Rationalization: A Neural Machine Translation Approach to Generating Natural Language Explanations

We introduce AI rationalization, an approach for generating explanations of system behavior as if a human had performed the behavior. We a rationalization technique that uses neural machine translation to internal state-action representations of an autonomous agent into language. We evaluate our technique in the Frogger game environment, an autonomous game playing agent to rationalize its action choices natural language. A natural language training corpus is collected from players thinking out loud as they play the game. We motivate the use of as an approach to explanation generation and show the results two experiments evaluating the effectiveness of rationalization. Results of evaluations show that neural machine translation is able to accurately rationalizations that describe agent behavior, and that are more satisfying to humans than other alternative methods explanation.