Group 4Fill 1Fill 1Page 1inGroup 11outlookpaperPDFStarFill 1Group 6zipfacebookinstagramlinkedinsocial/twitter@2xCombined ShapeCombined ShapeCombined ShapelunchCombined ShapeCombined ShapeCombined ShapeFill 1Artboardarrowicons/closeFill 1 Copy 2burgerArtboard

Lars Holmberg is a PhD student at the Computer Science department at Malmö University and affiliated with K2: Sweden’s national center for research and education on public transport. Lars holds a master degree in Computer science from Lund Institute of Technology and currently specialises in interactive machine learning. Before current PhD studies, Lars was responsible for the bachelor program in interaction design at Malmö University.


  • June 10: Easier said than done: doing right by the customer

    Interactive Machine Learning for Commuters: Achieving Personalised Travel Planners through Machine Teaching

    In Machine Teaching (MT) a person is acting as teacher while the Machine Learning (ML) agent acts as a student. In this work, a commuter app is trained using MT to make predictions on upcoming journeys. The predictions are based on the commuter’s location, time of day, weekday and activity. The ML agent presents departure times for the most probable journey when the smartphone app starts and an MT session can be initiated to address the situations were no training data exists or when the commuters travel patterns has changed. Designing the interaction for these sessions is a crucial factor in making the teaching understandable. In the teaching sessions the user can alter both what the agents already knows and teach the agent new travel patterns. This work concentrates on identifying design considerations for MT sessions given our design experience and evaluation results.