And the final speakers at the Bots Building Bridges workshop project are Mathias Orlikowski and Tony Veale, who begin by noting that bots can have a positive impact on online discussion by facilitating an increase in viewpoint diversity. Such bots might emphasise rational debate, but could also introduce more humorous perspectives.
One example of a polarised debate in Germany is the ongoing discussion about the introduction of a stricter speed limit on German freeways; in online discussions about this, a discussion bot could intervene by presenting an opposite point of view. Some such arguments could be pre-written responses to well-known and commonly occurring discussion stances, too. Ideally these would present thoroughly founded and non-hostile challenges to the original stance, rather than counterarguments; the aim is not to persuade a participant of an opposing view, but simply to increase overall viewpoint diversity.
This was tested in situ on Twitter, comparing the bots posting such authoritative replies to bots posting much more generic and unspecific replies. The authoritative bot interventions generate considerably more further responses from human users, though some 39% of those responses were also making fun of the bot itself.
However, such authoritative and rational bot interventions are not the only option. Another project instead used LLMs to generate a humorous visual message responding to prior posts: this could use an XML-format description of the intended content and then feed this to the image generator; but alternatively, the visuals could also be created first, with the LLM responding to the image content as it generates the conversation text. This also makes the images somewhat reusable.
Such visual bot interventions may then speak to their audience in a different way from more rational interventions. Comics are an engaging and unthreatening medium for highlighting the different views in a debate.
