Almende and Erasmus University Institute of Management (ERIM) offer the following PhD research project:

 

An Agent-based Approach to Human Preference Modeling in Complex Dynamic Networks 

(download the full description here: Agents_Complex_Networks_Preferences_ERIM_Almende_PhD_Proposal_2009.pdf)

 

Summary of the research problem and research question
The goal of this project is to create highly flexible and robust preference modeling techniques for multi-agent systems [Wooldridge and Jennings, 1995] that are based on principles of synergetics [Haken 1983, Haken 1993] and complex network analysis [Albert and Barabási 2002], and foster collaboration between humans and ICT systems in order to sustain or increase operational management performance levels [Beer 1959] within different business domains and across several dynamic and evolutionary scales. In addition, the project evaluates the developed multi-scale synergetics techniques, capable of handling the sparseness of data problem in building preference models, within recommender systems [Grcar, Mladenic, Fortuna and Grobelnik 2006] serving those domains.

Research question:
Which scientific and technological approach and related preference modeling techniques can dynamic and evolving networks of multi-agent systems, humans, and other ICT systems effectively employ to handle operational management issues, like last minute changes in logistics or the introduction of some new technology?

 

Description of the research project (min. 1000 words)
The problem of modeling incomplete, context-sensitive preferences that evolve over time is eminent in organizations and business settings. Business networks are besides highly dynamic also continuously evolving. Companies (departments) across an enterprise (institution) change and so do their customers (clients). In addition, their employees come and go; change position; gain/loose expertise, skills and competencies and alter relationships. Furthermore, ICT systems interact with, support and/or replace the employees more and more frequently and in an ever more diverse way in (helping) carrying out normal or even managerial tasks (e.g. logistic or medical workflow management systems). Last but not least, those ICT systems become more and more pervasive and ubiquitous, dependable and intricate (e.g. multi-project schedulers or Tom-Tom). Therewith companies, enterprises/institutions, customers and their employees/clients are facing besides these changes also complex managerial problems at different scales caused by uncertainty and incompleteness of information about the actual state of business operations; incidents may propagate and lead to a cascade of other incidents when not properly resolved. In such situations both actors and ICT systems of an organization need to be aligned and to ultimately cooperate to achieve shared or individual business goals, i.e. sustaining or even improving operational management performance. In such operational management situation it is clear that building and capitalizing on related preference models employed by the various network entities is a must.
The question how ‘living’ preferences models that are employed by multi-agent systems (but also the other entities) and that are adaptive upon changing network situations and circumstances both on short- and longer term can be generated boils down to:
1.    How can a multi-agent system be designed that interacts and co-evolves (by simulation) with existing or future (business) networks?
2.    Which preference modeling tools and techniques capitalize on the synergetics  and complex network analysis of dynamic and evolving networks? For example, how are business networks, including networks of multi-agent systems, described within a synergetics framework?
3.    What elicitation processes are available to the agents in the system and what data is typically available to model their own and the other network entities’ preferences? What techniques can be called upon by the agents to process these data in order to arrive at (implicit or explicit) feedback-based preference models with respect to the operational management of (business) networks, that are capable to tackle the sparseness of data problem?

Prof. Dr. Ir. Eric van Heck  - Professor of Information Management and Markets - 8 hours/month
Dr. Wolf Ketter – Assistant Professor for Business Intelligence, Networks and Markets - 12 hours/month

 

For more information contact Dr. Alfons Salden at +31 10 4049444 or out the ERIN website.