In open multiagent systems, agents depend on reputation and trust mechanisms to evaluate the behavior of potential partners. These evaluations (social evaluations) indicate how "good" or "bad" a given target agent results to be in a certain context. Along with direct interactions and observations, communicated social evaluations are a valuable source of information for the computation of evaluations. They carry out though several problems, being the detection of false information the most studied one. However, false information is often confused with unaligned information: A communicated social evaluation from an agent "A" may not be reliable for "B", even when "A" is completely honest, because the bases under which "A" has inferred the social evaluation cannot be accepted by "B". This can happen because agents may be using different inference rules, have been involved in different experiences, have different goals etc. One possible solution compels the definition of a complete dialog protocol that allows the agents to argue about the communicated values instead of considering just the communication of a single value. In the talk, we will detail an argumentation-based dialog protocol specifically designed for the exchange of reputation-related elements. It uses an instantiation of a weighted version of the classical Dung's abstract argumentation framework to define an acceptability semantics, and help agents decide whether communicated evaluations can be considered reliable.