Researcher in Logic/Cognitive Science
I am an associate professor at the Department of Philosophy, Linguistics and Theory of Science, University of Gothenburg and the Department of Applied Information Technology at Chalmers University of Technology.
PhD in logic from the Department of Philosophy, University of Gothenburg in 1996. Post-doc at the Department of Philosophy, Utrecht University in 1997-98.
I worked mainly with academic spinoff companies in the period 1999-2010. Since 2011 I focus on academic research. From January 2013 I work with research 80% of the time.
My PhD was about provability logic, which is a part of metamathematics. I am interested in human reasoning and the intersection of cognitive psychology and artificial intelligence. I am involved in the following research projects:
- Human reasoning in logic. In this project we explore human reasoning in logic using standard methods of experimental and cognitive psychology. Our cognitive models are based on proof systems that enable us to model logical reasoning with bounded cognitive resources. We use automatic theorem provers that generate the shortest proofs with limited cognitive resources of any given sentence. In the case of propositional logic, this approach yielded a 0.89 correlation between psychological and mathematical complexity measures. The research group includes Mic Detlefsen (Notre-Dame), Fredrik Engström, Abdul Rahim Nizamani, Lance Rips (Nortwestern), and Tony Strimple (Notre-Dame). The project is financed by the Swedish Research Council (VR).
- Automatic concept formation. In this project we use a cognitive architecture inspired by Daniel Kahneman’s dual process theory and Herbert Simon’s theory of bounded reasoning. It includes a long-term memory (LTM) and a working memory (WM). The LTM is modeled as a graph-structure (transparent neural network). This graph structure evolves in a developmental process in which memory structures are created, updated, merged, and deleted. We aim for a monolithic computational model that bridges the gap between the sub-symbolic and symbolic paradigms. The research group includes Christian Balkenius (Lund), Olle Häggström (Chalmers), Rickard von Haugwitz (Chalmers), and Johan Wessberg (Sahlgrenska).
- Automatic pattern discovery. In this project we develop programs that can solve certain pattern discovery problems appearing in IQ tests. We have one program for progressive matrix problems and one for number sequence problems. In the case of progressive matrices, our program scores around IQ 100 (without looking at the answer alternatives of the multiple choice problems). In the case of number sequences, our program scored above IQ 130 on a previously unseen IQ test, whereas the math programs Mathematica, Maple, and Wolfram Alpha scored below IQ 100. The special feature of our approach is that our programs include models of human reasoning. In the case of number sequence, we look for the shortest description of a number sequence with the side condition that it must be possible for a human with limited cognitive resources to decode it. This yields a computable version of Kolmogorov complexity, which is defined in terms of cognitive models. The research group includes Fredrik Engström (Gothenburg), Abdul Rahim Nizamani (Gothenburg), and Anders Sjöberg (Stockholm).
I also organize a seminar series in human reasoning.
I have taught courses in the fields of logic, artificial intelligence, linguistics, programming, and innovation. Currently I supervise one PhD student and several MSc projects.
I have been working with several academic spinoff companies, including the following:
- Safelogic. I started the formal verification company Safelogic in 1999 with the idea of using resource-bounded automatic theorem proving for analyzing and verifying integrated circuit designs. Safelogic was mainly financed by venture capital (about 3 MEURO) and had some 20 employees when it was acquired by Jasper Design Automation of Mountainview, California in 2004. Thus the sales and market channels of Jasper were combined with the product and technology of Safelogic. Jasper's product development is still mainly based in Gothenburg. The customers of Jasper include HP, Cisco, Sony, Oracle, ARM, and AMD. Jasper appears in the 2013 Red Herring Top 100 North America list.
- Optisort. I started the artificial intelligence company Optisort in 2008 with the idea of sorting waste automatically by means of artificial neural networks. The first application is a machine that sorts used batteries into environmental categoriesMachines of this kind were recently delivered to G&P batteries (which sorts about 1/3 of the batteries of Great Britain) and Renova (which sorts about half of the batteries of Sweden). Here is a video of the machine in action. The sorting algorithm is based on artificial neural networks. The company has mainly been customer financed and counted 10 employees at the end of 2012. Optisort appears in the 2013 33-list of NyTeknik and Affärsvärlden.