Moral evaluations influence our mental state ascriptions.

Gilbert Chin, in the "Editor's Choice" section of Science, gives a nice summary of work by Uttich and Lombrozo in their Cognition article:

A robust phenomenon established empirically during the past decade is the tendency of observers to regard morally bad consequences (such as harm to the environment) that occur as a secondary effect of actions taken by an agent (such as a corporate CEO) in the course of achieving the primary effect—an increase in revenue—as having been committed intentionally. In contrast, morally good consequences in a similar scenario are judged as being incidental. A number of explanations for this asymmetry (also known as the Knobe effect) have been put forth; most prominent, perhaps, is the proposal that the moral valence of the side effect alters the observer's inference about the agent's mental state; that is, whether the CEO acted with intent. Uttich and Lombrozo bring to bear a series of vignettes in which the type of social norm (moral versus conventional), the kind of behavior (norm-conforming versus norm-violating), and the outcome valence (helpful versus harmful) were varied independently. Their results support a "rational scientist" framework, so that the observer's computation of the agent's state of mind weights actions that flout commonly accepted rules of behavior as being more informative and hence diagnostic of intentionality than conformist ones.

Here is the abstract from the article:

Theory of mind, the capacity to understand and ascribe mental states, has traditionally been conceptualized as analogous to a scientific theory. However, recent work in philosophy and psychology has documented a “side-effect effect” suggesting that moral evaluations influence mental state ascriptions, and in particular whether a behavior is described as having been performed ‘intentionally.’ This evidence challenges the idea that theory of mind is analogous to scientific psychology in serving the function of predicting and explaining, rather than evaluating, behavior. In three experiments, we demonstrate that moral evaluations do inform ascriptions of intentional action, but that this relationship arises because behavior that conforms to norms (moral or otherwise) is less informative about underlying mental states than is behavior that violates norms. This analysis preserves the traditional understanding of theory of mind as a tool for predicting and explaining behavior, but also suggests the importance of normative considerations in social cognition.