skip to main contentskip to main menuskip to footer Universität Bielefeld Play
  • Research Seminar (Management)

    © Universität Bielefeld

Research Seminar (Management)

In the Research Seminar (Management) talks are mainly held by outside speakers. BiGSEM students (Profile Management) are required to attend the seminar during the semester.

 

Wednesday, May 11 2022, 16:00-17:00 in room V10-122

Prof. Dr. Oliver Müller (Paderborn University)
Using Natural Language Processing to Forecast the Impact of AI on the Labor Market

Abstract:
Artificial intelligence (AI), broadly understood as the capability of machines to perform cognitive tasks at a level that is comparable or even superior to humans, is a rapidly advancing general-purpose technology that holds the potential to reshape the nature of work. In recent years, so-called narrow AI systems outperformed humans in a growing range of specialized tasks requiring human-like cognitive capabilities like vision or natural language processing. However, we are still far from artificial general intelligence: Machines cannot do the full range of tasks that humans can do. This raises the question which tasks and, in turn, jobs will be most affected by the ongoing diffusion of AI. We are proposing a new measurement instrument to estimate the likelihood that tasks and jobs can be automated through AI in the near future. In contrast to previous measurement approaches, which require the description of jobs in terms of standardized numerical variables, our approach leverages state-of-the-art natural language processing and machine learning techniques to score textual job descriptions from online career portals according to their automation potential.

About the Speaker:
Oliver Müller is a Professor of Management Information Systems and Data Analytics at Paderborn University. He holds a BSc and MSc in Information Systems and a Ph.D. in Business Economics from the University of Münster's School of Business and Economics. In his research Oliver studies how organizations create value with (big) data and analytics; for example, by enhancing judgment and decision making, supporting knowledge management, or automating business processes. His research has been published in the Journal of Management Information Systems, Journal of the Association of Information Systems, European Journal of Information Systems, and various others.

Note: Although the talk will be in-person, you will also be able to attend virtually using the following Zoom information:
https://uni-bielefeld.zoom.us/j/64080294957?pwd=b0NxOG03YVI1WnZWcXVnTDBBS1hadz09
Meeting-ID: 640 8029 4957
Passwort: 732835 


Wednesday, June 1, 2022, 16:00-17:00  Zoom

Prof. Dr. Warren B. Powell (Princeton University)
Introduction to Sequential Decision Analytics: A unified framework for decisions under uncertainty

Zoom Link:
https://uni-bielefeld.zoom.us/j/62105454430?pwd=TXJTZEtKa1hqQkhvL2RHMTJSVWk3Zz09
Meeting-ID: 621 0545 4430
Passwort: 854083

Abstract:
Sequential decisions are an almost universal problem class, spanning dynamic resource allocation problems, control problems, stopping/buy/sell problems, active learning problems, as well as two-agent games and multiagent problems.  Application settings span engineering, the sciences, transportation, health services, medical decision making, energy, e-commerce and finance.  A rich problem class involves systems that must actively learn about the environment.  We also consider hybrid resource allocation and learning problems such as those that arise in disease mitigation, as well as complex multiagent supply chains.
These problems have been addressed in the academic literature using a variety of modeling and algorithmic frameworks, including (but not limited to) dynamic programming, stochastic programming, stochastic control, simulation optimization, stochastic search, approximate dynamic programming, reinforcement learning, model predictive control, and even multiarmed bandit problems. Particularly frustrating is that these communities do not use a common modeling framework.
We are going to introduce a universal modeling framework that can be used for any sequential decision problem in the presence of different sources of uncertainty.  Our approach is to define the problem first, which consists of sequential decision problems (decision, information, decision, information, …) where the challenge is to find policies for making decisions.  We claim that there are four (meta)classes of policies that are the foundation of any solution approach that has ever been proposed for a sequential problem.  Using a simple energy storage problem, we show that any of the four classes of policies might work best depending on the data, and hybrids can be created that combine two or more classes.  All of these ideas will be illustrated with applications drawn from different application settings.

About the Speaker:
Warren B. Powell is Professor Emeritus at Princeton University, where he taught for 39 years, and is currently the Chief Analytics Officer at Optimal Dynamics.   He was the founder and director of CASTLE Labs, which focused on stochastic optimization with applications to energy systems, transportation, health, e-commerce, and the laboratory sciences (see www.castlelab.princeton.edu), supported by over $50 million in funding from government and industry.  He has pioneered the use of approximate dynamic programming for high-dimensional applications, and the knowledge gradient for active learning problems, work that was the foundation for his books Approximate Dynamic Programming and (with Ilya Ryzhov) Optimal Learning. He developed a universal framework for sequential decision problems which is the basis for his latest book with Wiley: Reinforcement Learning and Stochastic Optimization: A unified framework for sequential decisions. He published over 250 papers and produced 70 graduate students and post-docs.  He is the 2021 recipient of the Robert Herman Lifetime Achievement Award from the Society for Transportation Science and Logistics, a fellow of Informs, and the recipient of numerous other awards.


!! Cancelled !!

Wednesday, June 8, 2022, 17:00-18:00  Zoom

Dr. Philipp Loick (Amazon)
Timing optimization for Amazon Middle Mile

Abstract:
Operating a logistics network of Amazon's scale gives rise to many fascinating mathematical challenges. In this talk, we will discuss Amazon's middle mile network and Amazon's approach to efficiently delivers millions of packages to customers every day. Specifically, we will discuss how timing optimization is performed in the Amazon network to allow customers to order packages as late as possible with delivery on the next day.

About the Speaker:
Philipp Loick works as an Applied Scientist in Amazon's European research science team on routing and timing optimization in Amazon's middle mile network. Prior to joining Amazon, he completed a PhD in mathematics on statistical inference and the theory of machine learning at Goethe University Frankfurt. He holds a Master in Operations Research from the London School of Economics and a Master in Computer Science from Georgia Institute of Technology.


Monday, June 20, 2022, 16:00-17:00 in room V10-122

Prof. Dr. Taïeb Mellouli, Thomas Stoeck (Martin Luther University Halle-Wittenberg)
Data-Driven Knowledge-Based Decision Support Systems involving both complex activity structures and complex systems

Note: Although the talk will be in-person, you will also be able to attend virtually using the following Zoom information:
https://uni-bielefeld.zoom.us/j/68824071093?pwd=amt2aklCS3BKcGRlYlJ2R2d0R1h2UT09

Meeting-ID: 688 2407 1093
Passwort: 016002

Abstract of the talk:
While the main focus of the talk will be the presentation of new business decision analytics, modeling and solution approaches for a “data-driven patient flow management in hospitals” (see abstract below of the respective research paper), the talk will give general reflections on how to tackle decision support problems in complex systems by approaches combining techniques from AI and OR. Adequately handling such business situations requires new intricate decision problems’ modeling and combined AI/OR solution approaches where dimensions of complexities are crossed. In the actual case, the complex activity structures are patient pathways and flows that are to be managed and optimized within a complex organization – the hospital – being a composition of interdependent wards/departments. Ingredients of our solution approach includes a disassociation of clinical and operational aspects of hospital data, separately mining hospital-independent clinical/medical and hospital-dependent operational knowledge, then adequately apply this knowledge to support physicians and dispatchers of the hospital to deal with situational patient-individual and hospital-sate dependent decisions. The developed policies and scenarios are evaluated by data science and hospital-wide simulation techniques.

As you well know from Prof. Römer, a business adequate and full flagged “decision analytics” requires a combination of data science and prescriptive analytics techniques of OR. As we showed in a former collaborative paper, this elevates to an AI/OR synergy that can be intricate, requiring not only process mining (advanced data science and AI), but also involving hierarchical planning (advanced prescriptive analytics/OR). Both AI and OR sides considers business relevant aspects – only as much as necessary – of the complex activity structures, here patient pathways. We show in this talk, that these business relevant aspects and the decision problems at operations (not planning phase) requires another, likewise innovative handling of these complex structures – and this at all levels: problem analysis, design of decision support, and solution approaches. A main general reflection inside these approaches of our actual paper is: Business adequate and full flagged “decision analytics” also needs “DATA-DRIVEN KNOWLEDGE-BASED Decision Support Systems” – that is, the integration of human or automated reasoning with learned knowledge in advanced decision support systems – for highly qualified experts.

Paper underlying the talk:
Thomas Stoeck, Taïeb Mellouli, Malek Masmoudi: Data-Driven Patient Flow Management via Mining Clinical and Operational Knowledge and Evaluation by Hospital-Wide Simulation

Abstract of the Paper:
Economic pressure in the healthcare market and restrictive governmental regulations force hospitals to ensure their competitiveness by improving their performance, which directly depends on increasing their occupancy rate. This can be achieved by treating more patients and managing higher levels of patient flow across the hospital through clever patient flow management and without additional investment in department or bed capacities and increased medical staff. Attempts to attain the highest occupancy rates will inevitably involve strain situations.
To manage and balance sporadic demand fluctuations between departments, we propose new flexible policies for smoothing the patient flow based on different types of patient admissions, redirections and relocations. As they depend on determining suitable alternative departments, we design a data-mining approach generating alternatives based on medical needs and service quality provided by departments. The resulting data-driven situational decision-making tool for hospitals uses standardized data to ensure wider applicability. In addition to this decision-making tool being used for local situation-dependent decisions, the global effects were evaluated using a generic object-oriented, hospital-wide simulation model.
The current practice of the hospital shows that for 9 % of patients, an individual non-documented ad hoc decision has to be made to ensure admission. Our policies, creating a structure for these decisions, allow for the admittance of 97% of patients without decreasing the overall quality. The evaluation shows that the rate of patients experiencing high-quality service increases from 87% to 92%. These results confirm that operational policies using redirections and relocations can overcome short-term strain situations without decreasing overall treatment quality.

About the Speaker:
Prof. Mellouli studied computer science and mathematics at the University of Stuttgart, then got his doctoral and post-doctoral degree (habilitation) from the University of Paderborn. After a practice phase at TUIfly (Hapag-Lloyd), where he initiated the development of the currently productive Crew-Optimizer-System, he switched to academia as a professor for “Business Information Systems and Operations Research” at the Martin-Luther University Halle-Wittenberg and still maintained the collaboration with selected practice partners seeking support for their complexly structured decision problems.
Main research interest involves business process analysis and design of information systems enriched by optimization and simulation modeling (operations research techniques) for business applications. Prof. Mellouli maintains his former interest on automated reasoning and knowledge-based systems (from doctoral thesis) and, by investigating AI/OR synergies, reveals insights leading to the design of innovations in decision support systems augmented by knowledge reasoning in complex business environments.  

15.12.2021, 15:00-16:00, zoom

Prof. Dr. André Martins (Universidade de São Paulo)
How does my neighbour think? Mental models and opinion dynamics update rules

https://uni-bielefeld.zoom.us/j/98801593901?pwd=VGpvcHBJMFdXRk4xRE5GaEJvdHBMQT09
Meeting-ID: 988 0159 3901
Passcode: 144036

Abstract:
Most models in opinion dynamics are created from simple ad-hoc rules. Different models match distinct applications better and very little work exists on how they relate to each other. In this talk, I will show how we can define general models from a framework for how agents communicate and how we can obtain update rules from agent mental models and simple uses of the Bayes theorem. That allows us to easily include specific aspects of any applications we might be interested in as well. And we can also understand, up to some underdeterminacy, which mental models seem to correspond to each of the update rules we find in the literature.

 

12.04.2021 von 10:00-12:00 Uhr, online

Vortragende: Prof. Dr. Matthias Baum, Universität Bayreuth
Endogeneity: Easy to get but hard to please – The Quest of writing an endogeneity paper and why it matters (not only for entrepreneurship)

Meeting-ID: 921 1581 3533
Passwort: 614943 


12.05.2021 von 15:00-17:00 Uhr, online

Vortragende: Prof´in Dr. Sabine T. Köszegi, TU Wien
I´d blush if I could: Eine Genderperspektive auf KI Technologien und Digitalisierung

Meeting-ID: 930 0813 8052
Passwort: 406284


- Canceled - 22.06.2021 von 14:00-15:30 Uhr, online

Vortragende: Prof Dr. Volker Breithecker und Dr. Michèle Kuschel, Universität Duisburg-Essen
Master Innopreneurship–Die Realisierung eines innovativen Studiengangkonzepts

Meeting-ID: 914 8373 9841
Passwort: 611206


15.07.2021 von 11:00-12:00 Uhr, online

Vortragende: Prof´in Dr. Brigitte Halbfas, Universität Wuppertal
Thema des Vortrages: Chancengleichheit in der Unternehmensgründung? Einblicke in die Forschung

Meeting-ID: 987 3346 0561
Passwort: 745087

 


back to top