Myths and Misconceptions about Machine Learning and How They Are Related to Software Engineering
Stefan Kramer, Johannes Gutenberg - Universität Mainz, Germany
What Can We Learn from Play?
Panos Markopoulos, Eindhoven University of Technology, Netherlands
Impact of End User Human Aspects on Software Engineering
John Grundy, Monash University, Australia
Myths and Misconceptions about Machine Learning and How They Are Related to Software Engineering
Stefan Kramer
Johannes Gutenberg - Universität Mainz
Germany
Brief Bio
Stefan Kramer is professor of data mining at the Institute of Computer Science at Johannes Gutenberg University (JGU) Mainz and Honorary Professor at the University of Waikato in Hamilton, New Zealand. He has been active in data mining since the first conference worldwide in 1995, had more than 30 funded projects, authored more than 200 publications, and authored award-winning papers at conferences such as IEEE ICDM, ACM SIGKDD, ILP, and IEEE ICBK. He was Vice Chair of IEEE ICDM 2013 and is 2021 Program Chair of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2021). His research interests include mining structured data (sequences, strings, text, time series, trees, graphs, data in logical formalisms, etc.), stream mining, the use of prior knowledge in machine learning, and machine learning under real-world constraints such as privacy, confidentiality, fairness, and explainability.
Abstract
In the talk I will discuss machine learning (ML) myths and misconceptions, with a perspective of software and systems engineering, including requirements as well as verification and validation (V&V). The often-cited ideas about ML include: ML is always data-hungry, always opaque, can only deal with association and not causation, cannot at all deal with uncertainty, that ML is per se not within the scope of computer science and software engineering, or, vice versa, can be dealt with completely by standard tools from these areas. The talk will clarify the concepts and relationship among concepts, will point to literature that may not generally be known to audiences outside ML, and thus hopefully contribute to a better understanding of the current state of the art, the chances and risks of the methods and technologies, and potential future developments.
What Can We Learn from Play?
Panos Markopoulos
Eindhoven University of Technology
Netherlands
Brief Bio
Prof. Panos Markopoulos is a computer scientist specializing in the field of Human-Computer Interaction. He is a professor in Design for Behaviour Change at the Department of Industrial Design at the Eindhoven University of Technology. He has worked on several topics including task analysis, awareness systems, ambient intelligence, and interaction design for children. His current research concerns designing interactive technologies for rehabilitation and for playful learning. He is a founding editor of the journal Child-Compute Interaction and is currently serving as chief editor of the Behaviour & Information technology journal.
Abstract
This talk shall review the design of a number of games designed with the purpose of supporting motor learning, social skills, and encouraging physical activity and social interaction for various user groups emphasizing on the role of embodiment in interaction. It will also demonstrate how games can be valuable media for learning about people and I discuss the potential and limits of player modelling. The talk shall conclude with some general lessons and challenges for future work in this area.
Impact of End User Human Aspects on Software Engineering
Brief Bio
John Grundy is Australian Laureate Fellow and Professor of Software Engineering at Monash University, Melbourne, Australia. His five year Laureate programme on Human-centric Model-driven Software Engineering aims to address some of the deficiencies in current software development practices that fail to take account of diverse software developer human characteristics and diverse software end user human characteristics. He leads the Human-centric Software Engineering (HumaniSE) lab and has published over 500 refereed papers in software tools, visual modelling languages, model-driven software engineering, software architecture, requirements engineering, and software security. He is a Fellow of Automated Software Engineering and Fellow of Engineers Australia.
Abstract
Software is designed and built to help solve human problems. However, much current software fails to take into account the diverse end users of software systems and their differing characteristics and needs eg. age, gender, culture, language, educational level, socio-economic status, physical and mental challenges, etc. I give examples of some of these diverse end user characteristics and the need to better incorporate them into requirements engineering, design, implementation, testing, and defect reporting activities in software engineering. I report on some of our work trying to address some of these issues, including: use of personas to better characterise diverse end user characteristics; extending requirements and design models to capture diverse end user needs; analysis of app reviews and JIRA logs to identify problems and ways developers try to address them; analysis of approaches to improve the accessibility of software designs for diverse end users; improved human-centric defect reporting approaches; and use of living lab co-design approaches to ensure end users are first class contributors during all phases of software development. I finish by outlining a research roadmap aiming to improve the incorporation of end user human aspects into software engineering.