[en] We introduce Alhazen — an approach that automatically determines the circumstances under which a particular program behavior, such as a failure, takes place. Alhazen starts with a run that exhibits this behavior and automatically determines input features associated with the behavior in question: (1) We use a grammar to parse the input into individual elements. (2) We determine features from the elements such as existence, length, or numerical values. (3) We use a decision tree learner to observe and learn which input features are associated with the behavior in question. (4) We use the grammar to generate additional inputs to further strengthen or refute hypotheses as learned associations. (5) By repeating steps 2 to 4, we obtain a theory that explains and predicts the given behavior. In our evaluation using inputs for find, grep, NetHack, and a JavaScript transpiler, the theories produced by Alhazen predict and produce failures with high accuracy and allow developers to focus on a small set of input features: “grep fails whenever the --fixed-strings option is used in conjunction with an empty search string.”