Computational methods for facilitating tradeoffs in team formation
OA Version
Citation
Abstract
The rise of online labor markets and human-resource management platforms has ignited a lot of research on team formation, where experts acquiring different skills form teams to complete tasks. The core idea in team formation has been that a team successfully completes a task if it acquires a superset of the skills required by the task and optimizes some other team objective. Even though this is a rather natural formulation, in its general form it lacks the ability to capture some practical concerns emerging in real-world scenarios that require compromising the team's competence to improve other important team characteristics. Towards addressing this shortcoming of the current team-formation models, in this thesis we introduce novel problem formulations and present efficient methods for their solutions.
This thesis comprises three main components. First, we address concerns raised from sociology studies indicating that team content presumes that team members have clear roles and mutually respect their teammates for the role they assume in the team. Therefore, the first part of the thesis formalizes this problem and provides methods for its solution that work well in practice and yield intuitive qualitative results.
Second, we consider settings where the entity posting a task might prefer trading off the number of task skills covered by the team in exchange for improving some other team objective. We provide this trade-off functionality for two team objectives; (i) the cost of hiring the team, and (ii) the workload of the team members. Furthermore, different computational challenges arise when considering each of the two objectives, and therefore we tackle each problem separately.
Finally, the third part of this thesis focuses on predicting team performance given some metric. More precisely, we assume the online team sports game setting where the designer's goal is to create exciting game matches. One intuitive way to do so is by launching competitively balanced matches in which both teams are likely to win. Thus, in this thesis we discuss the importance of competitive balance in online games and provide fast and interpretable models that allow achieving such matches.
Description
License
Attribution-NonCommercial-NoDerivatives 4.0 International