------ VOCAB ------
Feature lava_pit := ...
Feature current_position :=
Feature goal_position :=

Policy keep_walking := ...
Policy avoid_lava_pit := ...


------ USING GPT3 ------ * denotes a generated line
Original: it is a good idea to keep_walking if there is a lava_pit less than 4 spaces away
Rephrased: keep_walking is a good strategy when there is a lava_pit less than 4 spaces away

Original: avoid_lava_pit is a tactic where you turn_around if your current_position is next to a lava_pit and your current_direction is facing the lava_pit
Rephrased: if your current_position is within 1 space of a lava_pit and your current_direction is facing the lava_pit, avoid_lava_pit is a good policy to follow

Original: if goal_position minus current_position is decreasing then reward increases by 1
Rephrased: *if the goal_position is getting closer to the current_position, the reward increases by 1

Original: *if lava_pit is within 3 spaces of the current_position then reward decreases by 1
Rephrased: *if there is a lava_pit within 3 spaces of the current_position, the reward decreases by 1

------ OBSERVATIONS ------
- this vocabulary is small?
- it's hard to manually rephrase something
- do we need policy names?
