Program
Registration: https://conf.researchr.org/attending/RE-2024/registration
The workshop is planned to occur on June 25th, 2024 in Reykjavik, Iceland. The times in the following program are in Greenwich Mean Time (GMT, i.e., UTC+0).
Schedule
9:00 - 9:15 Welcome from the Organizers
Session 1: Keynote and Papers
Chair: Chetan Arora
9:15 - 10:15
Keynote: The proof is in the pudding - Real-world use cases for GenAI for Requirements Engineers
Jan-Philipp Steghöfer
10:15 - 10:45
Break
Session 2: Papers
Chair: Julian Frattini
10:45 - 11:10
Quest-RE - QUestion Generation and Exploration STrategy for Requirements Engineering
Hussein Hasso, Bettina Fischer-Starcke and Hanna Geppert
11:10 - 11:35
Classifying Ambiguous Requirements: An Explainable Approach in Railway Industry
Lodiana Beqiri, Calkin Suero Montero, Antonio Cicchetti and Andrey Kruglyak
11:35 - 11:50
Identifying Maintenance Needs through Machine Learning: a Case Study in Railways
Raihana Ferdous, Giorgio Spagnolo, Alessandro Borselli, Lucio Rota and Alessio Ferrari
11:50 - 12:15
Interpretable App Review Classification with Transformers
Momojit Biswas, Preethu Rose Anish and Smita Ghaisas
12:15 - 13:45
Lunch Break
Session 3: Papers
Chair: Fatma Başak Aydemir
13:45 - 14:10
Exploring the capabilities of large language models for the generation of safety cases: the case of GPT-4
Mithila Sivakumar, Alvine Boaye Belle, Jinjun Shan and Kimya Khakzad Shahandashti
14:10 - 14:35
Using GPT-4 Turbo To Automatically Identify Defeaters In Assurance Cases
Kimya Khakzad Shahandashti, Alvine Boaye Belle, Mohammad Mahdi Mohajer, Oluwafemi Odu, Timothy C. Lethbridge, Hadi Hemmati and Song Wang
14:35 - 15:00
Design of the Safety Case of the Reinforcement Learning-enabled Component of a Quanser Autonomous Vehicle
Mithila Sivakumar, Alvine Boaye Belle, Jinjun Shan, Oluwafemi Odu and Mingfeng Yuan
15:00 - 15:15
Formalization of Natural Language Requirements using Large Language Models
Johannes Norheim and Eric Rebentisch
15:15 - 15:45
Break
Session 4: Papers and Activity
Chair: Chetan Arora
15:45 - 16:10
Requirements-driven Slicing of Simulink Models Using LLMs
Dipeeka Luitel, Shiva Nejati and Mehrdad Sabetzadeh
16:10 - 16:25
80% Complete AI-Generated Functional Tests: Austrian Post Lightning Talk
Tomas Herda, Sandra Dertnig and Verena Homm
16:25 - 17:15
Joint activity: "Have large-language models (LLMs) made our previous AI tools obsolete?"
Interactive session exploring the possibilities and limitations of LLMs emulating previously hand-crafted tools from the AIRE workshop.
Material: https://github.com/aire-ws/aire24-activity
17:15 - 17:30 Closing
Keynote
Speaker: Jan-Philipp Steghöfer
Title: The proof is in the pudding - Real-world use cases for GenAI for Requirements Engineers
Abstract: The current frenzy about generative AI leaves me wondering about proper RE practices. It seems like we are faced with a novel technology and are now frantically trying to find use cases for it that range from the mundane (meeting summarization) to the absurd (cannabis sommelier). Software Engineering provides a more sober backdrop, but I am still left with the impression that LLMs are still a solution in search of a problem. We do not understand their possibilities and limitations well enough in the context of actual, real world practices, particularly in areas of SE that are not directly coding-related such as requirements engineering. I have worked with practitioners and academics in recent months to identify use cases and the research gaps we need to plug in order to address them with GenAI. In this talk, I am going to show you some of these use cases, why LLMs would be a good solution for them, and which steps we have already taken to address them. Some of these ideas have already ended up in research project proposals, for others there is already tooling available. Whether any of this actually works remains to be seen - but I will give you a first taste of the pudding that will contain the proof of LLM's usefulness for RE.
About the speaker: Jan-Philipp Steghöfer is a Senior Researcher at XITASO where he works on research projects on AI in healthcare, cybersecurity, and software engineering. He also supports XITASO's development teams as a requirements engineer, software architect, and security specialist. Previously, he worked as an associate professor in software engineering at the University of Gothenburg and Chalmers University of Technology where he did research on model-driven engineering, agile methods, and software and systems traceability. He is a reviewer for ICSE, TSE, JSS, RE, REFSQ and others and active in organising events in the software and requirements engineering community.
Joint Activity
Aim: The AIRE workshop series has - similar to comparable venues like the Natural Language Processing for Requirements Engineering (NLP4RE) workshops - brought forth many interesting AI- and ML-powered tools. These tools were often meticulously designed, authors collected and annotated training data manually, and tools were trained locally with much effort. But the wake of LLMs and generative AIs (GenAIs) like ChatGPT or Llama has put all this effort into question. These tools can perform several requirements engineering related tasks out of the box, without any design, and barely any data preparation or training. Was all previously invested effort obsolete? In this activity, we invite to explore this hypothesis. The activity is neither systematic nor conclusive, but shall rather stimulate discussions about the topic.
Preparation: You can prepare for the activity by making sure you have access to an LLM of your choice (e.g., ChatGPT). Optionally, prepare a GitHub account and set up a local git environment on your machine. You should be able to fork a repository, clone it to your machine, and push changes back to your fork. Check out the material linked below for further details.
Material: https://github.com/aire-ws/aire24-activity