Learning Heuristics HAL builds heuristics by comparing the two steps. HAL then generalises these heuristics. HAL learns by experience which heuristics are useful.