Dec 9, 2021, 9-11 am PST
YouTube link to panel recording
How can we conceptualize and construct artificial agents with rich autonomy? How can we use computational models to understand the agency of humans, and shape the collaboration between human and AI agents? Our panel brings a group of thinkers about artificial agency, motivation, emotion and sociality together, to discuss how intrinsic motivation gives rise to goal directed behavior, the organization of cognitive structure, multi agent collaboration and ethics.
Abstract: Two distinct notions of “goal”, of purposiveness: mental representations, goal-driven behavior vs. ‘functions’ of the behavior. Architecture for motivation and goal “value” as grounded on beliefs; belief-based goal processing and intention formulation.
Modeling different motives in Agents’ mind for/in Goal Adoption: help and guardianship; exchange; true ‘cooperation’ for a common goal; normative reasons, duties.
What is “autonomy” and why we need autonomy in an AI partner and initiative for good/efficient cooperation.
Bio: Cristiano Castelfranchi is the Director of the Institute for Sciences and Technologies, National Research Council , and a Full Prof. of Cognitive Science, University of Siena.
The guiding aim of Castelfranchi’s research is to study autonomous goal-directed behavior as the root of all social phenomena, at the same time highlighting how social life shapes individual cognition. The following contributions have earned him special recognition in this area: autonomous agency and goal dynamics; the cognitive structure of society; social and evolutionary functions and their relationship with individual goals; pathologies of goal-directed cognition.
Abstract: It is hard to devise a theory of motivation and emotion that fits perfectly with the everyday use of these concepts. The situation is not helped by the fact that most common words for emotions describe highly different functions that do not always operate in similar ways. This is why the idea of basic emotions is problematic and probably the explanation why such lists tend to be very different from theory to theory. Instead, I want to propose a theory of motivation and emotion based on their role in controlling behavior and learning. Such a theory allows for a straightforward implementations of these concepts in artificial systems. The model is based on cognitive processing at three different poles (motivation, emotions, attention), each of which influences the other two.
Each of the three processes are highly context dependent and by interacting with higher forms of cognition they give rise to a broad spectrum of different emotions.
Bio: Christian Balkenius is a professor of Cognitive Science at Lund University. His research focus is on computational modeling of cognitive and emotional processes and their use in the control of robots. His interdisciplinary research spans areas from neural engineering, AI, machine learning and robotics, to psychology and biology. He is the director of the research school of the Wallenberg AI, Autonomous Systems and Software Program – Humanities and Society, the goal of which is to study the consequences of AI for society. He has participated in four projects within EU’s framework programmes7. Balkenius has a large international network within academia that has resulted in many research collaborations and publications together with more than 100 people.
Abstract: Human behavior is driven by a variety of needs, which give rise to motivation along many dimensions. We can describe motivation as feedback control, and the entire motivational system as seeking dynamic homeostasis. Here, I will discuss the specific role played by the need for competence, which measures the ability of an agent to satisfy its needs. Subjectively, the need for competence manifests as the sense for efficacy (self worth). The competence motivation is used to develop and use heuristics and knowledge about the agent's ability to learn and explore, to estimate the probability of success for known and unknown skills, and ties into social behavior and creativity as well.
Bio: Dietrich Dörner (born 28 September 1938, Berlin) is emeritus professor for General and Theoretical Psychology at the Institute of Theoretical Psychology at the Otto-Friedrich University in Bamberg, Germany. In 1986, he received the Gottfried Wilhelm Leibniz Prize of the Deutsche Forschungsgemeinschaft, which is the highest honour awarded in German research.
The cognitive architecture Psi-Theory is developed under his guidance.
Abstract: We can understand agents as controllers with the ability to model the future, driven by a set of needs. What motivates a cognitive agent is not the reward itself, but the model of a future reward: the purpose. While competing needs do not form a hierarchy, purposes do. Purposes differ by their scope; scopes that reach beyond the individual agent and its expected lifetime give rise to collective agency. This understanding of shared purpose allows us to reason about ethics in human and machine contexts.
Bio: Joscha Bach, PhD, is a cognitive scientist and AI researcher with a focus on computational models of cognition and neuro-symbolic AI. He has taught and worked in AI research at Humboldt University of Berlin, the Institute for Cognitive Science in Osnabrück, the MIT media lab, the Harvard Program for Evolutionary Dynamics and is currently a principal AI researcher at Intel Labs, California.