author = {Marco Chiavaroli and Davide Crisante},
    howpublished = {Distributed Systems, A.Y. 2021/2022},
    month = dec,
    sort = {application},
    supervisor = {Ciatto, Giovanni},
    title = {Applicativo di tracciamento transazioni bancarie},
    type = {Application Project},
    url = {https://gitlab.com/pika-lab/courses/ds/projects/ds-project-chiavaroli-crisante-ay2122},
    year = 2021
}
@courseproject{p2pdistributedml-sd2122,
    apice = {P2pDistributedML},
    author = {Filippo Barbari},
    howpublished = {Distributed Systems, A.Y. 2021/2022},
    month = dec,
    sort = {technology},
    supervisor = {Ciatto, Giovanni},
    title = {Decentralized Distributed Machine Learning over a Peer-to-Peer network},
    type = {Technology Project},
    url = {https://gitlab.com/pika-lab/courses/ds/projects/ds-project-barbari-ay2122},
    year = 2021
}
@courseproject{bathousemanagement-sd2122,
    apice = {BathouseManagement},
    author = {Luca Panariello},
    howpublished = {Distributed Systems, A.Y. 2021/2022},
    month = oct,
    sort = {application},
    supervisor = {Ciatto, Giovanni},
    title = {App per stabilimenti balneari},
    type = {Application Project},
    url = {https://gitlab.com/pika-lab/courses/ds/projects/ds-project-panariello-ay2122},
    year = 2021
}
@courseproject{freddigardenal2324-ethicsofai20232024,
    abstract = {Predictive modeling for sequential diagnoses is often tackled with deep learn- ing mechanisms, which can achieve promising performance but may lack inter- pretability and struggle with limited sample sizes, especially for rare diagnoses. To address these challenges, GRAM [2] (GRaph-based Attention Model) sup- plements electronic health records (EHR) with hierarchical information from medical ontologies. By representing medical concepts as a combination of their ancestors in the ontology via an attention mechanism, GRAM aims to improve interpretability and mitigate the impact of data insufficiency. This approach allows GRAM to adaptively generalize to higher-level concepts when facing lim- ited data at lower-level concepts, resulting in enhanced predictive performance and more meaningful representations aligned with medical knowledge. In the original work, GRAM was implemented using the Theano [6] framework, which is no longer actively maintained and has been surpassed by more modern deep learning libraries. In this study, we explore the GRAM method in depth and propose an updated implementation using the PyTorch [5] library. Our primary objectives are to replicate the embedding methods proposed in the original work and to create a flexible structure that allows GRAM to be used with various models, such as RNNs [4, 1] and transformer encoders [7], under different con- figurations. This updated implementation aims to provide a more accessible and adaptable framework for leveraging the benefits of GRAM in predictive modeling for sequential diagnoses. The implemented model will be tested with the synthetic EHR dataset Synthea [8] together with the modern SNOMED CT [3] medical ontology.},
    apice = {FreddiGardenal2324},
    author = {Davide Freddi and Davide Gardenal},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Study, design, and implement the GRAM symbolic knowledge injection algorithm},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/freddigardenal2324}
}
@courseproject{dooja2324-ethicsofai20232024,
    abstract = {The goal of this project is to classify four distinct hand movements using machine learning techniques applied to data generated from high-density surface EMG (HD- sEMG) signals. We will evaluate three types of neural networks: Deep Neural Networks (DNN), Long Short-Term Memory Networks (LSTM), and Spiking Neural Networks (SNN). Each method will test for its ability to accurately identify movements associated with four types of isometric contractions flexion, extension, pronation, and supination performed at 10% of MVC. This study will highlights the comparative analysis in the context of movement recognition.},
    apice = {Dooja2324},
    author = {Kashaf Ad Dooja},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Classification of Hand Movements Using DNN, LSTM, and SNN: A Comparative Study},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/dooja2324}
}
@courseproject{eriksen2324-ethicsofai20232024,
    abstract = {The aim of this project is to develop a framework designed to assist users in constraining machine learning (ML) models. By utilizing this framework, users will have the ability to specify certain fairness metrics that should be upheld, thereby enhancing fairness in ML processes. For instance, one such metric could be ensuring demographic parity stays below a certain threshold. Given the opacity of neural networks (NN), this framework will especially focus this particular use case. Additionally, this will involve implementing a mechanism to penalize batches that fail to uphold the symbolic constraints defined in the model.},
    apice = {Eriksen2324},
    author = {Silje Eriksen},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta and Ciatto, Giovanni and Magnini, Matteo},
    title = {Framework for Symbolic Rules Constraints in Neural Networks},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/eriksen2324}
}
@courseproject{torzi2324-ethicsofai20232024,
    abstract = {In the past years, self-supervised learning emerged as a new paradigm to learn more expres- sive embeddings of the input itself. It is especially used in Large Language Models, like BERT, to learn contextual representations of words. Given the intrinsic structure of the Transformer Encoder learning block capable of accepting every type of input, in the successive years, this Architecture has been employed in the processing of other modalities, such as images; and, as a step forward, CLIP was presented with the aim of creating a Multi-Modal Model connecting image and text representations in a shared embedding dimensionality, learned from a huge amount of (image, text) pairs collected from the internet. This training methodology, that allows the model to learn from any type of data in a self supervised way, could highlight and enhance bias that exists in society and is inherited in the training data, generating discrimination if used in critical applications. The goal of this project is to understand if the use of a Multi-Modal Model, such as CLIP, that connects text and visual embeddings in a shared space, could increase the bias compared to a BERT model, that is trained on text data only.},
    apice = {Torzi2324},
    author = {Torzi Luca},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Assessing Bias in Multi-Modal Models: A Study of CLIP on Gender and Skin Tone},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/torzi2324}
}
@courseproject{merolajaspinderothmane2425-ethicsofai20242025,
    abstract = {Large language models (LLMs) have demonstrated impressive emergent abilities across diverse natural language tasks, yet their capacity for structured planning remains under-explored. This project investigates whether the LLaMA model can reliably solve planning problems and introduces a meta-network to evaluate and manage its performance. First, we construct a benchmark suite of planning prompts and record LLaMA’s responses. Next, we train a lightweight meta-network atop these examples to predict, for each prompt, (a) the likelihood that LLaMA’s answer is correct, and (b) a confidence score. When the meta-network predicts low confidence or high error risk, the system automatically delegates the task to alternative solvers or human experts. Our framework provides a generalizable mechanism for risk-aware deployment of LLMs balancing autonomy with reliable fallback strategies.},
    apice = {MerolaJaspinderOthmane2425},
    author = {Carlo Merola and Jaspinder Singh and Othmane Dardouri},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Robustness of LLMs},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/merolasinghdardouri2425}
}
@courseproject{castellifratinimone2324-ethicsofai20232024,
    abstract = {Bias in artificial intelligence algorithms used in healthcare is a growing concern due to its potential to negatively impact patient outcomes. AI tools, despite their power, can perpetuate systematic unfairness towards certain groups. This is particularly critical in healthcare, where existing biases can exacerbate inequalities. Historical examples, such as the Fram- ingham Risk Score, demonstrate how biases in traditional medical tools have led to variable accuracy across different ethnic groups and genders, due to the underrepresentation of diverse cohorts in their development. Women, for instance, experience higher rates of missed and delayed car- diovascular disease diagnoses, partly because of their underrepresentation in clinical trials. While some level of bias in data and tools is inevitable, the validation of AI tools to minimize bias offers a promising opportunity to reduce disparities in healthcare outcomes.},
    apice = {CastelliFratiniMone2324},
    author = {Giorgia Castelli and Alice Fratini and Madalina Ionela Mone},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Bias Mitigation in AI Models for Cardiovascular Diseases Prediction},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/castellifratinimone2324}
}
@courseproject{raimondigurioli2324-,
    abstract = {This literature review delves into the concept of explainability in the context of large language models (LLMs). It investigates the trade-offs between performance and explainability that are inherent in these models, comparing different approaches and methodologies for enhancing transparency. The review’s goal is to summarize current research, pinpoint gaps, and suggest directions for future work to enhance the explainability of LLMs while preserving their performance.},
    apice = {RaimondiGurioli2324},
    author = {Bianca Raimondi and Gurioli, Andrea},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Explainability of Large Language Models (LLMs): A Literature Review},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/raimondigurioli2324}
}
@courseproject{lianghuang2425-,
    apice = {LiangHuang2425},
    author = {Xiaojun Liang and Shengzi Huang},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Gender Bias in AI-Generated Texts: A Comparative Perspective between English and Chinese Language Models},
    type = {Standard Project}
}
@courseproject{fusaquarta2324-ethicsofai20232024,
    abstract = {Providing extensive support for categorical data in a ready to use ap- proach, enhanced computational performance, overfitting reduction and fast predictions, CatBoost is the state-of-the-art model for the tabular datasets. However, in research about fairness LightGBM gained the spot- light as gradient boost decision trees based algorithm with the well-known implementation called FairGBM that ingrains a in-process schema to guar- antee fairness. The proposed project aims to explore the effects of fairness approaches on CatBoost considering several datasets to assess gains and losses in terms of resulting fairness and overall performances. During the last quinquennium the research community focused its efforts on a set of datasets called ACS (used to empirically validate FairGBM), the proposed project adds other data sources as well in order to better comprehend the potentiality of CatBoost when fairness is involved.},
    apice = {FusaQuarta2324},
    author = {Edoardo Fusa and Angelo Quarta},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {On the Fairness capability of CatBoost},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/fusaquarta2324}
}
@courseproject{angilerimarzioshodi2324-ethicsofai20232024,
    abstract = {This project addresses the automation of the student loan eligibility process using a dataset focused on loan approval. Given the risk of inherent bias in machine learning models, particularly towards certain demographic groups, ensuring fairness is paramount. We will experiment with various machine learning models to achieve a balance between accuracy and fairness. By mitigating biases, such as those related to gender, our aim is to promote equitable decision-making and minimize potential discriminatory outcomes.},
    apice = {AngileriMarziOshodi2324},
    author = {Chiara Angileri and Niccol`o Marzi and Shola Oshodi},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Bias mitigation in automated loan eligibility process},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/angilerimarzioshodi2324}
}
@courseproject{saboonisamadzadeh2425-ethicsofai20242025,
    abstract = {This project investigates the integration of individual fairness metrics into Automated Ma- chine Learning (AutoML) pipelines, with the goal of aligning model selection and evaluation processes with ethical and socially responsible AI practices. Leveraging IBM’s inFairness li- brary, we design a training pipeline that evaluates models not only based on predictive accuracy but also on individual fairness criteria—specifically, the spouse consistency metric. By embed- ding fairness evaluation within the AutoML loop, the project aims to contribute toward the development of more accountable and equitable AI system},
    apice = {SabooniSamadzadeh2425},
    author = {Tara Sabooni and Yasaman Samadzadeh},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness in AutoML},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/saboonisamadzadeh2425}
}
@courseproject{masoudian2425-ethicsofai20242025,
    apice = {Masoudian2425},
    author = {Tannaz Masoudian},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness-Aware Deep Learning for Early Detection of Autism Spectrum Disorder},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/masoudian2425}
}
@courseproject{hassanibanihasheminazarmohsenifakori2425-ethicsofai20242025,
    abstract = {Our research proposal aims to explore the potential of mitigating social biases in Large Language Models (LLMs) by combining multi-agent frameworks and reflective reasoning tools. Specifically, we intend to leverage CAMEL’s multi-agent framework to enable multiple agents with distinct roles to collaboratively critique and refine outputs, along with Anthropic’s ”think” tool which invokes intermediate reasoning steps. By measuring bias reduction compared to traditional baselines, this project aims to show that agent collaboration and reflective self-correction can significantly boost the fairness and reliability of AI-driven systems},
    apice = {HassaniBanihashemiNazarmohsenifakori2425},
    author = {Hesam Sheikh Hassani and Mehregan Nazarmohsenifakori and Safoura Banihashemi},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Mitigating Social Bias in LLMs via Multi-Agent Collaboration and Reflective Reasoning},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/hassanibanihasheminazarmohsenifakori2425}
}
@courseproject{buzzoniromeo2425-ethicsofai20242025,
    abstract = {In the current landscape, characterized by the emergence of large- scale datasets, privacy preservation has become a central issue. The need for data to train recommendation algorithms has led to the collection of often sensitive and private information. A possible solution to ensure user privacy is the use of synthetic data, artificially generated, as a substitute for real data. The main challenge of this approach lies in balancing the quality and representativeness of synthetic data with the ability to preserve privacy, preventing the possibility of tracing back to private information. High quality synthetic data does not always translate into good generalization and privacy protection, making it necessary to identify an effective trade- off. This project aims to evaluate different synthetic data generation tech- niques by testing their performance in terms of privacy preservation and data quality across multiple real-world datasets. Models based on var- ious approaches will be analyzed in order to identify the most effective methodologies.},
    apice = {BuzzoniRomeo2425},
    author = {Mattia Buzzoni and Riccardo Romeo},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Evaluation of Privacy Preservation Capabilities through Synthetic Data on Multiple Datasets},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/buzzoniromeo2425}
}
@courseproject{albertazzigasperini2324-ethicsofai20232024,
    abstract = {Fairness in machine learning (ML) [8][5] is crucial, especially as AI systems increasingly influence various sectors, including healthcare and legal decisions. The application of AI must be cautious, as issues of fairness and algorithmic bias can have severe consequences. This work aims to try different methods to assess the fairness of machine learning models, regarding the risk of rehospital- ization in the case of diabetic patients. We use a kaggle dataset [3] representing 10 years (1999-2008) of clinical care at 130 US hospitals and integrated de- livery networks. It includes over 50 features representing patient and hospital outcomes. The data contains such attributes as patient number, race, gender, age, admission type, time in hospital, medical specialty of admitting physician, number of lab test performed, HbA1c test result [2], diagnosis, number of med- ication, diabetic medications, number of outpatient, inpatient, and emergency visits in the year before the hospitalization, etc..},
    apice = {AlbertazziGasperini2324},
    author = {Arianna Albertazzi and Lucia Gasperini},
    howpublished = {Ethics of AI, A.Y. 2023/2024},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Fairness in diabetes prediction dataset},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/albertazzigasperini2324}
}
@courseproject{dibuoreale2425-ethicsofai20242025,
    apice = {DibuoReale2425},
    author = {Christian Di Buò and Simone Reale},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Sustainable City via Trustworthy Digital Twin: a Use Case},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/dibuoreale2425}
}
@courseproject{cridligsangiorgifossa2425-ethicsofai20242025,
    abstract = {Aligning Language Models (LMs) with human moral values is a critical challenge in the development of responsible AI systems. Reward Models play a central role in this process by enabling post-training alignment through Reinforcement Learning from Human Feedback (RLHF) [2]. In this project, we leverage the ETHICS dataset [1] to investigate the extent to which modern Reward Models can represent and generalize ethical concepts. Using open-source tools, we plan to create morally congruent Reward Models aligned with shared human values, intended for integration into RLHF pipelines. Our goal is to develop a prototype framework capable of evaluating and guiding LMs toward ethically preferable outputs.},
    apice = {CridligSangiorgiFossa2425},
    author = {Nicolas Cridlig and Marco Sangiorgi and Andrea Fossa},
    howpublished = {Ethics of AI, A.Y. 2024/2025},
    sort = {classic},
    supervisor = {Calegari, Roberta},
    title = {Ethically Aligned Reward Models: A Language Model Post-Training Framework},
    type = {Standard Project},
    url = {https://dvcs.apice.unibo.it/pika-lab/courses/ai-ethics/projects/cridligsangiorgifossa2425}
}