₦airaland Forum

So I built an ai expert system on a legal domain focusing on homicide basically. It's knowledge based logic not an LLM. It's just raw logic but rule based basically, the earliest and plainest form of Artificial Intelligence currently but it's as accurate as the data provided. More accurate than most general AI chatbots for specifically trained knowledge based task. Try it out here, tell me what you think
https://verdant-bonbon-781b15.netlify.app/

Rule-based expert systems are underrated for this kind of thing. LLMs hallucinate legal definitions but a proper inference engine with well-defined rules will give you consistent outputs every time.

Did you model the rules from Nigerian criminal code specifically or more general common law? Curious how you'd handle edge cases like partial defenses — like if intent exists but there was provocation, does it reduce from first-degree to manslaughter in your system?

Common law yes. Haven't really thought of edge cages. And yes, this is far better than Gen AI cause of basically no hallucinations. I'm currently working on a research where both LLM and expert system can be combined, a hybrid. Bringing in the best of both worlds basically

as69:
Rule-based expert systems are underrated for this kind of thing. LLMs hallucinate legal definitions but a proper inference engine with well-defined rules will give you consistent outputs every time.

Did you model the rules from Nigerian criminal code specifically or more general common law? Curious how you'd handle edge cases like partial defenses — like if intent exists but there was provocation, does it reduce from first-degree to manslaughter in your system?