Hedda R. Schmidtke obtained her PhD from the University of Hamburg (Germany) in 2004 for work on granularity and level of detail mechanisms in artificial and natural cognitive systems. With a focus on the foundations of higher cognition and the interface between perception and reasoning, she worked on several application-oriented projects where this interface was of core concern, especially in the area of pervasive/ubiquitous computing (ubicomp). 

In ubicomp scenarios as well as in the brain, both distributed sensory data acquisition and (potentially distributed) logical reasoning about these data play a key role. During this time, she worked at research universities on four continents, including the Gwangju Institute of Science and Technology (South Korea), the Karlsruhe Institute of Technology (Germany), Carnegie Mellon University (USA), and the ICT Center of Excellence (Rwanda). Her last appointment was with the Department of Geography at the University of Oregon (USA). After a breakthrough on her long-term foundational research goal, she currently is a fellow at the Hanse-Wissenschaftskolleg Institute of Advanced Study (HWK). At HWK, she tests the next level of her cognitive architecture on a critical thought experiment from ethics, the Trolley Problem, combining causal reasoning and context-dependent levels of empathy.

Hedda Schmidtke

Research Interests:

With over 60 publications and 15 years of postdoctoral experience, Dr. Schmidtke has completed several externally funded research projects in different areas of practical, mostly industrial application. These range from reasoning conveyor belts in coal mining to intelligent leaflets for illiterate patients in Africa. The underlying common foundational research theme is the question how higher cognition, i.e., logical reasoning, can be founded upon perception, i.e., sensor value processing. This longstanding problem has recently regained public attention as key for the efforts towards explainable and ethical AI. In this area, Dr. Schmidtke recently achieved a research breakthrough using a novel cognitive architecture she developed and is now mostly interested in applying it.

The journey started from research on granularity in cognition. Human beings conceptualize differently depending on context: a road is a line, i.e., has negligible width, on a map, but a two-dimensional object, when we need to cross it. Depending on “how close our minds eye” is to an object, we use different conceptualizations. Granularity lets us focus on what is relevant. An effort to leverage this mechanism in reasoning included generating test data to evaluate the hypothesis. This led to the discovery that the test data, logical descriptions of spatial layouts, were not only correlated slightly with dimensions of granularity, as conjectured, but linearly equivalent replicas of the images that had been used to generate the description. The result implies that perception and language are not as different as we commonly think. 

The ramifications of this result, published in 2018 with several follow-up papers since then, are manifold. A range of predictions exist, and all tested so far could be confirmed. Among the most interesting applications is trustworthy small-scale AI.

Key Terms:

  • Cognitive Science, 
  • Artificial Intelligence (AI),
  • Ethical AI Systems, 
  • Symbol Grounding Problem, 
  • Meaning, 
  • Consciousness, 
  • Complexity Theory of AI