The deliberation system approach is currently one of the most discussed in democratic theory. Its attractiveness is determined by the fact that it does not set too high expectations towards citizens and does not call for the establishment of ideal conditions for deliberation (Parkinson 2006; Mansbridge et al. 2012). It takes into account the existing epistemic inequalities between citizens, adopting the familiar idea of an epistemic 'division of labour' (Kitcher 2001; Thompson 2006; Christiano 2008, 2012). In my paper I will focus on the place of expertise in this model, arguing that the fact that laymen are 'epistemically dependent' on experts makes the system approach much more problematic than it is usually admitted. Unlike some authors (Holst, Molander 2017, 2019), I will argue that this 'epistemic dependence', detrimental to any form of deliberation, cannot be overcome by any institutional design. The recent findings in social epistemology (Goldman 2011) and STS (Collins, Evans 2007, 2017) show that any potential solution to this problem is unreliable.
