Representing Users in a Negotiation (RUN): An Autonomous Negotiator Under Preference Uncertainty


Computers that negotiate on our behalf will soon become part of our personal lives. For instance, with the world-wide deployment of the smart electrical grid, devices in our living room will negotiate complex energy contracts automatically with other households. Another example is the rise of countless interconnected devices, known as the Internet of Things, which will negotiate the usage of our sensitive data.

Such systems need to make trade-offs autonomously between various (and beforehand unknown) personal interests of each individual user, such as cost effectiveness, privacy, sustainability, and functionality. For such complex negotiation spaces, it is impossible to fully elicit the user’s preferences. However, state-of-the-art automated negotiators do not support user preference uncertainty.

I aim to address this problem by enabling negotiation systems to learn about the user on-the-fly. This immediately opens up two important, and so far unaddressed research challenges:
- Developing negotiation strategies that are effective, even when it is uncertain how desirable certain proposals are;
- Identifying the most valuable information to elicit from the user in order to improve the negotiator’s performance.

My core contribution is an innovative framework that takes on these challenges by providing:
1) a stochastic representation of the user’s preferences in negotiation, based on probabilistic preference models;
2) new negotiation strategies based on recent methods from search theory and machine learning that can reason with incomplete and uncertain information through tractable performance-based metrics;
3) concurrent preference elicitation methods that can balance informational value against user bother.

I will test my framework by organizing an international automated negotiation competition, and I will apply my solutions in cooperation with smart grid users and industrial Internet of Things partners. My results will deliver an autonomous, multi-purpose negotiator that can represent users faithfully, thereby paving the way for a next generation of personalized negotiation systems.


Project number


Main applicant

Dr. T. Baarslag

Affiliated with

Centrum Wiskunde & Informatica


01/01/2018 to 31/12/2020