In Values and the Reflective Point of View (2006), Robert Dunn defends a certain expressivist view about evaluative beliefs from which some implications about self-knowledge are explicitly derived. He thus distinguishes between an observational and a deliberative attitude towards oneself, so that the latter involves a purely first-person point of view that gives rise to an especially authoritative, but wholly non-observational, kind of self-knowledge. Even though I sympathize with many aspects of Dunn's approach to evaluative beliefs and also with his stress on the practical significance of self-knowledge, I argue that his proposal seriously misinterprets the role of observation and evidence within the first-person point of view and, derivatively, in the formation of evaluative beliefs.
