Human errors remain the main source of incidents. They can lead to fatalities, traffic accidents, or product defects and cause high economic and social cost. While some errors can still be corrected if they are detected in time, many human errors cause high costs as soon as they occur or are even irreversible. In these cases, it is very important to recognize human errors before they occur. The goal of this project is therefore to develop methods based on artificial intelligence that forecast human errors from video data. We focus on erroneous and unintentional human actions and we aim to support humans to avoid them. In order to achieve this goal, we aim to solve three tasks jointly. We aim to develop methods that forecast human motion and intention with a very low latency such that unintentional actions can be recognized before they occur. Without the capability to interfere, however, even the best forecasting model does not prevent human errors. We therefore aim to develop a model that generates an auditory feedback if an error is forecast. The feedback, however, should not only warn humans, but also guide them such that they can successfully complete their intended action. Finally, we aim to model how humans will react to the feedback. We thus aim to develop a model that forecasts the motion of humans and objects they interact with, that recognizes human errors before they occur, and that guides the human motion via auditory feedback in order to prevent errors. The model should automatically decide if and what auditory feedback is generated by reasoning how the feedback will affect the motion of persons that are close-by. While we aim to showcase that the developed technology is able to prevent errors before they occur, this technology has the potential to drastically reduce the social and economic costs caused by human errors in the long term.

Medieval Islamic Archaeology in Southern Greater Syria is relevant to the sustainable agricultural intensification discourse due the implementation of state intensive land use for profit that affected the lives of peasants and the ecology of semi-arid regions. Medieval Islamic agricultural intensification practices included irrigation, fertilizers, overgrazing, deforestation and cash cropping, which increase agricultural yields. The project “Human Landscapes: agricultural intensification and peasant resilience in medieval Southern Greater Syria” (HumanLand), offers a deep-time perspective on these issues. According to surveys and historic resources, the later medieval Islamic eras were marked by the collapse of the Mamluk state, the decline of major agricultural centers, and a shift to a seasonal basis occupation partly due to climatic stressors of the end of the 14th and 15th centuries, as well as the lack of state resources and support. HumanLand will use new environmental and textual data for medieval land-use to investigate these three objectives: 1) To identify changes in agricultural intensification in relation to climatic, political and economic shifts that took place in the late 14th and much of the 15th centuries, 2) To investigate the use of sustainable strategies and impact of imperial regimes on medieval communities, and 3) To investigate the impact of political structures on the ecology in semi-arid regions. The project will use botanical micro remains (phytolith, starches and spherulites) and stable isotope data from crop remains to understand agricultural practices of six medieval communities of Jordan and Israel. This evidence will be evaluated with information on land-use in medieval texts. These methods combined, provide evidence for intensified agriculture making this research one of the few studies done, which combine the different approaches proposed to approach the big picture of medieval Islamic agricultural history.

An omnipresent but understudied environmental risk for our immune system is pollution by nano-sized plastics. Plastic particles have been detected in a wide variety of ecosystems and are speculated to enter and spread in the food web all the way to humans. Ingested nanoplastics can translocate from the gut to the lymph and circulatory systems and have the capacity to cross the blood-brain barrier in mammals. It has been recently shown that nanoplastics cause behavioural disorders in fish, and thus may also represent a risk for human health, in particular for brain function. However, the long-term bioavailability and toxicity of nanoplastics in the brain are unknown. Microglia as the main neuroimmune cells have not only a defence function required during inflammatory conditions, but constantly sense and response to environmental changes as part of their housekeeping functions that are essential for neuronal homeostasis. This places microglia at the interface between normal and abnormal brain development and function. In line with this, we have recently discovered that chronic microglial activation causes neurodegeneration. As highly phagocytic cells, microglia internalize nanoplastics reaching the brain. This process might in turn lead to their acute or chronic activation, thereby triggering neurological disorders. In NanoGlia, we will use rodent animal models to investigate behavioural as well as cellular and molecular changes in the brain that occur upon ingestion of nanoplastics. We will further determine nanoplastics-induced developmental reprogramming events in fetal microglia that may influence brain organogenesis and function. Understanding how nanoplastics triggers microglial activation during embryogenesis and postnatal stages and whether this immune activation leads to permanent changes in brain development and function will reveal ground-breaking mechanistic insights into the environmentally triggered pathogenesis of neurological disorders.

This proposal studies price negotiations in dynamic markets. The focus is on one of the most primitive economic problems: a seller and a buyer bargain over the price of a good. Their cost and values are private information and, if they do not reach an agreement today, they may continue bargaining tomorrow. Somewhat surprisingly, our knowledge about bargaining with two-sided asymmetric information is still quite limited. Intuitively, on the one hand, signaling forces induce the seller and the buyer to delay trade to obtain a higher share of the trade surplus. One the other hand, Coasian forces push prices down and make trade efficient. The balance between these two forces determines the efficiency of the market and how the trade surplus is shared between the seller and the buyer. I will provide a new systematic analysis of markets with asymmetric information. Using recent developments in the characterization of robust behavior and strategic stability, I will first analyze dynamic pricing by privately informed sellers in markets with independent values. Building on the understanding of the basic bilateral bargaining problem, I will then consider related problems. In particular, in the second subproject, I will consider interdependent values and, in the third subproject, I will consider a multilateral setting with multiple sellers. Markets studied in this project will include real estate markets, markets for durable goods, markets for intermediate goods, and financial markets. The results will provide guidance to assess the effect that currently debated policies regarding privacy or confidentiality have on social welfare or market efficiency.

The goal of the project is to automatically analyse human activities observed in videos. Any solution to this problem will allow the development of novel applications. It could be used to create short videos that summarize daily activities to support patients suffering from Alzheimer's disease. It could also be used for education, e.g., by providing a video analysis for a trainee in the hospital that shows if the tasks have been correctly executed. The analysis of complex activities in videos, however, is very challenging since activities vary in temporal duration between minutes and hours, involve interactions with several objects that change their appearance and shape, e.g., food during cooking, and are composed of many sub-activities, which can happen at the same time or in various orders. While the majority of recent works in action recognition focuses on developing better feature encoding techniques for classifying sub-activities in short video clips of a few seconds, this project moves forward and aims to develop a higher level representation of complex activities to overcome the limitations of current approaches. This includes the handling of large time variations and the ability to recognize and locate complex activities in videos. To this end, we aim to develop a unified model that provides detailed information about the activities and sub-activities in terms of time and spatial location, as well as involved pose motion, objects and their transformations. Another aspect of the project is to learn a representation from videos that is not tied to a specific source of videos or limited to a specific application. Instead we aim to learn a representation that is invariant to a perspective change, e.g., from a third-person perspective to an egocentric perspective, and can be applied to various modalities like videos or depth data without the need of collecting massive training data for all modalities. In other words, we aim to learn the essence of activities.