We are pleased to announce the following invited talks:
- Day 1:
1:00 PM EST: Gerd Gigerenzer – Director of the Harding Center for Risk Literacy, University of Potsdam
Title: Psychological AI: Simplicity and Transparency in Prediction
Abstract: Psychological AI is an elaboration of Herbert Simon’s original vision of AI, where the “I” refers to human intelligence as simulated by a machine. By studying the heuristics that experts use and programming these into software, it aims to improve the ability of computers to perform cognitive tasks. I argue that psychological AI originally failed because Simon and others applied it to stable situations such as chess, whereas its promise lies in making predictions in unstable environments, similar to those in which the human mind evolved. I discuss cases where amazingly simple psychological heuristics make more accurate predictions than complex big data algorithms such as Google Flu Trends. These insights offer an alternative to current attempts at understandable AI, and can introduce beams of transparency into an ever-darkening black-box society.
Gigerenzer, G. (March 2022). How to stay smart in a smart world. London: Penguin.
Katsikopoulos, K., Simsek, O. & Gigerenzer, G. (2021). Classification in the wild. Cambridge, MA: MIT Press. -
Day 2:
11:00 AM EST: Ute Schmid – Head of the Cognitive Systems Group, University of Bamberg
Title: Reconciling knowledge-based and data-driven AI for human-in-the-loop machine learning
Abstract: For many practical applications of machine learning it is appropriate or even necessary to make use of human expertise to compensate a too small amount or low quality of data. Taking into account knowledge which is available in explicit form reduces the amount of data needed for learning. Furthermore, even if domain experts cannot formulate knowledge explicitly, they typically can recognize and correct erroneous decisions or actions. This type of implicit knowledge can be injected into the learning process to guide model adapation. These insights have led to the so-called third wave of AI with a focus on explainablity (XAI). In the talk, I will introduce research on explanatory and interactive machine learning. I will present inductive programming as a powerful approach to learn interpretable models in relational domains. Arguing for the need of specific exlanations for different stakeholders and goals, I will introduce different types of explanations based on theories and findings from cognitive science. Furthermore, I will show how intelligent tutor systems and XAI can be combined to support constructive learning. Algorithmic realisations of explanation generation will be complemented with results from psychological experiments investigating the effect on joint human-AI task performance and trust. Finally, current research projects are introduced to illustrate applications of the presented work in medical diagnostics, quality control in industrial production, file management, and accountability.
1:15 PM EST: (Herbert A. Simon Prize Talk) James Allen – John H. Dessauer Professor of Computer Science, University of Rochester
Title: Conversational systems: Past, Present and Future
Abstract: A system that can carry on an extended conversation with a person involving substantive content is the holy grail of Artificial Intelligence research. It has been an active area of research for over fifty years. This talk will examine the history of ideas and advances, somewhat biased by the work from my research group over the years. I will also cover recent work in the area involving deep learning and consider whether such techniques bring the dream within reach. I’ll conclude by examining the tension between the desire for explanatory and inspectable models vs apparently high performing but completely opaque systems, and speculate on how the future of work in the area will unfold.
- Day 3:
1:00 PM EST: Anthony Cohn – Professor of Automated Reasoning, University of Leeds
Title: Manipulation in cluttered environments and interacting with robots
Abstract: Humans, in comparison to robots, are remarkably adept at reaching for objects in cluttered environments. The best existing robot planners are based on random sampling of configuration space- which becomes excessively high-dimensional with large number of objects. Consequently, most planners often fail to efficiently find object manipulation plans in such environments. I will present two methods to address this challenge; firstly, by learning from human demonstrations in VR and learning a model which can be used to build a high level plan which can then be used to guide a traditional stochastic planner; secondly exploiting human hints at run time to guide a stochastic trajectory optimising planner. Finally, if time allows, I will talk about how we can take account of human comfort whilst collaboratively working with robots.
1:45 PM EST: Panel Discussion: Research Directions for Cognitive Systems
|
Abstract: Given recent rapid advances in many areas of Artificial Intelligence, including but not limited to machine learning, access to large data, knowledge-graphs, robotics and autonomous vehicles to name a few, this community has a great opportunity to grow by demonstrating the relevance of its approaches through cross-fertilization with other research that could benefit from a cognitive systems architectural perspective and vice versa. This panel will lead a group discussion on how we might do that, and thereby encourage more diverse participation at this conference. Panelists: |