Seminar M.Sc Students
Predicting User Decisions in DSS-Assisted Signal Detection
Omri Bouchnick, M.Sc. student in the School of Industrial & Intelligent Systems Engineering
Advisor: Prof. Joachim Meyer &Mr. Yoav Ben Yaakov
Abstract:
Can single decisions be predicted in a Decision Support System (DSS) context? This thesis addresses this question through a controlled signal-detection experiment. A large-scale online study was conducted (N=569 participants, 3x11x11 factorial design) where users classified noisy stimuli as "Signal" or "Noise" while receiving DSS recommendations of varying reliability. Machine learning models were then trained to predict individual decisions using information available at decision time, including current-trial features and behavioral history.
The models achieved strong predictive performance: an XGBoost classifier reached 86.3% accuracy (AUC=0.937) for within-user prediction and 83.3% accuracy (AUC=0.910) when generalizing to entirely new users - substantially outperforming baseline heuristics. Feature importance analysis (SHAP) revealed that the DSS recommendation was the dominant predictor, followed by stimulus evidence and the known sensitivities of the DSS and human. Notably, removing behavioral history features reduced accuracy by only 2%, indicating that decisions are driven primarily by immediate context rather than past behavior.
Behavioral analyses provided additional insight: users followed the DSS on 76.6% of trials, and override decisions were significantly slower than follows, suggesting that disagreement requires cognitive effort rather than reflecting snap judgments. Despite reasonable performance (75.6% accuracy), participants left substantial value on the table: 60% would have done better by simply always following the DSS, and all fell far short of the optimal 86.2% achievable by integrating both DSS and stimulus evidence. These findings demonstrate that human-DSS decisions are highly predictable, supporting the design of adaptive interfaces that intervene at moments of likely conflict while highlighting persistent inefficiencies in human reliance strategies.
Bio:
Omri Bouchnick is an M.Sc. candidate in Industrial Engineering at Tel Aviv University and a Business-Data Science Analyst. His research focuses on human-automation interaction, trust calibration, and decision-making under uncertainty.
Contact:
E-Mail: bouchnick@mail.tau.ac.il
LinkedIn: https://www.linkedin.com/in/omri-bo/
Modeling Algorithmic Decision Support for Femur-Fracture Prevention Surgery in Prostate Cancer Patients
Snir David Malka, M.Sc. student in the School of Industrial & Intelligent Systems Engineering
Advisor: Prof. Joachim Meyer &Mr. Yoav Ben Yaakov
Abstract:
Bone metastases from prostate cancer frequently involve the femur and can lead to a pathologic fracture, an event associated with loss of function, urgent surgery, and likely morbidity. Clinicians therefore face a difficult decision: whether to recommend prophylactic femoral fixation to prevent fracture, or to continue surveillance. This decision is complicated by uncertainty in individual fracture risk, competing mortality risk, surgical complications, and the time-varying nature of both cancer progression and functional status. This research develops a decision-support model designed to evaluate the benefit of an algorithmic system that predicts impending fractures. We compare clinical outcomes under “wait” versus “fix” strategies, with and without algorithmic fracture risk predictions, and quantify how the system's value changes as a function of patient characteristics and system properties. The model explicitly represents key pathways (fracture, postoperative recovery trajectories, and death) and evaluates outcomes in terms of QALYs (Quality Adjusted Life Year) and economic impact. The goal is to clarify when and for whom a predictive algorithm meaningfully improves decisions and outcomes.
Bio:
Snir David Malka is an M.Sc. student in the School of Industrial and Intelligent Systems Engineering at Tel Aviv University, supervised by Prof. Joachim Meyer. His research focuses on decision-support modeling in healthcare, with an emphasis on dynamic, patient-specific clinical decisions under uncertainty. In his thesis, Snir develops a Markov decision process framework to guide prophylactic fixation decisions for prostate cancer patients with femoral metastases.
E-Mail: snirmalka123@gmail.com
LinkedIn: https://il.linkedin.com/in/snir-malka/
Decision Support for Adaptive Human Capital Management: A Machine Learning & Mathematical Programming Framework for Skill-Based Recruitment
Ayelet Dabush, M.Sc. student in the School of Industrial & Intelligent Systems Engineering
Advisors:
Prof. Yossi Bukchin, School of Industrial & Intelligent Systems Engineering, Tel-Aviv University, Israel,
Dr. Hila Chalutz-Ben Gal, The Alexander Kofkin Faculty of Engineering, Bar-Ilan University, Israel
Yariv Tabac, The Extreme Group - HR Tech Company
Abstract:
The nature of work is transforming driven by rapid technological change and the spread of artificial intelligence (AI) and automation. AI is reshaping the labor market and organizational processes by augmenting human skills, automating routine tasks, and enabling new modes of collaboration between humans and intelligent systems. These rapid changes enable greater workforce flexibility, demands new skill sets-both technical and soft skills-and accelerates the transition from traditional one-to-one (candidate-per-job) employment models, to more dynamic, many-to-many skill-based projects. The integration of AI in human resource functions further emphasizes the need for scaling the human capital assessment processes with robust decision support tools to identify required technical and soft skills.
This paper addresses this pivotal shift in employee recruitment toward data-driven, skill-based allocation frameworks that factor both technical and soft skills. The study introduces a novelle analytical framework for skill-based recruitment that combines machine learning and mathematical programming tools. We train multiple classification models on a large dataset of recruiter-labelled historical data, extracting feature embeddings from both candidate résumés and job descriptions. The resulting probability scores are then integrated in mixed-integer optimization models to solve large-scale assignment problems under real-world operational constraints. The experimental design includes over 150 simulation iterations per scenario to evaluate allocation performance and sensitivity across defined hypotheses. Beyond conventional one-to-one integer assignment (candidate-to-job), we operationalize a novel fractional assignment model (candidate-to-project) allowing candidates to partially fulfil multiple project needs, thus reflecting emerging managerial need for flexibility, skill-based and team-level configuration.
Research findings reveal that transitioning from recruiter-based to skill-based allocation is feasible, as the proposed skill-based model achieves comparable performance across multiple parameters. By comparing the integer and fractional allocation results, the study also demonstrates the boundaries, trade-offs, and practical value of prescriptive, fractional skill-based recruitment strategy for improved workforce management, particularly when technical skills or candidate pool size are high. These results advance the literature by providing a validated framework for optimizing human capital and talent allocation in the rapidly shifting labor market
Bio:
Ayelet Gvili (Dabush) is a Master's student in Industrial Engineering, specializing in Data Science. She holds a Bachelor of Science degree in Industrial Engineering from Tel Aviv University. Ayelet currently works as a Data Scientist at Menora Mivtachim, a leading insurance and financial services company, where she applies data analysis and machine learning techniques to support data-driven decision-making and process optimization.
