Learning and Work hybride seminars, lente/zomer 2026
De hybride seminarreeks van de Universiteit Maastricht - SBE biedt een podium aan onderzoekers uit verschillende disciplines, met aandacht voor zowel academische als beleidsmatige vraagstukken.
De hybride seminarreeks vindt elke dinsdag plaats van 12.00 tot 13.00 uur, ter plaatse en ook via Microsoft Teams.
Organisatoren L&W seminars
Raymond Montizaan: r.montizaan@maastrichtuniversity.nl
Katarina Wessling: k.wessling@maastrichtuniversity.nl
Steffen Künn: s.kuenn@maastrichtuniversity.nl
-
Aankomend seminar
Wanneer: dinsdag 26 mei 2026
Tijd: 12.00u - 13.00u
Waar: TS53 - C-1.05Spreker: Bart de Koning (VU)
Titel: Overconfidence, Information, and Vacancy Recommender Systems
Abstract: We conduct a field experiment with 2,147 job seekers in New York State to study how overconfidence, labor market information, and vacancy recommender systems interact. The experiment takes place on an online job search platform we developed. A random sample of job seekers gets to see vacancy recommendations generated for them by a content-based recommender system, and a random subset of those also receives information about their competitiveness across different submarkets (i.e., job titles). We first show that overconfidence in the labor market is hard to capture as a one-dimensional object, as there is substantial variation in overconfidence within individuals across submarkets. The recommender system seems to inhibit learning, as those who get to see its recommendations (but no information) hold moderately less accurate beliefs than the control group and are more likely to be overconfident in at least one submarket in subsequent months. This lack of learning can be more than compensated for by providing information: those who receive information hold the most accurate beliefs of all groups, with remarkably little decay over time. The behavior of job seekers who receive recommendations without information causes the recommender system to generate recommendations for vacancies with more competitive job titles, suggesting the recommender system perpetuates behavior based on inaccurate beliefs. At the 6-month mark, we do not find any differences in job finding rates or characteristics of jobs found across groups, though data collection will continue for 12 additional months. Results are preliminary and subject to change.Geplande seminars
09-06-2026: Hannah Klauber (PIK)
23-06-2026: ntb
07-07-2026: Amelie Schiprowski (Uni Bonn)