Psychoanalytic Defense Mechanisms in Cognitive Multi-Agent Systems

Friedrich Gelbard

• Publisher: Routledge
• Year: 2017
• ISBN:
  9781351867689
• Edition: 1
• Topic: Psychology
• Subtopic: Cognitive Psychology & Cognition
• Index: Psychology

Human cognitive processes and defense mechanisms, as described in psychoanalysis, bring about new notions and paradigms for artificial intelligence systems. One key reason is that the human cognitive processes and defense mechanisms in question can accomplish conflict detection functionalities, filter functionalities, and other system stabilizing tasks within artificial intelligence systems. Yet artificial cognitive architectures lack the capability to analyze complex situations as well as the universal competencies needed to orientate themselves in complex environments in various domains. Psychoanalytic Defense Mechanisms in Cognitive Multi-Agent Systems addresses this dilemma by exploring how to describe, model, and implement psychoanalytic defense mechanisms in the course of a project that provides a functional model of the human mind.

With discussions focusing on the development of a mathematical description for the implementation of conflict detection, the activation and selection of defense mechanisms, and the processing of defense mechanisms, Psychoanalytic Defense Mechanisms in Cognitive Multi-Agent Systems describes the decisive points for the application of defense mechanisms in artificial intelligence. Formulae that treat defense mechanisms as transformations are also provided. Interdisciplinary cooperation between the scientific fields of psychoanalysis and artificial intelligence is highlighted as the foundation of new research findings throughout the book.

Innovative and exciting, this book will be of great interest to academics, researchers, and postgraduates in the fields of cognitive science, artificial intelligence, and psychoanalysis.

1 Motivation, requirements, and methodology

The work at hand is part of the ongoing ARS (Artificial Recognition System) project [1]. The aim of ARS is to devise and implement a technical model of the human mind [2] according to psychoanalytic notions. In the present chapter, the scientific context and a brief overview of the scientific field and the ARS project in particular are given, followed by the methodology of how psychoanalytic notions can be transferred to be implemented in artificial intelligence. Furthermore, motivations for applying psychoanalytic knowledge in artificial intelligence are explained, and the scope of the present work is outlined. After that, the basic paradigms of the development process and implementation process applied in the work at hand are described. Six scientific statements that form the basic pillars for the work at hand are explained, and, finally, the process of interdisciplinary cooperation is outlined.

1.1 Context and overview

There are many projects that aim at devising a universal artificial intelligent system for problem solving in many domains. The first group of projects contains data mining and machine learning projects. The foundations for such projects are statistical methods to cluster sets of data, to categorize data, and to find patterns in given sets of data. An overview of data mining and machine learning concepts and techniques is given in [SSG10], [HK06], [WF05], [Jac02], and [CHY96]. Example architectures of artificial neural networks are given in [AB99].

Second, there are projects that claim to represent artificial general intelligence. They model semantic knowledge with the help of ontologies and model semantic relations of entities and semantic context. Members of this group are, for example, Novamente [GP07, pp. 63–129] or NARS [Wan04]. These projects rely upon logic based inference machines.

Furthermore, there are projects that are located in the field of cognitive science. An overview is given in [DOP08], [VHF10], and [LLR09]. These projects try to take human thinking processes as a role model and strive to understand and model humans’ cognitive capabilities.

Moreover, there are projects that aim at deploying software agents¹ and offer certain services via the Internet [SU10]. They use standardized agent communication languages such as FIPA_ACL [11] and are organized to communicate over distributed networks.

A totally different perspective is the field evolutionary computation [Fog06], genetic programming, and evolvable hardware (e.g., [GP07, pp. 158–174]). These projects aim at designing hardware or software that can follow a certain kind of evolution process. The idea is to alter hardware or software in random directions and to select in a second step the best-adapted systems for a given problem. This cycle of alteration and selection is repeatedly employed.

Another group of research includes projects based on neuroscience that try to reconstruct the human brain’s neuronal functions in the form of a new computer architecture and/or software. For example, the Blue Brain Project [Mar06] by Henry Markram at the École Polytechnique Fédérale de Lausanne tries to simulate the neuronal brain structure with the help of IBM’s supercomputer Blue Gene with more than 65,000 compute nodes [Mar06, p. 154]. Since 2013, the successor project has been the Human Brain Project [Mar12]. The École Polytechnique Fédérale de Lausanne coordinates it. The Human Brain Project is a European Commission Future and Emerging Technologies flagship. The project is a platform for the following research fields: neuroinformatics, brain simulation, high-performance analytics and computing, medical informatics, neuromorphic computing, and neurorobotics.

The last group of projects model the human mind according to psychoanalytic notions [NTNI99], [Bul03], and [Mos09]. These projects are subjects of the work at hand and will be discussed in Chapter 2. The projects want to improve artificial intelligence by incorporating new ideas taken from psychoanalysis. The work at hand, too, is a member of that group. In the present work and in the accompanying ARS project, psychoanalytic ideas and in particular the second topographical model [Eag11, pp. 28–30], [Lis09, pp. 95–97] (in German) are taken as the foundations for developing an implementable model of the human mind [DFZB09].

In 2000, Dietmar Dietrich elaborated on the foundations of the ARS project [Die00]. By that time, the idea of ARS was to generate specific conclusions of data provided by heterogeneous sensor networks in order to reach specific decisions. The aim was to develop decision units for complex applications that were not to be resolved with state of the art methodologies.

Other projects in the field of building automation and artificial intelligence followed: SmaKi started in 2000 [6], [Rus03], [Fue03]; SENSE started in 2006 [7], [ZF07]; PAIAS started in 2007 [8], [BKVH08]; ATTEND started in 2009 [9], [YB12]; and Computational Perception started in 2010 [10], [MPB11].

Close cooperation with psychoanalysts such as G. Fodor [RLFV06], [DLB⁺10]; E. Brainin [BDK⁺04]; and S. Teicher [DLP⁺06] brought the project to a state where S. Freud’s second topographical model and today’s view on Freudian psychoanalytic functionalities of the human mind are implemented in a simulation environment.

The ARS project was deeply influenced by the work of A. Luria [LP73]; A. Damasio [Dam94], [Dam98], [Dam00], [Dam03]; and M. Solms [KS00], [ST02], [Sol09]. In their work, they give comprehensive insight into psychoanalytic and neuropsychoanalytic ideas that can be used to model the human mind in artificial intelligence.

Several journal publications resulted from the ARS research such as [BPV⁺12], [BZD12], [DPD12], [BVP11], [VZD11], [BV10], [DBZP10], [DB10], [VZ10], [VB09], or [DZBM08]; conference publications such as [SWJ⁺14], [SDW⁺13], [WSG⁺13], [GB12], [YB12], [BG11], [GBD⁺11], [GBDK11], [DBM⁺10], [DLB⁺10], [DBZ⁺09]; and many more.

Furthermore, project ARS was presented to the public in different workshops, e.g.,

• The First International Engineering and Neuropsychoanalysis Forum (ENF 2007): Emulating the Mind, documented in the book Simulating the Mind [DFZB09], [2]
• A presentation at Wiener Arbeitskreis für Psychoanalyse in 2009 [3]
• Workshops at Wiener Psychoanalytische Akademie in 2010, March 2012, and October 2012 [4], [5]

1.2 Motivation

Defense mechanisms of the human mind resolve conflicts caused by inner drive wishes, internalized rules, external perceptions of the environment, and reality. Besides those primary goals of defense mechanisms, they serve as filter mechanisms to resolve conflicts and to reduce input data of internal sensors of the system and external perceptions. The defense mechanisms change (drive) aims of a software agent in case of a conflict, defense mechanisms help focus on perceptions that are assessed to be less conflicting than other perceptions, defense mechanisms filter/alter emotions of a software agent, and defense mechanisms cope with contradictory input data.

[Lis09, p. 91]: Defense mechanisms are (targeted) unconscious processes that prevent affects² and contents that cause unpleasure.³ (“Abwehrmechanismen sind (gezielte) unbewusste Vorgänge, welche vor Affekten und Inhalten, die Unlust bereiten, bewahren sollen.”)

All the aforementioned functionalities are important elements in artificial general intelligence. Defense mechanisms treat these elements in totally different ways ([GBDK11], [GBD⁺11]) compared to principles applied nowadays in biologically inspired cognitive architectures, such as described in [GLA⁺10].

Hence the advantage of using psychoanalytic defense mechanisms in artificial intelligence is that they regard artificial intelligence problems and tasks from a completely different perspective, and they solve these problems in their own way. Therefore, “defense mechanisms open a broad spectrum of new ways, opportunities, and insights for artificial intelligence which we do not even foresee now” [GBD⁺11, p. 4].

The three instances in psychoanalysis are the Id, the Ego, and the Super-Ego [Vel10b, pp. 17–19]. The Id represents inner drive wishes. The Super-Ego represents internalized, often social, rules. Now conflicts can arise between inner drive wishes, Super-Ego rules, and perceived reality. Here, in the third instance, the Ego diminishes the conflict strength and solves the conflict by using defense mechanisms. This way, conflict-free drive wishes can be satisfied and the interaction with the environment (reality) can be processed in a conflict-free manner.

Hence the main functionality of psychoanalytic defense mechanisms is to solve conflicts. The conflicts are caused by internal rules (Super-Ego rules), which contradict incoming data and/or desired aims of a software agent. Internal rules are necessary to maintain social behavior of interacting software agents.

Other tasks that can be accomplished with the help of internal rules and defense mechanisms are social interactions with other software agents according to certain social rules and to maintain the reliability of a software agent in social interactions with other software agents. To make social interactions possible, only reliable action plans are admitted. That is, defense mechanisms must alter or erase action plans and decisions that disturb the social processes.

So the following question may arise: “Why are defense mechanisms not applied in artificial intelligent systems up to the present?” The answer is, “Defense mechanisms are not applied in artificial intelligent systems up to present because of the intricacy of the psychoanalytic model of the defense mechanisms, the difficulty in translating the psychoanalytic model of the defense mechanisms into a technical implementable model, and the challenge to implement the devised technical model of defense mechanisms in artificial intelligence.” Efforts to develop a technical model of defense mechanisms in artificial intelligence, though, are part of some research projects [Bul03], [Mos09].

Two key elements make the human mind superior to artificial intelligence when it comes to the control of complex systems: The manifold valuation capabilities of the human mind of complex situations, objects, and persons, and the division of the functionalities of the human mind into a primary process and a secondary process. The primary process satisfies momentary desires. Whereas the tasks of the secondary process are decision making and action planning. The defense mechanisms serve as a kind of filter mechanism between the primary process and the secondary process. Decision making and action planning are only possible if momentary desires are withheld from the secondary process.

Furthermore, defense mechanisms are a pivotal part of the psychoanalytic model of the human mind. The aim of the ARS project is to transfer the psychoanalytic model of the human mind into a technical implementable model and to implement the technical model in a simulation environment. As of now, the technical model and the implementation of the defense mechanisms are missing in the ARS project.

Hence the aim of the work at hand is to transfer the psychoanalytic model of the defense mechanisms into a technical implementable model and to implement the technical model in a simulation environment.

The economic relevance of the implementation of a functional model of the human mind becomes evident in all areas of research where human’s behaviors, human’s cognitive capabilities, and human’s manifold assessment system of complex situations are explored. Funded projects in building automation (project ECABA) [ZHB⁺15] and in examining consumer behavior in renewable energy projects (project CogMAS) [SDG⁺15] underline the economic importance of such a functional model of the human mind. In these funded projects, the functional model of the human mind and the comprising defense mechanisms are implemented and used to improve people’s well-being in building automation, as well as the examination of human’s valuation capabilities to estimate and predict consumer behavior. The project manager and initiator of the funded projects mentioned earlier is S. Schaat [SD14] and [SDD14]. The aforementioned funded projects are not the subject of the present work.

1.3 Requirements

In order to implement the technical model of psychoanalytic defense mechanisms in a simulation environment, the author defines several basic scientific conditions and constraints. The basic conditions and constraints allow clear and distinct scientific work without misunderstandings between different scientific disciplines such as psychoanalysis, cognitive science, and software engineering. In some projects ...

Table of contents

• Cover
• Title
• Copyright
• Contents
• List of Figures
• Acknowledgments
• Abbreviations
• 1 Motivation, requirements, and methodology
• 2 Context of the field of research
• 3 Concepts to develop a technical model of the human mind
• 4 Technical model of the defense processes
• 5 Implementation of the defense processes
• 6 Simulation environment and performed experiments
• 7 Conclusion and outlook
• Index