Prof. Artinger at one of the leading machine learning conferences

Prof. Florian Artinger, who teaches and conducts research in Data Science and Behavioral Science at Berlin International University of Applied Sciences, was invited to present his research at the annual conference of The Institute for Operations Research and the Management Sciences (INFORMS) in Phoenix, USA. This conference is one of the largest and most prestigious worldwide, focusing on AI, analytics, machine learning, data science, and operations. Topics covered range from saving lives and money to solving problems more generally with the help of data and machine learning.

In a joint paper with three colleagues, Florian Artinger explores the labor supply decisions of taxi drivers. Taxi drivers provide an ideal lens through which to study how individuals supply labor, given that they typically have the freedom to start and stop working as they choose. In principle, one might expect drivers to offer extra hours of work when the demand for taxis is higher, and consequently, the hourly wage is also higher. However, classic studies in Behavioral Economics have revealed that taxi drivers often operate based on an income target per day, such as achieving $100. Once they reach this target, they cease working for the day. This implies that on busy days, drivers may finish early, while on days with lower demand for taxis, they need to work longer to reach their income target. This suggests biased and seemingly irrational behavior.

Using machine learning and a large dataset, Florian Artinger and colleagues investigated how well drivers can anticipate the next hour's hourly wage. Only if one can predict this would using an income target be irrational. What the scientists found is that predicting the next hour's wage with sufficient precision is actually impossible, even with very sophisticated algorithms and much more data than any driver would be able to handle. Instead, the taxi market is highly unpredictable in terms of the next hour's wage of a driver. In response, drivers have developed a set of simple adaptive strategies, or heuristics, that work well in this context. One such strategy is setting an income target given the unpredictability of hourly wage.

Overall, this suggests that it is important to understand under what conditions prediction works and when not. Given that this is not possible, the challenge is to develop clever strategies that still allow individuals to pursue their goals.

If you are interested in studying Data Science & Business, go to the online contact form or contact us through email!