Preserving Privacy in a (Timed) Concurrent Language for Argumentation
0 likes25 views
The document discusses a timed concurrent language for argumentation developed by Bistarelli, Meo, and Taticchi, which integrates privacy preservation into multi-agent decision-making processes. It outlines the framework of abstract argumentation, local stores for privacy, and illustrates its application through examples of agent interactions and decision-making scenarios. The conclusion emphasizes the functionalities of this language and proposes future enhancements to simplify model construction and extend applications in real-world scenarios.
![CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks
Example: Multi-Agent Decision Making
with Privacy Preserved problem
• Charlie’s, Alice’s and Bob’s beliefs1
13
[1] Yang Gao, Francesca Toni, Hao Wang, Fanjiang Xu: Argumentation-Based Multi-Agent Decision Making with Privacy Preserved. AAMAS 2016: 1153-1161](https://image.slidesharecdn.com/taticchi1cilc2024-240818092358-50457d1c/85/Preserving-Privacy-in-a-Timed-Concurrent-Language-for-Argumentation-30-320.jpg)
![CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks
Example: Multi-Agent Decision Making
with Privacy Preserved problem
• Charlie’s, Alice’s and Bob’s beliefs1
• Acceptable solutions:
13
[1] Yang Gao, Francesca Toni, Hao Wang, Fanjiang Xu: Argumentation-Based Multi-Agent Decision Making with Privacy Preserved. AAMAS 2016: 1153-1161](https://image.slidesharecdn.com/taticchi1cilc2024-240818092358-50457d1c/85/Preserving-Privacy-in-a-Timed-Concurrent-Language-for-Argumentation-31-320.jpg)
Preserving Privacy in a (Timed) Concurrent Language for Argumentation
- 1. Preserving Privacy in a (Timed) Concurrent Language for Argumentation Stefano Bistarelli, Maria Chiara Meo and Carlo Taticchi CILC 2024
- 2. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Overview • Abstract Argumentation Frameworks + Labelling • Timed Concurrent Language for Argumentation • Locality semantics • Preserving Privacy in Multi-Agent Decision • Conclusion & Future Work 2
- 3. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Abstract Argumentation • Represent and evaluate arguments • Abstract Argumentation Framework F = ⟨Arg, R⟩ 3
- 4. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Abstract Argumentation • Represent and evaluate arguments • Abstract Argumentation Framework F = ⟨Arg, R⟩ 3
- 5. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Abstract Argumentation • Represent and evaluate arguments • Abstract Argumentation Framework F = ⟨Arg, R⟩ • Argumentation Semantics (e.g. Labelling) 3 An argument is: • IN if it only attacked by OUT • OUT if it is attacked by at least one IN • UNDEC otherwise
- 6. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 7. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 8. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 9. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 10. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 11. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 12. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 13. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Argumentation-based communication between concurrent agents sharing a common store • Syntax: Timed Concurrent Language for Argumentation (TCLA) 4
- 14. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Parallel executions 5 • True concurrency: we assume infinite processors
- 15. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Parallel executions 5 • True concurrency: we assume infinite processors • Global clock for the the passing of time
- 16. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Parallel executions 5 • True concurrency: we assume infinite processors • Global clock for the the passing of time • We decrement the timeout environment T : ℐ ⇀ ℕ ∪ {∞} I 𝑇(I) Agent A 4 Agent B 1 Agent C 2
- 17. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks True concurrency • With 6
- 18. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks True concurrency • With • Possible implementation * (F, F′, F′′) := (F′∩ F′′) ∪ ((F′∪ F′′)∖F) 6
- 19. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks True concurrency • With • Possible implementation * (F, F′, F′′) := (F′∩ F′′) ∪ ((F′∪ F′′)∖F) • Alternative approach: interleaving 6
- 20. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Addition & removal 7 • Example: add({a,b},{(a,b)}) -> rmv({a},{}) -> success;
- 21. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Addition & removal 7 • Example: add({a,b},{(a,b)}) -> rmv({a},{}) -> success;
- 22. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Addition & removal 7 • Example: add({a,b},{(a,b)}) -> rmv({a},{}) -> success;
- 23. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Check 8
- 24. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Check 8
- 25. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Check 8
- 26. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Test semantics are similar to the check one but for the conditions to satisfy • Credulous test: ∃L ∈ ℒF σ ∣ L(a) = l • Sceptical test: ∀L ∈ ℒF σ ∣ L(a) = l Test 9
- 27. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks • Test semantics are similar to the check one but for the conditions to satisfy • Credulous test: ∃L ∈ ℒF σ ∣ L(a) = l • Sceptical test: ∀L ∈ ℒF σ ∣ L(a) = l • Example: ctest(2,{b},IN,complete) / stest(2,{s},IN,complete) Test 9
- 28. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Locality semantics • new 𝑆 in 𝐴 behaves like 𝐴 where arguments in 𝑆 are local to 𝐴 • 𝐴𝐹𝑙𝑜𝑐 contains information on 𝑆 which is hidden from the external 𝐴𝐹 12
- 29. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Locality semantics • new 𝑆 in 𝐴 behaves like 𝐴 where arguments in 𝑆 are local to 𝐴 • 𝐴𝐹𝑙𝑜𝑐 contains information on 𝑆 which is hidden from the external 𝐴𝐹 12 AF ↑ S S
- 30. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Example: Multi-Agent Decision Making with Privacy Preserved problem • Charlie’s, Alice’s and Bob’s beliefs1 13 [1] Yang Gao, Francesca Toni, Hao Wang, Fanjiang Xu: Argumentation-Based Multi-Agent Decision Making with Privacy Preserved. AAMAS 2016: 1153-1161
- 31. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Example: Multi-Agent Decision Making with Privacy Preserved problem • Charlie’s, Alice’s and Bob’s beliefs1 • Acceptable solutions: 13 [1] Yang Gao, Francesca Toni, Hao Wang, Fanjiang Xu: Argumentation-Based Multi-Agent Decision Making with Privacy Preserved. AAMAS 2016: 1153-1161
- 32. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks tcla for DMPP • We can write a tcla program emulating a DMPP problem using tcla agents in parallel N • Each agent builds its local framework by using an add step • The computation starts in the initial public argumentation framework 14
- 33. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Example 16 …
- 34. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Translation • 𝐴𝑔𝑒𝑛𝑡1 checks whether its preferred action choice is globally feasible ctest(1,{a},IN,admissible) a • If this is the case, 𝐴𝑔𝑒𝑛𝑡1 adds the 𝐴𝑔𝑒𝑛𝑡1:𝑎 to the public AF and checks its partial consistency, namely ∃s ∈ 𝑆𝑜𝑙 | ctest(1,{s},IN,admissible) • If 𝐴𝑔𝑒𝑛𝑡𝑖:𝑎 is consistent, either continues with other agents or terminate with success • If 𝐴𝑔𝑒𝑛𝑡𝑖:𝑎 is not consistent, 𝐴𝑔𝑒𝑛𝑡𝑖 removes it from the public AF • If no action is found which can be extended to find a solution, the computation terminates with failure 15
- 35. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Conclusion • Functionalities of the Timed Concurrent Language for Argumentation which can be used to implement decision- making processes 17
- 36. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Conclusion • Functionalities of the Timed Concurrent Language for Argumentation which can be used to implement decision- making processes • Local stores for enforcing privacy between agents 17
- 37. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Conclusion • Functionalities of the Timed Concurrent Language for Argumentation which can be used to implement decision- making processes • Local stores for enforcing privacy between agents • Illustrative example demonstrating how the Timed Concurrent Language for Argumentation can be used for modelling DMPP problems 17
- 38. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Conclusion • Functionalities of the Timed Concurrent Language for Argumentation which can be used to implement decision- making processes • Local stores for enforcing privacy between agents • Illustrative example demonstrating how the Timed Concurrent Language for Argumentation can be used for modelling DMPP problems • Automatic translation from a DMPP problem to a tcla program 17
- 39. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Future Perspectives • Explore other features of tcla to simplify the construction of the models and achieve more natural interactions with the native constructs 18
- 40. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Future Perspectives • Explore other features of tcla to simplify the construction of the models and achieve more natural interactions with the native constructs • Further develop illustrative examples to showcase the system’s effectiveness and highlight limitations across various scenarios 18
- 41. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Future Perspectives • Explore other features of tcla to simplify the construction of the models and achieve more natural interactions with the native constructs • Further develop illustrative examples to showcase the system’s effectiveness and highlight limitations across various scenarios • Extend tcla to model real-world applications where agents can coordinate autonomously and concurrently without being bound to a fixed agent ordering 18
- 42. CILC 2024 Deriving Dependency Graphs from Abstract Argumentation Frameworks Future Perspectives • Explore other features of tcla to simplify the construction of the models and achieve more natural interactions with the native constructs • Further develop illustrative examples to showcase the system’s effectiveness and highlight limitations across various scenarios • Extend tcla to model real-world applications where agents can coordinate autonomously and concurrently without being bound to a fixed agent ordering • Endow the agents with a notion of ownership to establish which actions can be performed on the shared arguments 18
- 43. Preserving Privacy in a (Timed) Concurrent Language for Argumentation Stefano Bistarelli, Maria Chiara Meo and Carlo Taticchi Thank you for your attention!