Two-day workshop on December 7th and 8th at the international conference ‘SMART Animals’ at the University of Amsterdam.
|Date||Start 7 December 2017||End 8 December 2017|
Consider this passage of text:
Wir können nunmehro urtheilen, ob die Thiere eigentliche Begriffe haben. Wenn wir nämlich nicht mit Worten spielen wollen, so ist ein Begriff (man mag darunter notiones oder ideas verstehen,) eine solche Vorstellung eines Dinges, dabey wir uns sowohl unserer eigenen Vorstellung, als des vorgestellten, deutlich bewusst sind. (“We can now judge whether animals have proper concepts. If we do not want to play with words, then a concept (one may take concepts to comprise notiones or ideas) is such a representation of a thing, in which we are distinctly aware of both our own representation and of what is represented”) (H. S. Reimarus, Allgemeine Betrachtungen über die Triebe der Thiere, 1762, p. 35)
Can a machine extract the knowledge required to understand properly this passage of text? If so, how exactly? The passage comes from a treatise of the 18th century philosopher Hermann Samuel Reimarus, “General observations on the drives of animals”. The passage is in fact a transcript from a picture from Reimarus’ book written in an old script, Blackletter Gothic; it’s in German, and it uses words that, although they seem rather common German words, such as ‘Begriffe’ (concepts), ‘urtheilen’ (judging), and ‘Vorstellung’ (representation), they are instead highly technical terms in the history of philosophy. Reimarus supposes here a particular model of animal cognition, that is, he supposes that human perceptual experience is conceptually structured whereas animal experience is not, because animals lack concepts; he also presupposes that awareness can be clear and distinct. This model of animal cognition was typical of 18th century (animal) psychology, and was later rejected by philosophers who did attribute concepts to animals, and who adopted a different notion of what a concept is. Hence, the notion of concept was drifting during the 18th century.
Only highly trained and multiply skilled philosophers can extract this kind of knowledge properly from a text like this one: machines won’t. What would be needed for machines to approximate how we read and comprehend conceptually difficult historical texts? Reading like a human will be devoted to this question, and to the state of the art of technologies in historical natural language processing and philosophical concept drift.
Astrid van Aggelen (CWI Amsterdam)・Hein van den Berg & Arianna Betti (ILLC, UvA Amsterdam)・Antske Fokkens (VU Amsterdam)・Aurélie Herbelot (Trento/Barcelona・Laura Hollink (CWI Amsterdam)・Michael Piotrowski (Lausanne)・Martin Reynaert (Tilburg) Stefan Schlobach (VU Amsterdam)・Caroline Sporleder (Göttingen)・Nina Tahmasebi (Gothenburg) ・Kalliopi Zervanou (Utrecht)