Prof. Marek Rusinkiewicz

Title: From Search Trees to Blockchains: A History of Landmark Ideas in Computing

Abstract: In my talk, I will discuss the emergence of several foundational technologies that have helped shape today’s information landscape, using as examples projects in which I was involved. I will first describe the early work in artificial intelligence leading to the world’s first computer chess match. I will then talk about the InfoSleuth project at MCC, which was one of the early attempts to create a large self-organizing community of software agents collaborating to solve complex information processing tasks -- a precursor to many of today's distributed intelligent systems. Finally, I will describe how practical challenges in document versioning and information retrieval led to the development of two technologies that have since become fundamental to modern computing: cryptographically linked timestamping (blockchains) and Latent Semantic Indexing

Biography: Marek Rusinkiewicz is a computer scientist, an educator, and a former research executive. Currently he is the Dean Emeritus of the College of Computing at the New Jersey Institute of Technology (NJIT). He joined NJIT after retiring from his position as a Senior Vice President and the General Manager of Applied Research Laboratories at Telcordia Technologies (formerly Bell Communication Research), which included R&D centers in New Jersey, Texas, Taiwan and Poland. Under his leadership, Telcordia Labs carried out research in telecommunication networks, mathematical sciences, software engineering, information processing, and information security.

Prof. Stefan Tai

Title: TBA

Abstract: TBA

Biography: Stefan Tai is a Full Professor and Head of the Chair of Information Systems Engineering at TU Berlin, a role he has held since 2014. His academic career includes prior appointments as Full Professor at the Karlsruhe Institute of Technology (KIT) and as a Research Staff Member at the IBM Thomas J. Watson Research Center in New York.

Prof. Lech Madeyski

Title: The Model Is a Service: LLM4SCREENLIT for Credible Secondary Studies in Service-Oriented Computing

Abstract: Systematic reviews consolidate the evidence a research community relies on, and screening titles and abstracts is the most labour-intensive step in producing one. Large language models (LLMs) promise to compress that work from months into hours, and Service-Oriented Computing — already applying them to service composition, microservice decomposition and business process discovery — will reach for them here too. This talk argues that the model is no longer an artefact a researcher holds but a service they call, and that this turns credibility into a services problem.
Two failures compound. The first is measurement. On a 9,695-article secondary study, the LLM that maximises Accuracy loses 63.3% of the relevant evidence. The Matthews Correlation Coefficient (MCC), the standard remedy for class imbalance, still loses 43.9%. A cost-sensitive criterion, the Weighted Matthews Correlation Coefficient (WMCC) proposed in LLM4SCREENLIT, selects a model that loses only 5.8%.
The second failure is delivery. On a benchmark built from a Service-Oriented Computing review of 1,130 records, recording the model name turned out not to record what ran. One unpinned run was served by thirteen providers at precisions from fp4 to bf16. Nine endpoints were pinned, one per model, and each was sent the same request twice. Four changed their own decision between the two calls. A local run reproduced exactly across all 1,130 records. No hosted endpoint has been shown to do the same.
The talk presents LLM4SCREENLIT — the WMCC, the Lost Evidence metric, decision trees for benchmarking and deployment, and the provenance record a screening result must carry — and asks what a service contract must guarantee before a model may be used as a scientific instrument.

Biography: Lech Madeyski is an Associate Professor and Deputy Head of the Department of Applied Informatics at Wrocław University of Science and Technology. He has held visiting appointments at Keele University, Lancaster University and Blekinge Institute of Technology. His research applies AI/ML and data science to software engineering — most recently the credible evaluation of LLMs for screening in systematic literature reviews — alongside software defect prediction, vulnerability prediction, code-smell detection, code and test quality, mutation testing, and robust statistical methods for empirical research. With Barbara Kitchenham and Martin Shepperd he co-authored LLM4SCREENLIT, guidelines for the credible use of LLMs in study screening that underpin this talk. With Kitchenham and David Budgen he established SEGRESS, the unified reporting standard for secondary studies in software engineering — an extension of PRISMA 2020, the medical standard. His collaboration with Kitchenham spans over a dozen joint papers. He is the founder and Editor-in-Chief of the e-Informatica Software Engineering Journal, appears in the Stanford/Elsevier "World's Top 2% Scientists" ranking, and has served on the program committees of ICSE 2026 and 2027, FSE 2025, and ASE 2023 (all CORE A* conferences).