When working with assurance case arguments, a simple text argument format can be useful. This should be understandable to humans and usable with AI tools, while also being formally defined so this can be imported into assurance case editors. For this purpose, we are introducing LTAC – Lightweight Text Assurance Case. It’s a simplified text format for presenting assurance case arguments. Let’s look at an example argument. Is it easy to read and understand?
- Claim G1: System is acceptably safe to operate
- Context C1: System Definition
- Strategy S1: Argue by mitigating all hazards
- Claim G2: All hazards have been identified
- Evidence E1: Hazard analysis report
- Claim G3: All identified hazards have been mitigated
- Strategy S3: Decompose mitigation argument by identified hazards
- Claim S5: All identified hazards have been accurately assessed
- Claim S6: Mitigation strategies have been implemented for all hazards
- Claim S7: The effectiveness of the mitigation measures has been validated
- Claim S4: Continuous monitoring of risks is in place The above argument fragment is human-readable, which is crucial. It can easily be reviewed and modified as needed.
When this argument is pasted into PREMIS, we get the GSN diagram as shown below.
LTAC arguments can be created manually, but also using AI tools or scripts. If you want to use an AI tool, first give it an instruction to use LTAC notation according to the LTAC specification. Once the AI tool learns the LTAC notation, it can easily create and analyse arguments in this way.
As the argumentation gets larger, this textual form becomes more difficult for humans to review. However, the argument fragments in LTAC are still easy to use.
There’s another major limitation to LTAC. This text-based argumentation is effective when there are no argument element descriptions or metadata, such as ‘undeveloped’ tags or other similar information. In such cases, more advanced textual notations, including those using YAML, can be used. However, when the data for a single argument element is more complex and not presented in a single line, it becomes more difficult for manual analysis.
LTAC is an effective solution when you use AI tools and want to review the argument before inserting it into the assurance case tool. It works well with PREMIS.