Traditional online comment boards suffer from significant communicative deficits, including informational cascades, scattered content, and low information-to-noise ratio. Recent events (e.g. Brexit, Trump campaign) have raised the profile of further distortions such as trolls, fake news and the polarization of online discourse. To address this, technologies that exploit argument visualization have been developed with the aim of improving the communicative quality of online interactions. In this paper, we explore the results of the Scholio project http://www.scholio.net/ that delivered a randomized controlled field experiment to over 1500 participants from across the world to test the feasibility and effects of adopting two argument visualization technologies to support quality discussion of online news. Pol.is combines user generated comments, survey responses, clustering of information, argument visualization and commenting. Deliberatorium takes a different approach, engaging participants in the construction of a structured argument map. Alongside the variations in platform, the experiment also tested an empathy inducing perspective taking treatment assigned at the group level and orthogonal to the platform treatments. Results indicate that, contrary to our hypotheses, neither the argument visualization platforms nor the empathy induction significantly reduced participants' reports of having negative feelings during the experiment. Deeper analysis indicates, however, that these effects are likely the result of participants' unfamiliarity with, and frustration about, the platforms. This suggests that developers of online commenting platforms that aim to reduce polarization and its negative consequences must remain cognizant of the limitations of complexity.
