A DSL for Explanatory Decision Making

on YouTube ↗15m

" summary

A Haskell-embedded DSL called Mad Max encodes multi-attribute, multi-layer (AHP) decisions and produces numeric justifications: it normalizes attributes to 0–100, computes component-wise scores, and emits valuation traces that break the 50.36 score for Honda into contributions (e.g., price-personal = 15.43). It extracts minimal dominating sets to show the smallest components that flip a result and runs sensitivity analysis that reports concrete thresholds (Honda +$1,085.34, BMW −≈$1,100, Toyota −$20,500) to swap rankings.

AI-written summary · about

" chapters (15)

1. 0:00Mad Max decision making intro
2. 1:06Car purchase decision in Mad Max
3. 2:17Normalization in Mad Max
4. 3:48Comparing attribute values
5. 4:41Adding alternatives with union
6. 5:26AHP in Mad Max
7. 6:40Hierarchical decision maps
8. 7:21Final scores and winner
9. 7:43Valuation traces for explanation
10. 9:17Cleaning valuation traces
11. 10:04Element-wise differences
12. 10:54Minimal dominating sets
13. 11:43Focusing explanations
14. 12:58Sensitivity analysis
15. 14:09Conclusion & Q&A