A good decision is rational in a broad sense - it takes into account as much information as is reasonably available, attempts to anticipate outcomes and implications (or their likelihoods) accurately, and evaluates the relative merits of those outcomes. It is impossible to have complete information and knowledge relating to a decision; consequently it is always necessary to combine logical analysis with more intuitive modes of reasoning. In all cases, these processes use confidence-qualified knowledge as their raw material.
For example, a manager of a team may be faced with a budget shortfall and instructions from above to lay off one team member. The manager has relatively complete knowledge of the amount of the budget shortfall and the salaries and other costs associated with each individual on the team. She has varying levels of experience with each team member's productivity, skill sets, and cultural contribution, such experience varying mostly with duration of tenure. She also has some sense of the relationships among the team and the likely emotional reactions people will have to a termination in general as well as to particular terminations.
All of this knowledge contributes to the decision, but note that not all of it is of equal quality and confidence. A rational decision must take knowledge confidence into account. For example, one of the employees might be clearly the least productive, but also well-liked. The manager has a high level of confidence in the productivity impact of a termination, but less confidence in her prediction of the emotional reaction of the team. A low level of confidence on a crucial item may suggest gathering additional information if possible, but this cannot be pursued indefinitely, and at the moment of the decision one must always take knowledge confidence levels into account, even if only intuitively.
After the decision is made (the moment of decision itself is amusingly both vacuous and encompassing), implementation normally follows. For our purposes here, I assume the decision-maker and the implementer are the same, though interesting intricacies arise when the roles are separated. When other people are involved, the implementation can be quite complex, and may include persuasion, amelioration, motivation, and other leadership behaviors and skills. Essential among these is the confidence projected by the leader. Support for a decision (whether required in advance or important in arrears) will not be forthcoming if the leader is publicly tentative about the decision after it is made.
The extent of deviation between a decision-maker's public presentation of confidence after the decision, and his actual confidence in the correctness of the decision, is a question of ethics; by this I mean not just a purely "moral" ethics but also including the practical, long-term reputational impact. Let us not address the details of that question here, but simply assume that there will naturally be some difference.
We have now arrived at the heart of the matter. It is essential that the decision-maker does not propagate this publicly projected confidence backward to his knowledge confidence levels that figured in the decision itself. This would be a clear epistemic error and an irrational confusion, detrimental in two important ways: first, the particular confidence levels that are adjusted might later be used in future decisions, which will then have an implicit bias. Second, this behavior can easily become a progressive habit, whereby the leader mis-leads himself about his knowledge. Among other consequences, he will then be less inclined to seek further information at times when it is rationally warranted, and increasingly use uninformed instinctual processes. Ironically, the risk of this is greatest when a decision-maker has a history of making good and correct decisions, since such outcomes reinforce the public leadership confidence spread.
Returning to the previous example, our manager presents the termination to the team with a confident countenance, and erroneously increments her confidence in the productivity and cultural contribution of one of the remaining team members who was also being considered for termination. Later, she selects that team member over another for an important role, in part based on the incorrect confidence, and this decision backfires. Or, she fails to spend the requisite time required to familiarize herself with a new team member, because she began to confuse her confidence in her decision abilities with confidence in her knowledge.