פורום הדוקטורנטים ייערך
בתאריך 6.4.16 בשעה 17.00 - מרצה מר אבשלום אלמליח
Overcoming People's Irrationality Using Agents
I will talk about innovative methods for collaborative agents for better-assisting people facing complex decision situations. In many decision-making situations, people's decision is different from the theoretical optimal one. There are various explanations for the phenomena, some of these are psychological such as peoples' tendency to gather facts that support certain conclusions and disregard those that do not. Other relies on people's computational capabilities, meaning that in some decision situations it is difficult (computationally) for people to calculate the optimal policy.
I will demonstrate the use of two automatic methods for assisting people in their decision-making process in various domains. The first deals with restructuring information about choices to improve the performance of agents/people on recurring sequentially dependent decisions. The domain in which this method was evaluated simulates a real-life decision-making scenario where people engaging search in platforms such as electronic marketplaces need to decide which opportunity to explore next and when to terminate their search process.
The second method for assisting people in their decision-making process is suboptimal advising that can be considered as a new paradigm for automated advice provisioning agents. Contrary to what was previously thought, this work empirically demonstrates that agents that dispense optimal advice will not necessary facilitate the best improvement in people's strategies. I will provides results demonstrating the effectiveness of a suboptimal advising approach in extensive experiments in two canonical mixed agent-human advice-giving domains.