About  |  Study  |  Members  |  Projects  |  Publications  |  Events  |  Search  |  Internals  |  Contact & Imprint

ICCL Summer School 2008  -  Course Program

Computational Logic and Cognitive Science: An Overview

Helmar Gust    (Universität Osnabrück, Germany)
Kai-Uwe Kühnberger    (Universität Osnabrück, Germany)

The relation between computational logic and cognitive science is not trivial: On the one hand, logical techniques seem to be clearly essential to model certain cognitive abilities like planning or reasoning. On the other hand, the history of cognitive science is full of examples, where several researchers tried to abandon logical means for modeling cognitive abilities, due to the fact that computational logic does not to seem to fit perfectly to explain certain cognitive phenomena. This course will give an overview of classical examples of such findings in cognitive science together with a discussion of potential consequences for computational logic as a means for modeling cognitive abilities. More precisely, the course will cover examples of human reasoning and rationality that are seemingly hard to model with classical logic like the famous Wason Selection task, analogical reasoning (Gentner), ecological rationality (Gigerenzer), the learning of certain abilities like to learn natural language (Chomsky), or the modeling of context and situatedness of agents. In parallel, the course gives a glimpse of how the usage of non-classical types of computational logical can nevertheless be used (and is in fact necessary) to overcome these problems. Examples are the usage of non-monotonic logics, neuro-symbolic reasoning mechanisms, frameworks for analogical reasoning, learning from inconsistencies, and model-based reasoning.

Slides:

Lecture 1 (pdf) | (ppt)
Lecture 2 (pdf) | (ppt)
Lecture 3 (pdf)
Lecture 4 (pdf)
Lecture 5 (pdf)

Human Reasoning and Cognitive Science

Michiel van Lambalgen    (University of Amsterdam, The Netherlands)
Keith Stenning    (Edinburgh University, United Kingdom)

The course will be based on the book of the same name by Keith Stenning and Michiel van Lambalgen at MIT Press. Currently available here. The book uses defeasible logics to show that the data used in the psychology of deductive reasoning has to be regarded chiefly as reasoning TO interpretations, rather than reasoning FROM them. This data is modeled in the defeasible logic `logic programming with negation as failure'. Neural implementations for this defeasible logic are given and implications for the evolution of human reasoning and language, and for understanding developmental syndromes are drawn. The course will be aimed at logic and linguistics students who are interested in getting to grips with the empirical literature on reasoning, or for cognitive psychologists wanting to gain formal understanding.

Slides:

Part 1 (pdf)
Part 2 (pdf)
Part 3 (pdf)
Part 4 (pdf)
Part 5 (pdf)

Computational Logic and Connectionist Systems

Steffen Hölldobler    (Technische Universität Dresden, Germany)

Three long-standing open research problems in connectionism are the questions of how to instantiate the power of symbolic computation within a fully connectionist system, how to represent and reason about structured objects and structure sensitive processes, and how to overcome the propositional fixation, i.e.\ how to use connectionist systems for symbolic learning and reasoning beyond propositional logic. In the course, I will present a method for the generation of models of logic programs using recurrent connectionist networks. After discussing in detail the propositional case, I will focus on first-order logic programs and the corresponding neural networks to provide a missing link between symbolic and connectionist computation.

Slides:

Part 1 - Propositional Logic (pdf)
Part 2 - First-order Logic (pdf)

Computational Logic in Human Reasoning

Robert Kowalski    (Imperial College, United Kingdom)

Formal logic was originally developed as a normative model of human reasoning. However, numerous psychological experiments, including the Wason selection task, suggest that logic plays little role in human reasoning. As a consequence, other, more computationally oriented approaches, such as production systems, have had a greater impact in cognitive science. In this tutorial, I will review some of the psychological and philosophical literature on human thinking and argue that computational logic can reconcile logical and computational models of human thinking.
Computational logic is a wide-spectrum language used in computing for both high-level specifications and low-level implementations. Moreover, the same kinds of reasoning that are used to execute logic programs at run time can also be used to transform specifications into implementations at compile time, and sometimes to decompile lower-level programs into higher-level form. This use of logic at multiple levels is analogous to the use of different levels of thinking in the dual process model.
In cognitive psychology, the dual process model hypothesizes that two kinds of thinking operate in tandem. Intuitive thinking operates automatically, effortlessly and subconsciously, while deliberative thinking operates serially, effortfully and consciously. Logic is normally associated only with deliberative thinking. However, I will argue that computational logic can be used to model both intuitive and deliberative thinking, as well as many of the relationships between them.
Logic is a general-purpose, domain-independent reasoning mechanism. In cognitive science, on the other hand, it is generally held that the mind is composed of special-purpose modules, in which general-purpose thinking is problematic and relatively unimportant. However, in computational logic, modularity and special-purpose knowledge are compatible with general-purpose reasoning. I will argue that this compatibility can also model the way in which special-purpose knowledge and general-purpose reasoning interact in the human mind.

Slides:

Part 1 (ppt)
Part 2 (ppt)
Part 2-3 (ppt)
Part 3 (ppt)
Part 4 (ppt)
Part 5 (ppt)

Logic-based agents

Fariba Sadri    (Imperial College, United Kingdom)

Intelligent agents in Artificial Intelligence are computer systems that interact with an external environment, observing changes and performing actions. A logic-based intelligent agent is one that uses logic to assimilate observations, to reason about goals and to decide what to do. In this tutorial, we will review a number of such logic-based agent models and show how computational logic can be used to implement their functionalities.
The notion of intelligent agent in AI can be understood as an extension of production systems, in which "thinking" is modeled by forward chaining with production rules. AgentSpeak is one of the most well-developed and most popular agent model of this kind. In production systems, AgentSpeak, and many other agent models, forward chaining is used to implement both reaction rules and goal-reduction procedures. We will show how such agent models can be modeled by logic-based agents that use abductive logic programming (ALP) as their "thinking" component.
ALP agents combine logic programs to represent beliefs, integrity constraints to represent goals, and abducible (undefined) predicates to represent observations and actions. They use logic programs to reason backwards, to reduce goals to sub-goals; and they use integrity constraints to reason forwards, to monitor both observations and candidate actions.
We will describe an extension of the ALP agent model, called the KGP (Knowledge-Goal-Plan) agent model, which has been developed in the EU SOCS (Societies of Computational Entities) project. The KGP model provides a hierarchical and modular architecture of capabilities, state transitions, and control theories, which can be used for planning, reactivity, goal decision and temporal reasoning. The control theory module allows different agent personality profiles to be implemented, providing a variety of different dynamic, context-dependent behaviours.

Slides:

Lecture 1 (ppt)
Lecture 2 (ppt)
Lecture 3 (ppt)
Lecture 4 (ppt)

The logic of generalized truth values
A tour into Philosophical Logic

Heinrich Wansing    (Technische Universität Dresden, Germany)

The notion of a truth value is among the most important notions of modern symbolic logic and analytic philosophy. It has been explicitly introduced by Gottlob Frege, who considered exactly two classical truth values, the True and the False, which played in his theory the role of references for sentences. In this connection Frege characterized logic as the discipline that should investigate the ``most general laws of being true''. Another prominent supporter of this view was Jan Lukasiewicz, who defined logic as the science of truth values. In Belnap's useful 4-valued logic of ``how a computer should think'', the set 2 = {T,F} of classical truth values is generalized to the set 4 = P(2) = {emptyset ,{T},{F},{T,F}}. In this course, I shall consider non-classical logics obtained form generalizing the set of classical truth values along the lines of Belnap's 4-valued logic. The focus will be on the bilattice FOUR_2 and the trilattice SIXTEEN_3 of generalized truth values. Another topic will be Suszko's Thesis, the claim that ``there are but two logical values, true and false'', which is given a formal contents by the so-called Suszko Reduction, the proof that every Tarskian n-valued propositional logic is also characterized by a bivalent semantics. The critical discussion of Suszko's Thesis will highlight two ways of defining many-valued logics, namely by means of sets of designated and antidesignated truth values and by means of ordering relations on the set of truth values.

For slides see
Lectures 1\&2 .
Lecture 3.
Lecture 4.
Lecture 5.
References.

Computational Logic Applications in Cognitive Sciences

Luís Moniz Pereira    (Universidade Nova de Lisboa, Portugal)

The general relevance of Computational Logic to the Cognitive Sciences will be examined by means of a number of detailed diverse applications, each corresponding to a running program. The applications concern the modeling of moral reasoning; the modeling of decision making by prospecting alternative uncertain future outcomes; the modeling of joint preference handling in a society of agents; the modeling of ascribing intention to others; etc. Moreover, general epistemological conclusions are drawn and an encompassing framework is set forth. The contents of the course are supported on readily available published papers by the lecturer.

Final versions of course material

Slides (tgz). Supporting Documents (tgz). Acorda (tgz). Lecture Recordings (tgz).

Preliminary course material

Slides (zip). Documents (zip). Acorda (zip). Evolution Prospection Software (zip).

You might need to rename the files FILE.zip.x.zip to FILE.zip before unzipping.

Cognitive Human Performance Models and Human Machine Interaction Design

Leon Urbas    (Technische Universität Dresden, Germany)

Human Machine Interaction (HMI) Design is an important task, in particular for safety-critical systems. Due to the complexity of the human operator's mind and his tasks in dynamic systems and also a lack of tools to cope with more than basic tasks formal methods are rarely used to test in advance if the interaction specification fulfils the requirements. In this course, I give an introduction to human performance modelling in highly specialized production systems - in particular ACT-R - that provide means to develop models showing behaviour similar to human behaviour based on psychological theories of memory, vision and motor system. It will make clear that current systems don't hold the promise of a convenient abstraction level to solve real world tasks. Therefore the second part of the course focuses on an approach that borrows from model driven architectures. I will present a method that allows to derive formal models of human information processing directly from a semi-formal task analysis with no 'hand waving' in between. This approach has the potential to help human factor experts to effectively and efficiently develop and apply cognitive human performance models for HMI-Design.

For slides see (update of literature in part 2 will follow)
Part1
Part2

Last update:   Mon, 22 Sep 2008 12:03:13