Software testing from the system perspective has been the cornerstone of testing for decades, ensuring software satisfies customer requirements, is efficient, and dependable. And our community has spent significant effort formalizing and leveraging models that support this goal. Strong models lead to better testing and have provided the foundations for automated program repair and self-adaptation. As we have traveled through time we have utilized different types of models for specifying program inputs, system configurability, and for coverage of sequences of actions; all driven by the assumption that we can build those models. However, as we move into the next generation of automated software engineering, some of our long-held assumptions are breaking. Specifications may no longer be available or are less complete, our systems are becoming more complex and harder to define, and we have an increasing number of competing objectives. This is making our software more difficult to formally specify, and hence to test. Yet, we also have new powerful tools at hand that can help mitigate these challenges. In this talk I will look back and forward across space and time with an eye towards a new future of automated system testing.

@inproceedings{paperbot12284,
  title = {Lost in Automated Specifications: A Time and Space Odyssey in System Testing (Keynote)},
  author = {Myra B. Cohen},
  booktitle = {ASE},
  year = {2026},
  doi = {10.1145/3832783.3857285}
}