In the past, we emphasized the important role of transitive delegations for the idea of Liquid Democracy: stripping transitivity from Liquid Democracy would give advantage to people according to their technical abilities and/or social integration, while a fully-transitive delegation model allows an equal treatment of all voters regardless of whether they are directly participating in a vote or delegating the decision to experts. [PLF, subsection 2.4.2]
Occasionally, we were inquired as to whether it was possible to incorporate the idea of a preferential delegation model into LiquidFeedback. Due to the previously stated reasons (equal treatment of all voters), the transitivity would need to be combined with (and not replaced by) a preference model where each voter can provide a list of delegates for a single issue instead of selecting one particular person as delegate.
At first, our main reason to not incorporate such a feature was the complexity of combining transitivity and precedence (e.g. the impossibility to use simple delegation chains for graphic representation). But recently we discovered that extending LiquidFeedback by adding a preferential delegation model would always break certain (mathematical) properties of the system. The two articles of this issue #3 will deal with the proof and the consequences for online decision-making systems.