764.2
Argumentative Agents with an Empirical Test
Such model highlighted some limitations of abstract argumentation that motivated us to further our investigation by consider structured argumentation.
We defined a new, cognitively-aware computational argument model. These model reflects patterns of reasoning and communication observed in a number of works in empirical psychology. We consider the set of experiments done by Hugo Mercier (Trouche, Sander & Mercier, 2014) aimed at measuring the effect of a single argument on intellective tasks. We describe our model and experiments with one of these logical problem, defined as:
“A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost?”
Mercier’s experimental results showed that good arguments can change people’s minds on intellective tasks.
Our agents engage in dyadic dialogues, with a typical turn-taking structure, up to a point when they stop because: they agree in the intuitive but wrong answer, or they agree on the correct answer, or they agree on the correct answer, but without having really understood the problem.
The results of our simulations agree with what observed in Mercier’s experiment. Circa 50% of our initially wrong agents change their mind when compelled with arguments favoring the right answers.
ABM models in Sociology rarely explicit the level of actor’s reasoning: actors either adopt some conclusions or they don't, because of some sort of “pressure” that somewhat wins their resistance (many mechanisms have been proposed in the literature). Our research instead puts an emphasis on reasoning.
The empirical results obtained so far make us believe that our approach could open new avenues in opinion dynamics research.