According to Lee Smolin
Like many academics, I spend a lot of time writing and reading confidential letters of recommendation; such secret information is the core of our system of hiring and promotions. Is this really the wisest way to inform our bets on who is going to do important science? The claim is that the advice you get is more honest than it would be in a system of open assessments, but is this really the case? The information one wants is often in the confidential letters, but so is a lot of exaggeration, bias and sloppiness, that would not survive some transparency. There is also the slowdown in the rate of progress due to the power differential which arises when older academics are confidentially evaluating the work of people significantly younger.
A system of open evaluations, in which candidates are allowed to see what is written about them, would take some getting used to, but it might lead to wiser decisions at less of a cost of time. I would also guess that a system of open evaluations would be weighed more favorably towards independent thinkers who do high risk/high payoff science than the present system, which, by its emphasis on confidential evaluations by senior scientists, is weighed towards scientists who follow well established research programs.