Here are some research topics in which I am especially interested.
Cooperation in Multi-agent systems
I
am very interested in models of multi-agent cooperation and
teamwork---especially, the role that information plays in such models.
Most realistic settings nowadays are characterized by uncertainty;
hence information plays a key role in the ability of agents to achieve
their goals in an efficient manner. In that sense, acquiring and
sharing information often requires that some resources be consumed,
hence agents need to carefully reason about what information they
should obtain and share with others when working together. The methods
I work on apply to both cooperative and self-interested agents. In
particular, I put significant emphasis on the detrimental effect that
information sharing can have in models of multi-agent cooperation, if
not managed properly.
Economic paradigms
Much
of my work deals with electronic markets, auctions and mechanism
design, bargaining and negotiation. In these models I tend to focus on
the way the equilibrium analysis changes upon augmenting the legacy
model by introducing information broker/provider in the MAS. Alongside
the analysis of individual strategies and social welfare aspects this
research provides important foundations for theories of information
providers in multi-agent systems, the way they should price their
services and the influence of their existence and pricing on the MAS as
a whole. It turns out that in many cases, the sole existence of the
information provider actually lowers the benefit of the agents, and
social welfare can be substantially improved by pushing the information
provider off the market.
Agent-based simulation
My
interest in agent-based simulation is mainly in finding ways for
reliable representation and modeling of people and bounded rational
agents acting on behalf of people. One turn-key technology I am
especially interested in is PDAs. PDAs stands for peer-designed agents
– computer agents developed by typical programmers, not experts in
decision-making. In recent research, I have managed to demonstrate the
effectiveness of this technology in generating a variety of realistic
strategies and behaviors and how it can be used for alleviating the
simulation and behavior analysis of systems populated by human
individuals with diverse strategies.
Agents and People
I
am interested in intelligent user interfaces, both for adversarial and
cooperative settings. Much of my work in this area deals with
developing methods for selective information disclosure, to be used by
autonomous agents when interacting with people in order to influence
them. This way, agents and platforms can improve the benefit of people
or agents that use their services, to attract more customers, or
alternatively to benefit from their users' overconsumption of their
services.
Economic Search
I
use economic search, which has many real-life applications, as a
modeling framework and a test-bed for my research. Economic
search problems involve both uncertainty and costs of information
gain. These models represent a wide variety of activities,
including many in which agents are likely to participate (e.g., a
consumer searching for a product, a saver searching for an investment),
mostly due to their recurring nature. In an economic search, an agent
needs to choose among several opportunities, each of which has an
associated distribution of gains; in order to obtain the actual gain
from this distribution, the agent incurs a cost. In such problems,
agents need to take into consideration the trade-off between the cost
of further exploration and the additional benefit of knowing more
values. Unlike typical AI problems of exploration---exploitation
tradeoffs (e.g., multi-armed bandit problems), in economic search, the
exploration yields a specific value (rather than simply narrowing the
distribution) and the end goal is the choice of an opportunity (rather
than of an entity with which to continue interacting). It is also
distinguishable in that it has a closed-form solution.
Over
the years I have extended theories of optimal economic search to the
case of multi-agent cooperative search and investigated parallel
interaction aspects in distributed matching.