Skip to content
Programs : Brochure
This page is the brochure for your selected program. You can view the provided information for this program on this page and click on the available buttons for additional options.
IIP-Max Planck Institute for Ornithology, Department of Collective Behavior
Konstanz, Germany (Outgoing Program)
Program Terms:
Program Terms: Summer
Layout wrapper table for buttons
Layout table containing buttons
Request Advising
Layout table containing buttons
Partner Institution/Organization Homepage: Click to visit
Program Dates & Deadlines: Click here to view
Restrictions: Princeton applicants only
Fact Sheet: - unrelated header
Fact Sheet:
#i18n(14)# parameter/value output
Dept Offering Program: International Internship Program (IIP) Program Type: Internship
Language Prerequisite: No Program Features: Academic Study, Field Work, Lab Based Work, Research
Degree Level: 1st year u/g students, 2nd year u/g students, 3rd year u/g students Time Away: Summer
Housing options: Student Responsibilty with support from IIP and/or Host Organization Program Group: International Internship Program
Program Description:
Program Description:
Max Planck Institute for Ornithology, Department of Collective Behavior
The goal of the Max Planck Society for the Advancement of Science is to support excellent fundamental research in the natural, life, and social sciences, as well as arts and humanities. This goal is achieved in more than eighty Max Planck Institutes, each of which focuses on a single area of research. The Institutes of the Max Planck Society are independent and autonomous in the selection and conduct of their research pursuits. Each institute has its own, internally managed, budget, which is supplemented by third-party funds through competitive research grants and collaborations. The Max Planck Institute for Ornithology has four departments: Department of Behavioral Neurobiology, Department of Behavioral Ecology and Evolutionary Genetics, Department of Migration and Immuno-Ecology and Department of Collective Behavior. The Max Planck Department of Collective Behavior consists of three labs which work on a wide range of organisms in both the laboratory and field, including fish, insects, arachnids, mammals and birds. Their department is a highly interdisciplinary environment with a closely integrated experimental and theoretical research program to understand the fundamental principles that underlie collective behavior across levels of biological organization. The systems used are both observable and readily able to be manipulated, and are ideal subjects with which to develop and test mathematical models that predict dynamic group-level properties from the behavior of smaller components. By integrating research at all levels of organization – from the neuro-biological mechanisms of social interaction all the way to movement ecology of social groups of large vertebrates – their research provides unrivaled opportunities to quantify the behavior of individual components within the context of the collective.

Intern Responsibilities: IIP interns will work on one or more of the following projects:
  • Project 1 - Collective Sensing:  In the natural world, individuals constantly face the challenge of acquiring, interpreting and responding to complex sensory information. Empirical evidence suggests that animals deal with this challenge by living in groups and processing information collectively. While these collective properties are manifested at the group level, they are an outcome of decisions made by individuals. In general, it is unclear how selection on behavioral rules adopted by individuals leads to evolution of group level properties such as collective information processing and distributed sensing. Previous models in collective behavior have been successful in explaining experimental data in a range of contexts and species, including leadership and consensus decision-making in fish and baboons. Even though these models work well predicting group behavior in moderate to large groups, they fail to reliably reproduce behavior in small groups and solitary individuals. This limitation constrains our ability to truly examine benefits of collective information processing and distributed sensing because it prevents comparison across various group sizes.

    The IIP intern will be involved in analyzing movement data from lab based recordings of fish in isolation and in varying group sizes and building on top of existing schooling models to account for individual behavior.  This more realistic models will then enable comparison across scales from individuals to collectives.
  • Project 2 - Neural Network Based Collective Decision-Making: While there are many advantages to living in a group it is also often advantageous to cheat on one’s responsibilities in the group.  This free rider problem can lead either to instability in groups, or, to groups never forming in the first place.  The team builds simulations that examine the conditions from which stable collective hunting and collective prey evasion emerge from naive random walk-like behavior.  Their simulations are inspired by work from DeepMind on deep reinforcement learning. 

    IIP interns would create learning based simulations, either starting from scratch or building off existing simulations in the lab, that examine how the abilities of individuals and the condition of the environment in which they exist affect the emergence and type of collective behavior in predators and prey.  How, for instance, does changing individuals’ range of possible movement actions affect group level behavior?  How does adding visual occlusions in the environment (like boulders in nature for instance) affect the learned hunting strategies of predators and how do prey respond? 
  • Project 3 - Quantitative approaches to the study of animal behavior: When individuals use socially acquired information during search, they are trying to simultaneously reduce their uncertainty about the environment and avoid personal costs by exploiting the efforts of others. However, relying on social information is not always adaptive. While social information may be less expensive to acquire, it is potentially unreliable. In contrast, personal information is more reliable but risky and effortful to obtain. Under conditions of limited time and cognitive resources, how do individuals navigate this trade-off and decide which information to use? To investigate these ideas I run a series of experiments that track human eye movement while humans solve visual search tasks.  I hope to understand 1) when and how searchers use social information during complex, difficult search tasks and 2) the effects of social information on search behaviors and performance. 

    IIP interns would work to help analyze the results of these experiments exploring how humans weigh personal and partner information while solving these problems.  How, for instance, does the cognitive difficulty of a task affect how reliant an individual is on those around it?  How do people respond to varying levels of noise in their own information vs. the information around them? If the intern is willing to be trained at Princeton in the spring they can also run their own experiments.
  • Project 4 - Collective behavior in locust swarms: Understanding how organisms process sensory information in the brain to produce behavior is one of the most exciting scientific problems of the 21st century. More specifically, understanding the sensory and behavioral mechanisms that animals use to successfully migrate long distances is one of the great scientific challenges of our time. Movement in migrating swarms of locusts is driven by cannibalistic interactions where individual movement decisions are made based on the threat of being cannibalized from behind and the motivation to cannibalize others ahead. The behavior of individuals within these marching bands is the result of visual and physical contact between individuals. The lab studies questions related to how sensory information and individual behavior influence the movement dynamics of group migration. Marching behavior in juvenile desert locusts is used as a model system to address two broad questions: 1) How do sensory information networks drive individual decision-making and group-level movement dynamics in migrating animal groups? 2) How do individual differences in behavioral state and group composition influence movement dynamics in migrating animal groups? The team conducts experiments in behavioral arenas that are filmed using multiple synchronized high-resolution 4K video cameras. Using computer vision techniques we measure the movement of individuals while also maintaining individual identities with 2-D bar code tags (similar to QR codes) attached to individuals. The visual fields of individuals are calculated using ray casting algorithms like those used in video game engines. By measuring visual and physical interactions between individuals we can infer the underlying social networks that drive both individual and group-level movement.

    IIP interns will help design and conduct experiments and analyze these data using unsupervised machine learning methods to classify behavior of individuals and describe changes in behavioral state across time and context.
  • Project 5 - The dynamics of group hunting and collective evasion: One benefit of sociality in prey animals is collective predator detection.  For collective detection to occur, information regarding the presence of predators must be transferred from knowledgeable individuals (detectors) to naive individuals (non-detectors). For this project, the lab will use drone-mounted cameras to capture aerial videos of ungulate (hoofed animal) groups in Kenya to study individual and group level vigilance patterns. To observe the process of information transfer, we will present model predators to groups and record their reactions. From these videos, the lab will extract continuous movement and behavioral (head up/head down) data for every member of the group, and calculate distances between individuals. Because of the complex background of these aerial videos, the data can only be extracted from the videos with deep learning based object detection and recognition algorithms. Deep learning models developed by Facebook Artificial Intelligence Research (FAIR) are particularly promising for this project ( Using the data extracted from these videos we can computationally reconstruct visual fields of individuals and, along with 3D habitat models, explicitly consider the information available to each individual and investigate how this information affects individual behavioral decisions.  Other people will carry out the actual filming, but the IIP intern can help with every other aspect of this project spending the most time working on what most interests them.
  • Project 6 - Revealing the structure of sensory interaction networks in animal groups: The predominant paradigm in the study of animal collectives has been to consider individuals as self-propelled particles which interact via social forces such as local repulsion, and longer-range attraction. This approach fails to consider key aspects of biology for many group-living species such as identity, social status, relatedness and informational status. The social relationships within these groups can change the interactions among individuals and have strong effects on the function of animal groups, yet understanding of social hierarchy in the context of collective behavior is limited.
    In this project, the team explores collective decision-making in an organism that forms stable, highly coordinated and socially stratified groups – the damselfish, Dascyllus marginatus. This is a tropical marine species that forms stable size-based social hierarchies of unrelated individuals in close association with branching coral species. I employ multi-camera imaging technology in order to track simultaneously the motion and behavior of each member of D. marginatus groups, in three dimensions, in the field (Red Sea, Israel). Using various stimuli, the team will explore the relationship between individual- and group-behavior in three ecologically relevant contexts: (A) Detection of potential threats (B) Individual and socially-mediated escape manoeuvres and (C) Decision-making regarding emergence. Acquiring data from the videos constitutes a challenging computer vision problem. Interns can be involved in developing programs to automate data acquisition, as well as developing tools to visualize the data in three dimensions.
Qualifications: IIP candidates with interests in mathematics, physics, electrical engineering, computer science, operations research and financial engineering, mechanical and aerospace engineering, economics or ecology and evolutionary biology or related fields are encouraged to apply. Technical programming skills in python, C++ and an interest in machine learning and computer vision would be an asset.

Dates / Deadlines: - unrelated header
Dates / Deadlines:
Tabular data for Dates / Deadlines:
Term Year App Deadline Decision Date Start Date End Date
Summer 2019 04/01/2019 04/15/2019 TBA TBA
Layout wrapper table for buttons
Layout table containing buttons