Plan of lecture and motivation

In this lecture, we are interested two things:

• Exploring if gradual semantics can be approximated using ML techniques. This could lead to the following benefits:
  • The creation of new gradual semantics that better reflect human reasoning
  • Studying how semantics that do not converge or are not defined on specific graphs (e.g., acyclic graphs) could behave.
• Studying the question of extension enforcement in argumentation. This could lead to the following benefits the creation of strategies to persuade agents in debates (by fetching new information or gathering proofs to strengthen particular arguments).

5B.1 Learning gradual semantics (Al Anaissy et al.)

5B.2 Enforcement Heuristics with Deep Reinforcement Learning (Craandijk and Bex)

5B.3 Conclusion