The 13th International Conference on Modeling Decisions for Artificial Intelligence Sant Julià de Lòria, Andorra September 19 - 21, 2016 http://www.mdai.cat/mdai2016

Submission deadline: DEADLINE EXTENDED: April 11th, 2016

Abstract: To be determined.

Abstract: The fundamentals of insurance are introduced and alternatives to risk measurement are presented, showing how to quantify the size and likelihood of future losses. Real data indicate that insurance companies observe a lot of small losses and only a few large claims or extremes, which occur very rarely. This is known as the skewness of the profit and loss variable, which is specially troublesome for risk quantification, but is successfully addressed with generalizations of kernel estimation. In connection with this approach, distorsion risk measures study the expected losses of a transformation of the original data. GlueVaR risk measures are presented. The notion of sub-additivity and tail-subadditivity is discussed and an overview of risk aggregation is given with some additional applications to insurance.

Abstract: Moderation poses one of the main Internet challenges. Currently, many Internet platforms and virtual communities deal with it by intensive human labour, some big companies –such as YouTube or Facebook– hire people to do it, others – such as 4chan or fanscup – just ask volunteer users to get in charge of it. But in most cases the policies that they use to decide if some contents should be removed or if a user should be banned are not clear enough to users. And, in any case, typically users are not involved in their definition. Nobel laureate Elinor Ostrom concluded that societies –such as institutions that had to share scarce resources– that involve individuals in the definition of their rules performed better –resources lasted more or did not deplete– than those organisations whose norms where imposed externally. Democracy also relies on this same idea of considering peoples’ opinions. In this vein, we argue that participants in a virtual community will be more prone to behave correctly --and thus the community itself will be "healthier"—if they take part in the decisions about the norms of coexistence that rule the community. With this aim, we investigate a collective decision framework that: (1) structures (relate) arguments issued by different participants; (2) allows agents to express their opinions about arguments; and (3) aggregates opinions to synthesise a collective decision. More precisely, we investigate two aggregation operators that merge discrete and continuous opinions. Finally, we analyse the social choice properties that our discrete aggregator operator satisfies.

Abstract: Bitcoin has emerged as the most successful cryptocurrency since its appearance back in 2009. However, its main drawback to become a truly global payment system is its low capacity in transaction throughput. At present time, some ideas have been proposed to increase the transaction throughput with different impact on the scalability of the system. Some of these ideas propose to decouple standard transactions from the blockchain core, and to manage them through a parallel payment network, relegating the bitcoin blockchain only for transactions which consolidate multiple of those off-chain transactions. Such mechanisms generate new actors in the bitcoin payment scenario, the Payment Service Providers, and new privacy issues arise regarding bitcoin users. In this paper, we provide a comprehensive description on the most relevant scalability solutions proposed for the bitcoin network, we analyse its impact on user's privacy and we outline some research challenges on that topic.

University of Skövde

MDAI - Modeling Decisions