I showed this video to the argumentation class that I took over for the last 6 weeks. This was shown (well most of it) just after everyone had done some in-class debates.
One of the biggest goals I have in teaching debate and argumentation is to address fact addiction. Students strongly believe that access to facts and repeating what they accessed is all that’s needed to resolve controversies and disagreements. If the other side rejects the facts, you just say them louder. Or you pause the debate for 10 minutes while you go print out the statistics you think will do the job.
Treating poor debate and argument performance not as failure, stupidity, inability, or the like does not help in teaching people how to argue. Treating it like an addiction, and addressing the causes of addiction to facts seems to be the way to go. More on that perhaps in another post, but for now I’ll say that being addicted to facts, like a lot of addictions, is a way to feel comfortable in a world that continuously exposes its contingency. The arbitrary symbol-systems we’ve invented to keep things together are pretty frail, and in a world where students have been taught that there are correct answers to everything, they simply need to be found, and science and math are the ways to find them, they feel pretty powerless against the big contingencies. Facts help them think there’s some bottom and avoid the idea that the elevator has no ground floor.
So I show this video - and the students do not like the IBM Debater. I ask them why.
“She sounds fake,” one student said (Note: If you watch the video you’ll see the IBM team chose a female sounding voice simulation for the AI and indicated that the AI should be called she during the debate).
“She’s just listing a bunch of facts about the issue,” another said.
I said, “Now you know how I feel” - and although I said it as a joke, it resonated with the class. They all paused, as it really sort of opened something up for them.
The conversation continued into a discussion about the other factors (bad pun) needed to do well in debate, and how those factors might be more important than factual information. This was a far cry from the first in-class debate, where most of the students said the debate would have been better if there were “more facts.”
Showing this video after students have tried to do some debates was a really accidentally inspirational teaching move. They saw for themselves what was lacking in their rhetorical performance by watching an AI debater do the same things they did. Although AI is really keen on teaching a system how to debate, what happened here was something akin to the Turing Test - a moment where we learned more about human capabilities than we did about making AI.
One student asked, “Why are they doing this? Having an AI that can debate doesn’t help anyone out at all, what help could it possibly offer?” A great question about the goals of AI, but also answered by the rest of the conversation. We see through AI faults that are hard to pin down; habits that we don’t know we have. I believe this is also a theory of how and why science fiction is so great, it allows us to see something about ourselves or society very clearly by mucking it up with some really unfamiliar context.
Our conversation turned toward this question of why make it - students brought up the Turing test and the belief that debate is a marker of human intelligence. If the AI can debate, then it’s intelligent. I pointed out that since debate is a learned thing, doesn’t that make us artificially intelligent? What part of our intelligence, the way we talk about it, is natural?
I suppose in answer to this question one could turn to the scholarship on argumentation coming from evolutionary psychology and cognitive psychology and say that yes, arguing is a part of natural human communication, but it serves communicative ends (not logic, reason, or rationality). Indeed, those are life-long learning pursuits for everyone as they rely on contingent and unique factors for their judgement and appreciation every time. But debating? That seems very un-natural. Most people use debate and argument as interchangeable terms, but this is a huge mistake that is responsible for a lot of grief out there. Argumentation is a communicative tool whereas debate is an epistemic tool is the best way to put it. But instead we tend toward the model of arguments live inside debates which are ways of determining what’s right or true, or a means of getting people to act in the interests of what’s right and true. Of course there are many problems with this definition namely that people can agree that one side won a debate and go on acting in their lives to the contrary of that decision.
So if AI must be taught how to debate, perhaps we are all AI, there is no “natural” intelligence, and if debating well is a marker of intelligence, that means we must invest the time and energy into creating intelligence among people through schools and all sorts of other programs.
The students came to another conclusion in the light of the Turing Test and now this - that it is more and more impossible to determine what makes a human qualify as human. Instead of hosting the debate for AI to attend and learn, we all realize we are late to the party instead. We are all learning, and intelligence, like it’s harbinger debate, is something we have to practice and work on together all the time.
Argumentation’s role is as a facilitator of communication whereas debate’s role is a check back on argumentation’s power, making sure that we realize the contingency of our knowledge. This is the approach I’m taking toward theory now when teaching the debate or argumentation course. Pretty good result from watching what was a proof of concept debate hosted by IBM.