1. Proposing new ways of resolving online conflicts: an intelligent facilitation of forgiveness in CMC Asimina Vasalou, Jeremy Pitt Intelligent Systems and Networks Group Electrical and Electronic Engineering Department Imperial College London Presented by Paolo Petta Humaine WP8 Workshop

2. Overview  Online communities  Norm mechanisms  Forgiveness –Why it is important –Model proposal  Application domain

3. The broader picture  Online communities –MUDS, emotional support groups, social networks, seller-buyer  Recent media theories have mostly focused on the benefits of online communication –Shyness –Decreasing social distance  equal contributions –Anonymity  social identity  group goals –Role playing towards healing in RL

4. The problem  What happens when users have divergent goals? –Hostility –Deception –Online rape

5. Examples of current solutions  Human moderator  Peer to peer recommendations  Trust and reputation mechanisms  Successful?

6. Our claim  The “quantification” of human behavior (i.e. performance ratings) removes important human coping mechanisms which in physical worlds add value to human relationships and provide closure during their disruption

7. Why forgiveness  Law, psychology, theology and organizational management  Motivations –Healing for victim and offender –Reversal of action –Unjust punishment  anger, low compliancy behaviors –Issuing forgiveness  voluntary actions of repair –Health

8. What is forgiveness  A number of positive motivational changes which reverse one’s initial desire to adopt negative strategies towards the offender

9. What is forgiveness (cont.)

10. 1. Judgment of offence  Severity of action  Frequency of action  Intent

11. 2. Reversal and Restitution  Apology predicts forgiveness  Reversal of an offence with a good deed

12. 3. Historical Interactions  History –Prior familiarity –History of commitment –Costs or benefits of previous interactions

13. 4. Empathy  Empathy predicts forgiveness and its intensity correlates with the amount of forgiveness issued –Apologies  Empathic embarrassment “imagining oneself in another’s place” Offender’s visible embarrassment Some prior-familiarity Similarity in personality or characteristics (e.g. culture) Victim’s propensity to embarrassment

14. Challenges in building a model  Theoretical work has looked at each motivation individually  How does one motivation weigh against the other and which one is most influential?

15. Our approach  Fuzzy Inference System –Each motivation is a separate decision maker, as is the final forgiveness inference –Each constituent motivation (e.g. intent) can carry different weights  How does the model collect the constituent values? –Computed –Supported by the interface

16. Visualizing the model

17. Domain  Distance learning community  Team activities, assignments etc  Violations consist of not delivering assignments, being late with delivering work, not communicating efficiently etc.

18. A hypothetical scenario  Alice delivers low quality work to her teammate Bob who in turn rates her negatively. Upon receiving this rating, the forgiveness mechanism is instantiated and presents Alice with the possibility to repair her offence i.e. by offering apology and reparative action outlets. Alice chooses to apologize. The forgiveness mechanism then computes all the motivations together and recommends that Bob forgive Alice. Bob is shown all the relevant information on his screen.

19. Forgiveness facilitation

20. Future work  Evaluating the weights of the computational model with questionnaires  Designing the tool so that it does not mislead users (e.g. colloquial understanding of forgiveness)  Giving this application to users to determine whether they will use it while in a conflict  Uncovering the connection between current systems and users aggression in contrast to a forgiveness application

21. Questions?  Email minav@luminainteractive.comminav@luminainteractive.com  Special thanks to Paolo Petta for presenting this work