normativeai / frontend

GNU General Public License v3.0
6 stars 3 forks source link

Support defeasibility in the editor #71

Open shaolintl opened 4 years ago

shaolintl commented 4 years ago

We would like to support engines capable of defeasible reasoning. for that, we need to be able to mark which formula defeats which.

Jason has suggested we use two levels approach. The current annotator generates formulas and then, one can mark on the formulas which one defeats the other. He suggested using Blawx.

We need to determine which approach to use for the task and then implement it. Once it is ready, it should not be too difficult to add support for a tool like Flora-2.

shaolintl commented 4 years ago

Another idea from Jason:

If there was some way to have the original law and the pseudo-code version of the law in the same window, but visually distinct from one another, that might serve the purpose. I'm imagining something like "track changes" in Microsoft Word, where the stuff that has been changed and added is visibly distinct from what was original.

Then, if you had the law+pseudo-code on the left, and on the right you had some representation of the code that was actually being generated by the annotations, even if that code wasn't directly editable, that could work.

Ideally, it would be visually obvious which sections of code were being generated by which sections of text. Perhaps by color coordination, or just highlighting the opposite component when you hover over its pair.

Defeasibility would then be represented in pseudo-code, by inserting names for rules, and text like "rule x overrides rule y".