<div dir="ltr"><div><div style="font-size:13px">Dear all,</div><div style="font-size:13px"><br></div><div style="font-size:13px">In this Thursday's game theory reading group, Dr Tim Baarslag will give us a talk about his research in negotiation, especially the application of Pandora Problem.</div><div style="font-size:13px"><br></div><div style="font-size:13px">Time/Venue: 21 May, 1:15pm - 2:00pm / B32 3073</div><div style="font-size:13px"><br></div><div style="font-size:13px">Online participation: <a href="https://sites.google.com/site/sotonecsreadinggroups/home" target="_blank">https://sites.google.com/site/sotonecsreadinggroups/home</a></div><div style="font-size:13px"><br></div><div style="font-size:13px">Title: Pandora's Problem - a technique from search theory to deal with uncertainty and cost.</div></div><div style="font-size:13px"><br></div><div style="font-size:13px">Abstract:</div><div style="font-size:13px">In this talk, I will introduce a problem from search theory called Pandora Problem. The problem deals with optimizing reward while searching through boxes with uncertain prizes, which can be uncovered against a certain cost. Pandora's Problem not only has a wide range of possible applications, it also has a surprisingly elegant and optimal solution. I will provide some intuition behind the solution and then show an application of this technique in the form of a negotiating agent that can adapt itself to the user. In such a negotiation setting, it is essential for the agent to understand the user's preferences, without exposing them to elicitation fatigue. I discuss a new model, using Pandora's Rule, in which a negotiating agent may incrementally elicit the user's preference during the negotiation. This yields an optimal elicitation strategy that decides, at every stage of the negotiation, how much additional user information to extract at a certain cost.<br></div><div style="font-size:13px"><br></div><div style="font-size:13px">To access all previous talks: <a href="https://sites.google.com/site/sotonecsreadinggroups/resources" target="_blank">https://sites.google.com/site/sotonecsreadinggroups/resources</a><br></div><div style="font-size:13px"><br></div><div style="font-size:13px">Best,</div><div style="font-size:13px">Dengji</div><div><div class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div>--------------------------------<br></div><div>Dengji Zhao<br></div><div>Research Fellow<br><div>Agents, Interaction, and Complexity Group (AIC)</div><div>School of Electronics and Computer Science </div><div>University of Southampton </div><div>Southampton SO17 1BJ, United Kingdom</div></div><div><a href="http://dengji-zhao.net" target="_blank">http://dengji-zhao.net</a><br></div><div><br></div></div></div></div></div></div></div></div></div></div></div></div>
</div>