PAPERS INSTALLED IN THE YEAR 2004 (APPROXIMATELY)
This file is
Maintained by Aaron Sloman -- who does not respond to Facebook requests.
It contains an index to files in the Cognition and Affect Project's FTP/Web directory produced or published in the year 2004. Some of the papers published in this period were produced earlier and are included in one of the lists for an earlier period http://www.cs.bham.ac.uk/research/cogaff/0-INDEX.html#contents
Last updated: 10 Oct 2009; 13 Nov 2010; 7 Jul 2012
In some cases other versions of the files can be provided on request. Email A.Sloman@cs.bham.ac.uk requesting conversion.
Title: Damasio's Error
Date Published: 4th Quarter 2004
In The Philosophers' Magazine 2004, pp 61-64
Abstract (Opening Paragraphs):
In 1994 Antonio Damasio, a well known neuroscientist, published his book Descartes'
Error. He argued that emotions are needed for intelligence, and accused Descartes and
many others of not grasping that. In 1996 Daniel Goleman published Emotional
Intelligence: Why It Can Matter More than IQ , quoting Damasio with approval, as did
Rosalind Picard a year later in her book Affective Computing.
Since then there has been a flood of publications and projects echoing Damasio's
claim. Many researchers in artificial intelligence have become convinced that
emotions are essential for intelligence, and they are now producing many computer
models containing a module called `emotion'.
(This article criticises the current fashion.)
Title: Do machines, natural or artificial, really need emotions?
Talk to Birmingham Cafe Scientifique & Culturel Birmingham, 7th May 2004
Revised version presented on 24th June 2005 in Utrecht at The 3rd multi-disciplinary symposium organized by the NWO Cognition Programme: How rational are we?
Author: Aaron Sloman
Date installed: 11 Aug 2004 (Updated several times)
For full abstract follow link above. (Includes critique of Damasio's fashionable view that emotions are required for intelligence).
in Cognitive Systems Research, Volume 6, Issue 2, June 2005, Pages 145-174
Online since Sept 2004 at ScienceDirect.
NOTE: (by August 2005) this paper was third in the list of top 25 most downloaded articles in the journal. See
Much revised version of paper originally presented at: International Workshop Biologically-Inspired Robotics: The Legacy of W.Grey Walter, 14-16 August 2002, Bristol, UK http://www.ecs.soton.ac.uk/~rid/wgw02/home.html
Animals and robots perceiving and acting in a world require an ontology that accommodates entities, processes, states of affairs, etc., in their environment. If the perceived environment includes information-processing systems, the ontology should reflect that. Scientists studying such systems need an ontology that includes the first-order ontology characterising physical phenomena, the second-order ontology characterising perceivers of physical phenomena, and a (recursive) third order ontology characterising perceivers of perceivers, including introspectors. We argue that second- and third-order ontologies refer to contents of virtual machines and examine requirements for scientific investigation of combined virtual and physical machines, such as animals and robots. We show how the CogAff architecture schema, combining reactive, deliberative, and meta-management categories, provides a first draft schematic third-order ontology for describing a wide range of natural and artificial agents. Many previously proposed architectures use only a subset of CogAff, including subsumption architectures, contention-scheduling systems, architectures with `executive functions' and a variety of types of `Omega' architectures. Adding a multiply-connected, fast-acting `alarm' mechanism within the CogAff framework accounts for several varieties of emotions. H-CogAff, a special case of CogAff, is postulated as a minimal architecture specification for a human-like system. We illustrate use of the CogAff schema in comparing H-CogAff with Clarion, a well known architecture. One implication is that reliance on concepts tied to observation and experiment can harmfully restrict explanatory theorising, since what an information processor is doing cannot, in general, be determined by using the standard observational techniques of the physical sciences or laboratory experiments. Like theoretical physics, cognitive science needs to be highly speculative to make progress.
Keywords: Architecture, biology, evolution, information-processing, ontology, ontological blindness, robotics, virtual machines
Title: Interactions between Philosophy and Artificial Intelligence:
The role of intuition and non-logical reasoning in intelligence,
Author: Aaron Sloman
NOW TRANSFERED TO: http://www.cs.bham.ac.uk/research/projects/cogaff/62-80.html#1971-02
This paper is a sequel to my invited contribution to PPSN2000. It attempts to identify and analyse a collection of issues implicitly taken for granted in the earlier paper and in a great deal of literature which assumes that biological organisms do information processing. Normally it is assumed that we all understand intuitively what it means for something to be an information-processor, whether natural or artificial. I attempt to offer the beginning of an analysis which attempts to justify many of the ordinary ways of talking about information in organisms -- some of which attract critical comments from those who are sceptical about attempts to talk about computation and representations in organisms. In the long run I hope to show that that scepticism is misguided.
Title: Simulating Infant-Carer Relationship Dynamics
Presented at cross-disciplinary workshop on Architectures for
Modeling Emotion at the AAAI Spring Symposium at Stanford University in
Author: Dean Petters
Date added: 15 Feb 2004
Advances in autonomous agent technology have resulted in the potential for implementations of multiple agents to act as psychological theories of complex social and affective phenomena. Simulating attachment behaviours in infancy provides a relatively simple starting point for this type of theory development. The presence of neurophysiological, psychological and other types of data facilitates the validation of architectural theories by constraining these architectures at multiple levels. A seven part design process is described which details how requirements are specified and how design, implementation and evaluation processes are carried out. Two competing theories are proposed, one that involves some deliberation and one that is reactive only.
For movies see http://www.cs.bham.ac.uk/research/poplog/figs/simagent
The slide presentation is here
Title: What are emotion theories about?
Invited talk at cross-disciplinary workshop on Architectures for
Modeling Emotion at the AAAI Spring Symposium at Stanford University in
Date added: 29 Jan 2004
The slide presentation is here
This is a set of notes relating to an invited talk at the cross-disciplinary workshop on Architectures for Modeling Emotion at the AAAI Spring Symposium at Stanford University in March 2004. The organisers of the workshop note that work on emotions "is often carried out in an ad hoc manner", and hope to remedy this by focusing on two themes (a) validation of emotion models and architectures, and (b) relevance of recent findings from affective neuroscience research. I shall focus mainly on (a), but in a manner which, I hope is relevant to (b), by addressing the need for conceptual clarification to remove, or at least reduce, the ad-hocery, both in modelling and in empirical research. In particular I try to show how a design-based approach can provide an improved conceptual framework and sharpen empirical questions relating to the study of mind and brain. From this standpoint it turns out that what are normally called emotions are a somewhat fuzzy subset of a larger class of states and processes that can arise out of interactions between different mechanisms in an architecture. What exactly the architecture is will determine both the larger class and the subset, since different architectures support different classes of states and processes. In order to develop the design-based approach we need a good ontology for characterising varieties of architectures and the states and processes that can occur in them. At present this too is often a matter of much ad-hocery. We propose steps toward a remedy.
Title: The St. Thomas common sense symposium: designing architectures for human-level intelligence.
To Appear in The AI Magazine in 2004.Authors: Marvin Minsky, Push Singh and Aaron Sloman,
To build a machine that has common sense was once a principal goal in the field of Artificial Intelligence. But most researchers in recent years have retreated from that ambitious aim. Instead, each developed some special technique that could deal with some class of problem well, but does poorly at almost everything else. An outsider might regard our field as a chaotic array of attempts to exploit the advantages of (for example) Neural Networks, Formal Logic, Genetic Programming, or Statistical Inference with the proponents of each method maintaining that their chosen technique will someday replace most of the other competitors.
We do not mean to dismiss any particular technique. However, we are convinced that no one such method will ever turn out to be best, and that instead, the powerful AI systems of the future will use a diverse array of resources that, together, will deal with a great range of problems. In other words, we should not seek a single unified theory! To build a machine that s resourceful enough to have human-like common sense, we must develop ways to combine the advantages of multiple methods to represent knowledge, multiple ways to make inferences, and multiple ways to learn. We held a two-day symposium in St. Thomas, U. S. Virgin Islands, to discuss such a Project to develop new architectural schemes that can bridge between different strategies and representations. This article reports on the events and ideas developed at this meeting, subsequent thoughts by the authors on how to make progress.
Title: The ways to improve intelligence of interacting agents
Author: Marek Kopicki
Semester 1 Miniproject submitted as part of work (Oct-Dec 2003) for MSc in Advanced Computer Science, University of Birmingham.
A video demonstration of the program can be found here (look for the hybrid reactive/deliberative sheepdog). The whole program can be downloaded and run within the Free Poplog environment on a PC running Linux or a Sun, or Windows PC using VMware. The SimAgent Toolkit with this program included is part of the linux PC poplog package (21 Mb).
Abstract (expanded 11 Aug 2004):
Path planning is not a trivial problem of artificial intelligence. An agent has to find a path from one state (or position) to another whilst avoiding contact with obstacles. The configuration space used for representation of all agent states is usually continuous, which makes the problem even more complex. Skeletonisation is one of approaches, which discretises continuous space and reduces it to a graph search problem.
The sheepdog demo is a Pop-11 written computer simulation of an artificial world consisting of a dog, sheep, trees (obstacles) and a pen. The program uses the SimAgent toolkit to implement all the agents, the objects in the scene, and the various concurrently active components of the dog's 'mind'. The task of the dog is to drive all the sheep to the pen avoiding collisions with trees and other agents. An earlier version of the sheepdog, produced by previous MSc students, was purely reactive, so that it could not cope with complex barriers and mazes, which require planning if they are to be traversed in a sensible way. This version of the program adds a sophisticated planning capability. Probabilistic roadmap and A* graph search algorithm play a major role in the current refined version of the simulation, changing an original stimulus-response paradigm. The program combines deliberative planning with reactive plan execution including reactive local plan optimisation during execution. I will present advantages of using agent planning, and I will attempt to confront this traditional AI conceptual approach to a concept-free, perception-action architecture proposed by Rodney Brooks.
Even though the program still needs lots of improvements the overall result of the simulation is promising - the dog is able to complete the task, avoiding dynamically obstacles and changing the plan if necessary. The future version of the program might involve smarter skeletonisation procedure, more extensive use of a sim agent toolkit, and possibly one of approximate search algorithms to tackle more complex environment.
See also the School of Computer Science Web page.