Project UmU1: Socially Intelligent Systems for Human-Agent Collaboration

Project UmU2: Self-driving transparent analytics

Project UmU3: Long time no see – Communicating with Autonomous Systems

Link To Application Portal:

Link To local pages:

Application deadline: June 15 2017

Project UmU1: Socially Intelligent Systems for Human-Agent Collaboration

Supervisor: Helena Lindgren, Dept of Computing Science, Umeå University

Contact info: Phone +46 70 3657463, Mail: helena.lindgren@umu.se

Short Project description: The aim is to develop socially intelligent software agents for human-agent collaboration. The fundamental challenge lies in how intelligent autonomous agents may collaborate with humans in decision making tasks to achieve goals, and making prioritizations among potentially conflicting goals, needs, motivations, preferences and choices of actions, e.g. in medical situations. To provide socially intelligent systems that humans can trust enough to collaborate with, algorithms for explaining automated learning, reasoning, and values of arguments and decision outcomes will be developed. Artificial intelligence-based methods for user modelling, adaptation and acting in a socially acceptable way tailored to a situation will also be developed, partly by formalising theories about human behaviour.

Research group: http://www.cs.umu.se/forskning/forskargrupper/uikm/

Project UmU2: Self-driving transparent analytics

Supervisor: Martin Rosvall, Integrated Science Lab, Dept. of Physics, Umeå University

Contact information: +46702391973, martin.rosvall@umu.se , www.mapequation.org

Short Project description:  Many autonomous system rely on state-of-the-art machine learning algorithms. The best-performing algorithms are also the most opaque. This transparency problemthe fact that we do not understand how they workcauses a great challenge for verification. That is, for an autonomous system that is supposed to work in dynamic conditions, we only know that it works in tested conditions. And as the autonomous system acquires new knowledge, how do we know that the new machine still handles previously tested conditions? This challenge calls for more transparent machine learning algorithms that allow us to look under the hood and understand the machinery. This project seeks to integrate transparent but domain-specific inference techniques from network science with versatile but opaque machine learning algorithms and enable self-driving transparent analytics. Join interdisciplinary Integrated Science Lab with great opportunities.

Project UmU3: Long time no see – Communicating with Autonomous Systems

Supervisor: Assoc. Prof. Kai-Florian Richter, Department of Computing Science, Umeå University

Contact info: Phone: +46 90 786 68 31 , Mail: kai-florian.richter@umu.se

Short project description: Autonomous systems, such as self-driving cars or robots in household or healthcare settings, are on the verge of becoming integral parts of our everyday lives. While autonomous, the systems are not self-sufficient. Eventually they need to interact with humans, which particularly holds for the ‘consumer’ systems mentioned above. For them interaction is particularly challenging since users’ knowledge about and experience with them is varying and possibly unknown. This PhD project will explore interaction between human and autonomous system. It will focus on how to establish and communicate meaningful spatial and temporal references, given that the last interaction may have been minutes or even hours ago. The aim is to develop methods for both understanding and producing such references.