A Brief Overview of Artificial Intelligence Focusing on Computational Models of Emotions

Brigitte Lorenz and Etienne Barnard

The development of intelligent systems has been a significant research goal for many years, and has proceeded from a number perspectives. We briefly review three of the most influential approaches, namely those derived from symbolic logic, from statistical models, and from the study of embodied agents. We then focus on the role of emotion in cognition, and review several computational models of emotions. These models raise a number of issues on the role of emotions in behavior, and show great promise to extend the capabilities of artificial intelligence.

The slides of this presentation can be downloaded here.

Considering a Technical Realization of a Neuro-Psychoanalytical Model of the Mind - a Theoretical Framework

Dietmar Dietrich, Georg Fodor, Wolfgang Kastner and Mihaela Ulieru

We use the psychoanalytical model of the psychical apparatus to define a unified coherent model for intelligent bionic systems. The terms intelligence, feelings and emotions are central topics within the fields of psychology, pedagogy and psychoanalysis. When engineers use these terms, they have to consider the concepts of those scientific fields. Our heterogeneous team joining engineers and psychoanalysts attempts to map Sigmund Freud’s model of the "psychical apparatus" in combination with Luria’s Dynamic Neuropsychology into a machine. Following up on the first paper of this forum which outlined the state-of-the art in Artificial Intelligence, this paper outlines the motivation of our new scientific step and describes visions and constraints we have encountered to date. Research results are presented in the following papers.

What is the Mind?

Mark Solms

This paper attempts to characterise the essential features of ‘the mind’ from a psychoanalytical standpoint. The perspective of psychoanalysis is a subjective one (first-person, introspective). This perspective can only be used in relation to minds. What is unique about the psychoanalytic first-person perspective is that it infers unconscious subjective events where discontinuities occur in the flow of conscious subjective experiences; in other words it fills the explanatory gaps in mental life with inferred unconscious mental experiences. This conceptual innovation generates a complete causal sequence of purely subjective events, and thereby renders possible a natural science of the mind.

This paper then considers from the psychoanalytical standpoint the relationship between (a) mental events and (b) brain events. It concludes that these are different observational perspectives on a unitary part of nature. The underlying substrate of both brain and mind is an abstract entity called the ‘mental apparatus’. The mental apparatus has the same ontological status as other natural kinds that are inferred from empirical observation but cannot be observed directly, e.g., ‘gravity’, ‘electricity’, ‘quarks’. Attempts to artificially engineer such things are best conducted at the same (abstract, functional) ontological level.

Lastly, and to this end, the cardinal functional characteristics of the human mental apparatus are identified. They appear to be: (1) consciousness, (2) intentionality and (3) agency. These characteristics are deeply interrelated, in that they all reflect the demands made upon the mind to perform work on behalf of the body, to maximise the chances of the body surviving to reproduce. An artificial mind must possess analogous characteristics, in the service of analogous values. To monitor the performance of the apparatus in these respects, it should also ideally be equipped with a capacity for reflexive (quasi-subjective) report.

Machines in the Ghost

Aaron Sloman

This paper summarises a subset of the ideas I have been working on over the last 35 years or so, about relations between the study of natural minds and the design of artificial minds, and the requirements for both sorts of minds. The key idea is that natural minds are information-processing machines produced by evolution. What sort of information processing machine a human mind is requires much detailed investigation of the many kinds of things minds can do. In particular, it is not clear whether producing artificial minds with similar powers will require new kinds of computing machinery or merely much faster and bigger computers than we have now. Insofar as psychotherapy is analogous to run-time debugging of a virtual machine, in order to do good psychotherapy we need to understand the architecture of the machine well enough to know what sorts of bugs can develop and which ones can be removed, or have their impact reduced, and how. Otherwise treatment will be a hit-and-miss affair.

The slides for this presentation and additional information can be found at http://www.cs.bham.ac.uk/research/projects/cogaff/talks/#enf07.

Simulating the Primal Emotions of the Mammalian Brain: The Affective Feelings of Mental Life and Implications for AI-Robotics

Jaak Panksepp

This paper discusses the manner in which emotional feelings may be created in biological brains. The affective neuroscience approach to understanding these issues is premised on the fact that basic emotions reflect large-scale neurodynamics of various kinds of emotional-instinctual action circuits, that are built into the brain as large-scale network functions that are better described in terms of "state-spaces" rather than "information-processing" algorithms. Cognitive mentality is built upon a solid foundation of emotional-affective processes which require explication of network-doctrines much more than traditional information-processing neuron-type doctrines that seem to suffice to model cognitive processes. The only experiencing minds that currently exist in the world are complex carbon-based biophysical, neurochemical, network-based whose brain capacities are intimately linked to the dynamics of living bodies. It is possible that such affective-emotional properties of biological brains need to be closely emulated in virtual machines before there can be any semblance of success in simulating the core properties of primary-process consciousness (e.g, the varieties of affective experiences), which may be essential foundations for higher forms of cognitive activities. Sloman’s impressive cognitively driven virtual machines fail to adequately confront the affective nature of the human mind.

The slides of this presentation can be downloaded here.

Cognitive and Affective Automation: Machines Using the Psychoanalytic Model of the Human Mind

Peter Palensky, Brigitte Lorenz and Andrea Clarici

Technical systems get into serious troubles, if they are confronted with a certain degree of complexity. An analytical, deterministic description of a complex problem is often not possible, and so its solution. This is especially true for automation systems of the future. Far away from scalar control loops and PLC (programmable logic controller) based machinery control, future systems are supposed to process a tremendous amount of information coming from millions of sensors and complex information sources like cameras. Large numbers of inexpensive and diverse sources of information can increase the performance of automation tasks in buildings, factories, transport systems, or machinery. However, the complex and uncertain semantics of such large amounts of data make bit-by-bit processing and traditional rule-based decisions impossible. A new trail from the sensor to the decision is necessary.

This paper describes a bionic approach to this problem. The human mind, as described in the latest findings of neurology and psychoanalysis, gives a blueprint of a system that is potentially capable of filtering, evaluating, and judging situations and scenarios. The relationship between system/environment interactions, memory, emotions, learning, and higher mental processes is believed to be the key for the success of our species. Translating psychoanalytic insights of the mind to computers and machinery is a big challenge, but there is the promise of creating a more capable technical systems.

The slides of this presentation can be downloaded here.

Response to Palensky, Lorenz and Clarici: Issues at the interface of artificial intelligence and psychoanalysis

David Olds

This comprehensive and thoughtful paper by Palensky, Lorenz and Clarici discusses the attempt to use neuropsychoanalytic theory to provide models and questions for the developing undertaking of artificial-intelligence scientists as they increase the effectiveness of machines that provide functions similar to those of the human brain. They give a succinct and sophisticated summary of psychoanalytic theory and its major contemporary variants, a summary compatible with the thinking of most analysts identified as ego psychologists. They give important weight to emotions as part of the cognitive apparatus, since emotions may be needed for the more complicated tasks of sophisticated robots. At the same time they hold the prevailing view that consciousness is an even more complex phenomenon than emotion, but one that we are not yet ready to confront. This commentary will raise a different possibility, namely that we should introduce consciousness into AI thinking, and even into bionic AI machines, from the very beginning. I will try to envision how consciousness could be seen as a kind of affect system that evolution installed very early in the history of motile organisms. The crucial criterion is salience, or "worth attending to;" and this may be as primitive as the basic building blocks of the emotion system, approach and avoidance, which, in mammals, are rooted deep in the brainstem. My hypothesis is that the basics of motivation in motile creatures are approach, avoidance and salience; thus the birth of the motivational system includes a building block of consciousness from very early on. Consciousness, as we know it, evolved in its current complexity from such a beginning. If this is so, then it may be useful to build into our robots a salience device, which will evaluate the importance of things. Even if my argument is completely wrong from an evolutionary point of view, it may still provide for interesting discussion.

Lunch break

The Prometheus Phantasy

Gerhard Pratl and Elisabeth Brainin

It is our goal to create technical systems that are able to understand their environment in a way similar to the way human beings understand their environment. Consequently, they shall also be able to interact with their environment – including humans. To create such systems we need models of human abilities. We need to understand how humans perceive their environment, how knowledge about the environment is created and stored, how motor interaction works, how decisions are taken and actions are executed. And the models that we require have to allow us to implement them in a technical, not a biological way. We can only use technology that is currently available (we cannot "grow" our own biological brain). This paper identifies principles and functions of the human psyche that make up a set of basic requirements for developing technical systems that can perceive and interact with the environment – including human beings – in a more human-like way.

The slides of this presentation can be downloaded here.

Return of The Zombie – Neuropsychoanalysis, Consciousness, and the Engineering of Psychic Functions

Yoram Yovell

In order to design technical systems that are based on a top-down analysis of human mental functioning, a working model of the mind is needed. Neurobiological and philosophical formulations of the mind often equate it with consciousness. As Pratl and Brainin (2007) have noted, the prospects of building a truly conscious machine appear remote at this time. However, neuropsychoanalysis does not view consciousness as synonymous with the mind, but rather as a property of one part of it. Consciousness is not a unitary phenomenon, and it may occur on several levels. Furthermore, there are situations in which high-level, complex, and distinctly human cognitive and emotional processing of information may occur without conscious awareness. It might therefore be possible to design technical systems that emulate important aspects of the mind without being conscious.

The slides of this presentation can be downloaded here.