In September 2017 we organize a summer school at TU Dresden, Germany.
The topic of this year’s summer school is Bridging the Gap between Human and Automated Reasoning
As in the past summer schools at TU Dresden in 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2010, 2013 and 2015 people from distinct, but communicating communities will gather in an informal and friendly atmosphere. This two-week event is aimed at graduate students, researchers and practitioners. The summer school is supported by the German Academic Exchange Service (DAAD) and therefore, a limited number of grants for students and university employees will be available, which includes a waiver for the participation fee. The grant includes a waiver of the participation fee, a lump-sum for living costs of 500 EUR and a lump-sum for travel costs, where the amount depends on the country.
I consider competing views of the nature of the mental representations and cognitive processes that underlie understanding and reasoning about counterfactual alternatives. One view is that counterfactuals are based on the computation of possibilities and another is that they are based on probabilities. I report experimental results on judgments of the truth of counterfactuals that indicate that people judge their truth based on the computation of possibility rather than probability. I also consider alternative views about the format of the mental representation of counterfactuals. One view is that it can incorporate symbols, e.g., a counterfactual to a set of facts can be represented by a symbol for negation, and another view is that it relies on embodied meaning, e.g., a counterfactual to a set of facts is represented by alternates. I report experimental results on inferences from counterfactual conditionals that indicate that counterfactuals can be represented symbolically rather than by embodied meaning.
Automated Reasoning by Ulrich Furbach (Universität Koblenz-Landau, Germany)
The course will introduce basic aspects of logical calculi for automated reasoning. The resolution calculus will be presented and basic properties of proof procedures are given. Techniques for controlling the search space are discussed and methods for equality handling are introduced. As an alternative, tableau calculi are introduced and the relation to resolution and decision prodedures for description logics are discussed. In a hands-on tutorials the participants will learn to use automated theorem provers from the Systems-on- TPTP website, where all the available systems can be used with own examples or the TPTP benchmark suite.
Tableau Calculi and Applications by Ulrich Furbach (Universität Koblenz-Landau, Germany)
This course will cover first order tableau calculi with a focus on hyper tableaux. The calculus rules together with a comparison to other logical systems are introduced and an extension for an effcient handling of equality is given. Various applications for a hyper-tableau-system are discussed and, in particular, the Loganswer-System (www.loganswer.de) as an example of the cognitive computing paradigm is introduced. Based on these applications some extensions and requirements for a proof-system are explained; in particular, handling of large knowledge bases (like e.g. Cyc), webservices and abductive answers.
Abstract Argumentation – Reasoning, Expressiveness and its Connection to Answer Set Programming by Sarah Gaggl (Technische Universität Dresden)
Argumentation is one of the major fields in artifcial intelligence (AI) and non-monotonic reasoning (NMR). Nowadays, the concept of abstract argumentation frameworks (AFs) is one of the most popular approaches to capture certain aspects of argumentation. This very simple yet expressive model has been introduced by Phan Minh Dung in 1995. Arguments and a binary “attack” relation between them, denoting conflicts, are the only components one needs for the representation of a wide range of problems and the reasoning therein. Nowadays numerous semantics exist to solve the inherent conflicts between the arguments by selecting sets of “acceptable” arguments. Depending on the application, acceptability is defined in different ways. Some semantics are based on the idea to defend arguments against attacks, while others treat arguments like different choices and the solutions stand for consistent sets of arguments. In this course we will first focus on the expressiveness of AFs, in particular we will study if, and under which conditions, a given set of arguments can be accepted at all in an AF under a given semantics. Furthermore, we will analyze different notions of equivalences for AFs, for example when two different AFs posses the same solutions under a semantics, even if we apply modifcations to them. Finally we will observe the connection between answer set programming (ASP) and AFs.
The course gives a comprehensive introduction into propositional and first-order logic: propositional formulas; interpretations and models; satisfiability, falsifiability, validity, unsatisfiability and their relations; truth tables; logical consequences; satisfiability problems (SAT); conjunctive normal form; SAT-solving; first-order formulas; substitutions; semantics. In the tutorials participants will be asked to turn a real world problem into a SAT-problem and to solve it using a state-of-the-art SAT-solver.
In the last nine years we have developed a new cognitive theory. It is based on the weak completion of logic programs, the three-valued Lukasiewizc logic, abduction and revision, and has been successfully applied to adequately model various human reasoning tasks like the suppression task, the selection task, the belief bias effect, syllogistic reasoning, spatial reasoning as well as reasoning about conditionals. In the course we will give an in-depth introduction into the new theory as well to its applications to different human reasoning tasks. In addition we will do experiments in order to evaluate certain reasoning tasks, in particular, how humans reason with conditionals.
Cognitive Science by Marco Ragni (University of Freiburg, Germany)
This course aims at a core introduction into cognitive science with a special emphasis on foundational aspects and general assumptions about human representation, memory, and reasoning. The following questions will be approached: Which methods are used to learn about behavioral regularities in human reasoners? What is the influence of perception on internal mental representation? What can be considered a `good’ cognitive models for human perceptive and reasoning processes?Which special aspects do we know about internal mental models for different descriptions? This course will provide several “do-it-yourself” examples and prepare for the advanced course “Modern approaches to human reasoning.”
Modern Approaches to Human Reasoning by Marco Ragni (University of Freiburg, Germany)
Since the times of Aristotle humans have discussed what can be regarded as an acceptable inference striving the border of formal logical inferences and a form of “natural deduction” humans are applying. The ability to gain new insights from given knowledge by reasoning is one of the most fundamental cognitive abilities of humans. Psychological findings show a difference between inferences drawn by humans and by formal reasoning systems implementing classical logic. These differences can be found for all domains: in reasoning about relations, in reasoning about conditional statements, and in reasoning about syllogisms. In this course I will first introduce several examples for each domain demonstrating specific effects in reasoning, including content effects and illusions. In a second step different theories of reasoning are presented and applied to the different effects. In a third step the theory is evaluated and differences to other theories
and alternative approaches are discussed.