Certus: A domain specific language for confidence assessment in assurance cases
This paper introduces Certus, a domain-specific language that enables quantitative confidence assessment in assurance cases by utilizing fuzzy sets for linguistically meaningful judgment representation and transparent expression-based propagation, addressing common limitations of existing methods.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a judge in a high-stakes courtroom, but instead of deciding if someone committed a crime, you are deciding if a self-driving car is safe enough to hit the road. In the world of engineering, this isn't just a gut feeling; it's a formal process called an "Assurance Case." Think of an Assurance Case as a giant, logical puzzle where engineers stack up pieces of evidence—like test results, code reviews, and safety checks—to build a tower of proof that says, "We are confident this system won't hurt anyone."
For a long time, engineers have struggled with how to measure that "confidence." Some try to be super precise, assigning exact numbers like "93.4% confidence." But this is tricky. Is 93.4% good enough? Is 93.3% a disaster? It's like trying to describe the taste of a strawberry using only a ruler; the numbers feel cold, confusing, and easy to misinterpret. Others use vague words like "pretty sure" or "very worried," but then it's hard to do math with those words to see if the whole tower of proof holds up. The big question is: How do we mix the precision of math with the natural, fuzzy way humans actually think about risk?
This is where a new tool called Certus steps in. The paper introduces Certus as a special "language" designed to help engineers talk about confidence in a way that feels natural but still works like math. Instead of forcing engineers to pick a specific number like 0.87, Certus lets them use fuzzy words like "high," "low," or "very low." But here's the magic trick: Certus treats these words like "fuzzy sets." Imagine a sliding scale where "high" isn't just one point, but a soft cloud of possibilities that overlaps with "medium." This allows the system to do calculations with these clouds, carrying the "fuzziness" all the way up the argument tower to the final verdict.
The authors, Simon Diemert and Jens H. Weber, suggest that this approach solves several headaches. First, it stops people from misinterpreting numbers (no more panic over a 0.07% failure chance). Second, it admits that human judgment is naturally vague and lets engineers express that nuance without losing it in a calculator. Third, it handles "defeaters"—which are like little "gotcha!" arguments that try to knock the tower down. Certus has a special way of letting these "gotchas" subtract from the confidence without breaking the whole system.
To test their idea, the team built a prototype of Certus and applied it to a real-world example: a safety argument for an Adaptive Cruise Control (ACC) system in a car. They showed how the language could take simple statements like "the code inspection is very high confidence" and combine them with "the hardware might fail" to produce a final, calculated confidence level. When they compared their fuzzy results to an older method that used strict numbers, the results were similar, but Certus felt more intuitive and transparent.
The paper doesn't claim that Certus is the finished, perfect product ready for every factory tomorrow. Instead, it suggests that this is a promising new direction. The authors admit they need to do more work to turn this prototype into a tool that engineers can use every day, including creating better visual diagrams and proving that the results are trustworthy enough for life-or-death decisions. But for now, Certus offers a playful yet powerful new way to build trust in the machines that will soon be driving us.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.