Building user argumentative models

Abstract

Knowing how a user builds his/her arguments during a discussion gives useful advantages if we want to assist the user or analyse his/her argumentative skills. This paper presents a novel mechanism to build user argumentative models, which captures the argumentative style to generate arguments. To this end, we observe how users generate arguments, and apply a generalised association rules algorithm to discover rules for argument generation. These rules depict the argumentative style of the user. They are composed of an antecedent, which represents the conditions to build an argument, and a consequent, which represents such argument. To evaluate this proposal, we show results obtained in the domain of meeting scheduling. We discovered interesting rules from a group of users discussing in that domain, and checked that about 60% of the arguments that users had generated in a test situation can be also generated from the rules previously learnt, at least partially. Finally, although this work focuses on modelling users' argumentative style, we discuss how this promising approach could be applied in different knowledge domains.

Assisting students with argumentation plans when solving problems in CSCL

“…The first ones depict the argument generation process as in rules, whereas the second represent the counterpart's acceptance of the argument. However, we can personalise these actions building a user argumentative model (Monteserin & Amandi, 2008), which captures the rules for argument generation that a user applies during the negotiation.…”

“…In particular, we use a planning algorithm based on preferences in which student preferences impact on the action selection process and, as a result, on the argument selection mechanism, which decides what argument must be uttered in a given situation when there are several alternatives to choose. Finally, we can use an argumentative user model (Monteserin & Amandi, 2008) to personalise the argumentation plan construction. This model is composed of the action that a user (in this study, the student) uses to generate arguments and it is built from the arguments that he/she uttered in previous argumentations.…”

A reinforcement learning approach to improve the argument selection effectiveness in argumentation-based negotiation

“…After improving the policy, the personal agent can assist the users by indicating which argument should be uttered according to the context. Moreover, we are going to integrate this approach with an approach to build user argumentative models [16] in order to incorporate to the model the preferences about argument selection.…”

Agents that Learn How to Generate Arguments from Other Agents

“…As we need argument generation rules with a variable format, it is not enough to use a traditional algorithm to mining association rules, since the arguments and the facts observed are expressed in constant terms [18].…”

“…In a previous work, an approach to build user argumentative models was presented [18]. A user argumentative model captures the argumentative styles of the user and is depicted by the argument generation rules that the user utilises during the argumentation.…”