English  Sprachen Icon  |  Gebärdensprache  |  Leichte Sprache  |  Kontakt

Aktuelles aus den Elitestudiengängen

Internship at the Potsdam Institute for Climate Impact Research COPAN (Coevolutionary Pathways)

Katharina Funk had the opportunity to do her internship from March 13th to April 23rd 2017 at the Potsdam Institute for Climate Impact Research (PIK) in Potsdam. She worked with the COPAN-Group, whose members are doing research on the mid- to long term coevolution of natural and socio-economic subsystems of the Earth system and its relationship to planetary boundaries.

Foto: Katharina Funk

Foto: Katharina Funk

During the first two weeks her main task was to do literature research. She read about behaviour experiments within the area of public good games, and wrote a report about distinguishing under which circumstances persons tend to cooperate (e.g. communication, punishment, dynamical networks). Additionally she created a table with relevant information for building a model of these experiments and the stylized facts of the paper.

They worked mainly with reinforcement learning and agent-based modelling; in this setting, agents (actors which could represent persons or nations) can use a certain natural environment in different ways (e.g. sustainable or non-sustainable ways). The natural environment reacts to these management practices in a certain way, which can be observed by the agents. Depending on these observations (e.g. harvesting rate, strategy of the neighbours...) agents can also change their strategies.

In order to create a simple model herself, she taught herself the basics of reinforcement learning and agent-based modelling. She chose one paper (Milinski et al., 2008) in which they performed a behavioural experiment related to climate politics: During each of the ten rounds every player could invest 0€, 2€ or 4€ in a "climate account". They had to reach a threshold of 120€ to keep all their savings, otherwise they would lose all savings (catastrophe) with a probability of 90% (50%, 10%). The question was how high the probability of a catastrophe had to be in order to encourage the player to invest in the climate account.

Katharina Funk implemented this experiment within a reinforcement learning environment and performed the same experiment with learning agents. She was able to reproduce the results of the study: if the probability for a catastrophe to happen was high (around 90%), the agents learned fast to contribute to a total of 120€. If the probability was low (50% or lower), the agents were not able to reach the threshold.

Personal experience

There is a good working atmosphere at PIK and there are lots of possibilities to learn, talk to people and listen to talks and presentations. PIK has a wonderful location on the Telegraphenberg, where it is quiet and green, together with other research institution. Katharina Funk worked in the COPAN-group with really nice people who were always happy to help her and answer her questions. She had flexible times and a very independent work. On the downside, she was not very well connected with my working groups as they worked within other buildings/places. Therefore she had to figure out a lot by herself and was working mainly alone. Nevertheless she always had the possibility to ask her supervisors or anyone else in the group and always got help. Her supervisors were Wolfram Barfuss and Jonathan Donges.

Text: Katharina Funk



Milinski, Manfred, et al. "The collective-risk social dilemma and the prevention of simulated dangerous climate change." Proceedings of the National Academy of Sciences 105.7 (2008): 2291-2294.

veröffentlicht am