Gayani Dias Pathirathna Senanayake: Discovering the Recipe Behind Real-World Patterns: A Human-Centred IGSS Framework

<p dir="ltr">We present a framework for inverse model discovery that starts from observed patterns and works backwards to uncover simple, human-readable rules. Rather than hand-coding behaviour, we use evolutionary search to generate many candidate “recipes” from theory-compatible ingredients (e.g., distance, incentives, social influence). Each candidate is evaluated through two gates: <br>(1) empirical fit (does the simulated outcome reproduce the data? ) and (2) theoretical plausibility (is the rule sensible under established domain knowledge and theories) To scale expert involvement, we combine targeted expert review with an AI plausibility rater trained on expert rationales, filtering out “absurd-but-fitting” rules early and focusing attention on interpretable finalists. Illustrative examples show how data-only selection can admit unrealistic generators, while our dual-gate approach yields compact rules that generalise and support explanation. The result is a transparent, human-centred IGSS workflow that reduces modeller bias and turns complex patterns into actionable, trustworthy insights.</p>