Thinking About Thinking: Evolutionary, Developmental, and Epistemological Aspects

Converging evidence from comparative psychology suggests that some non-human species have the ability to offer reliable confidence judgements about what they can  remember or perceptually discriminate. There is also some evidence that young children are sensitive to the fluency of perceptual contents and to existing consensus about a belief, before they develop the capacity to metarepresent their own ability of perceiving or of forming beliefs.  This raises the issue of the preconditions for controlling one's cognition, i.e. for having metacognitive abilities. On a classical view, metarepresentations involving mental concepts are needed to represent normative awareness about first-order representational contents ("I evaluate my [belief that P, reasoning that R, categorizing as C] as [true, relevant, etc.]". Such metarepresentations are claimed to apply in a similar way to self- and other- attribution of mental states. Formal results about belief revision and conditional belief, however, indicate that metarepresentation does not capture how introspection should work. Furthermore, different cognitive processes seem to be used, and different judgments reached, when evaluating one's own chances to succeed while performing oneself a mental task, versus evaluating others' while observing them performing the same task.  
An alternative explanation emerges from metacognitive studies. It is hypothesized that a set of procedures and heuristics is implicitly learned by subjects in order to control their mental actions (such as trying to remember, or trying to discriminate), similar to the procedures and heuristics used in the control of one's physical actions. In both cases, comparators deliver evaluations of the anticipated need and feasibility of the action, or of the observed results attained once it is performed.  Just as proprioceptive signals play a major role in evaluating one's physical actions, emotional signals ("epistemic feelings") express heuristic-based evaluations for mental performances, such as remembering or discriminating. Mental agents, on this view, have an immediate subjective, phenomenological access to the comparator’s verdict. They can use such feelings to guide their epistemic decisions, even though they are not able to represent concepts such as representation, memory, or accuracy.
A promising line of research consists in identifying the various  epistemic norms to which human children and adults or non-human agents are implicitly sensitive, by studying their practical abilities for monitoring and controlling their own cognition in different contexts. A second important topic is to examine whether training in specific mental tasks might  allow agents to develop metacognitive abilities that they would not otherwise acquire.