The chatbot shouldn't respond this! Words have a meaning and thus an effect. They may solve problems or create problems. They may be beneficial or not. If a dialogue system is not able to evaluate the consequences of words it may cause damage. This is the lesson drawn from chatbots that "naively" replicated phrases from ill-intentioned users.

The debate on Ethics in AI seeks to address these issues. An ethical AI system needs to be able to evaluate the consequences of possible actions and to take them into account when choosing an action. Finding a good enough action simply is not good enough! Creators of chatbots should ensure that their systems respect some ethical preferences (e.g. prefer phrases that may solve problems to phrases that may create problems) when choosing responses.

Those ethical preferences may be categorical and follow another learning cycle than that for learning preferences about a user's choice behavior. The latter are about objects in the world such as movies to watch, items to shop, or places to visit. Hence, those preferences concern concrete objects and a preference learner needs to take various characteristics of those objects into account. Ethical preferences, however, concern important consequences of actions. Indeed they might not be learned from observed data, but either be pre-programmed, taught, or acquired as insights about decision making behavior in difficult situations. 

AI and Ethics may thus lead to new questions for preference learning and preference elicitation. It will be interesting to see whether the topic of ethical decision making will be studied at any of the forthcoming events on preference handling listed below.

The 10th Workshop on Advances in Preference Handling M-PREF 2016 will be held in conjunction with the 25th International Joint Conference on Artificial Intelligence IJCAI-16, which takes place from July 9th to 15th, 2016 in New York City, USA. The workshop organizers are Markus Endres (University of Augsburg), Nicholas Mattei, (Data61 and UNSW), and Andreas Pfandler (TU Wien and University of Siegen). As in recent years, the workshop covers preference handling in artificial intelligence, preference handling in database systems, preference handling in multiagent systems, applications of preferences, preference elicitation, preference representation and modeling, properties and semantics of preferences, and practical preferences. The deadline for paper submission has been extended to May 8th, 2016. The precise date of the workshop will be announced as part of the IJCAI schedule.

The 28th European Conference on Operational Research EURO 2016 will be held on July 3-6, 2016 in Poznan, Poland. Three of the conference streams are of interest for preference handling, namely the stream on preference learning chaired by Krzysztof Dembczynski, Salvatore Greco, and Roman Slowinski, the stream on multiple criteria decision aiding chaired by Salvatore Corrente, José Rui Figueira, Milosz Kadzinski, and the stream on multiple criteria decision analysis chaired by Milosz Kadzinski, Chin-Tsai Lin. The conference also features an invited talk related to preference handling. The title of the talk is “Preference elicitation and learning in a Multiple Criteria Decision Analysis perspective: specificities and fertilization through inter-disciplinary dialogue" and it will be given by Marc Pirlot (Univ. of Mons).

Multiple criteria decision aid and preference learning are also the topics of the EURO Mini Conference DA2PL'2016 (From Multiple Criteria Decision Aid to Preference Learning), which provides an occasion to researchers from decision analysis and machine learning to discuss topics and research challenges that are of common interest for both fields. The mini conference will be held on November 7-8, 2016 in Paderborn and organized by Eyke Hüllermeier. Submissions are due on August 22, 2016.

Other conferences of interest for preference handling include the International Conference on Autonomous Agents & Multiagent Systems AAMAS 2016, which will be held on May 9-13, 2016 in the School of Information System (SMU) in Singapore, the Sixth International Workshop on Computational Social Choice COMSOC-2016, which will take place in Toulouse, France on June 22–24, 2016, and the 2nd Workshop on Conflict Resolution in Decision Making COREDEMA 2016. The latter will be held in conjunction with the 22nd European Conference on Artificial Intelligence ECAI 2016 in Den Haag, The Netherlands on August 29-30, 2016. Furthermore, AAMAS 2016 hosts the 3rd Workshop on Exploring Beyond the Worst Case in Computational Social Choice EXPLORE-2016 on May 10th, 2016.

Finally, we give an outlook on our major event next year. In 2017, the working group on preference handling will meet at the 5th International Conference on Algorithmic Decision Theory ADT 2017 in Luxembourg on October 25-27, 2017.