EPSRC-funded research could improve quality of online reasoning

Researchers at the University of Dundee have developed the Analysis Wall – a four-square-metre touchscreen intended to facilitate more effective debate. In addition to providing a tactile tool to support focused discourse, it is hoped that the project will serve to improve the quality of reasoning online.

The Analysis Wall, which was funded by the Engineering and Physical Sciences Research Council (EPSRC), can be used to plot the path of a debate. Acting as a shared workspace, the device allows analysts to rearrange different arguments and contributions in order to formulate easy-to-follow narratives for complex discussions. The touchscreen has already been road tested in conjunction with the BBC Radio 4 programme The Moral Maze. A team of analysts and stenographers placed each of the participants’ comments onto the screen and rearranged them into a more coherent structure.

I had the opportunity to interview project leader Professor Chris Reed from Dundee’s School of Computing. I began by asking him to talk me through the development of the Analysis Wall.

"For the last decade or so, my research group has been trying to understand the ways in which people argue, disagree and express their opinions," Professor Reed explained. "We have been building philosophical and linguistic models of these processes. More recently, the EPSRC has supported our efforts to develop more practical applications for these models. We had the idea to develop the ‘Argument Web’. As its name suggests, this system is capable of connecting the various claims, counterclaims, reasons and critiques that are proffered during the course of a debate. This sort of system could prove useful both when talking about issues of global interest, such as climate change, and those that affect people locally, such as where an electricity pylon should be situated. Debate, whatever its subject, involves a wide range of aspects. We wanted to connect these aspects to make them more accessible and concrete."

Professor Reed went on to outline how the software utilised by the Analysis Wall could be exploited to improve the standard of online debate.

"Consider online message boards and forums," he began. "The quality of the arguments that take place within these spaces is not always as good as it might be. I don’t think that this indicates a widespread inability to argue; quite the contrary. You can actually see examples of people trying to circumvent the limitations of the technology that they are using. You see participants using OP and PP to represent original and previous posts. You see people using @ to identify the particular person that they are addressing. Users are inventing their own tools because the currently-available technology does not allow them to argue effectively. This is utterly mad! We should be designing technologies to facilitate and encourage high quality argument. This idea led to the Argument Web and the Argument Web led to the Analysis Wall."

Chaired by Michael Buerk, The Moral Maze offers lively debate for Radio 4 listeners as panellists argue over the ethical conundrums of our time. Professor Reed and his colleagues identified the programme as an excellent opportunity to test drive their new device in real time.

"When developing the Analysis Wall, our aim was to cultivate gold-standard argumentation," he explained. "We looked around and found an excellent existing source of argumentation in the form of The Moral Maze. Each week, panellists engage in a very structured and apparently logical discussion concerning a particular topical issue. The discourse is fantastically rich and sophisticated, which made it absolutely ideal for our purposes.

"The traditional method for analysing such a debate would have been for a highly-trained researcher to hide away for three or four weeks before returning with their results. This is fine from an academic point of view, but what if you want to get people to engage with and join the discussion in real time? Obviously, you will need something much faster. Previously, we might have simply thrown a lot of people at this problem. The trouble with this approach is that a debate is not easily parallelisable. You can’t tell one person to take the first two minutes, another to take the second two, and so on. There are too many interconnections. What is being said now might link back to something that was said five minutes ago, and this might well link to something else that has yet to be said. In light of this, you need a big, collaborative analysis base where people can come together and talk about their analyses in real time.

"When we first started working on the project, it wasn’t even clear whether this would be possible," Professor Reed continued. "The large screen and bespoke software that comprise the Analysis Wall, however, make it very quick and easy to conduct intuitive analysis. This is vital because the cognitive load of analysing in real time is extremely high. During the broadcast, a team of stenographers in London sent a live text stream into our laboratory. Two of our analysts were on hand to segment the text stream into the individual molecules that comprised the debate. These components then appeared on the analysis wall. A further seven or eight analysts rearranged the molecules to construct an easy-to-follow representation of the debate in real time."

Finally, I asked Professor Reed what he and his colleagues plan to do next. He is excited by the wide range of potential applications offered by the technologies that his team has developed.

"There are a number of avenues that I find particularly exciting," he explained. "The first is crime-scene investigation. In order to reconstruct events, police officers must create a narrative relating to what might have happened. When you are investigating a crime scene, new evidence is uncovered frequently. Assembling and critiquing this evidence is, essentially, the same task that we were performing in conjunction with The Moral Maze.

"Another hot topic that we have started to explore is the domain of intelligence. Intelligence analysts can become utterly swamped by the vast swathes of information that they receive. One of their most vital roles is to judge the reliability of human intelligence data. This involves teasing apart reasoning in real time, and so a large and collaborative workspace might well be ideal for their purposes."