STUD: Disputas - A platform for analysis and critical reflection.

Machines are catching up on us. The past several years has seen significant advances in the development of artificial intelligence. That is, pattern-recognizing computers that match, or outperform human capabilities. The method driving the development of this exciting technology is incredibly flexible, and powerful. It is called machine learning. With large datasets, smart algorithms, and strong computers, we may train models that are able to recognize all patterns we can perceive ourselves, in addition to abstract patterns that are too complex for human biology to detect. Private and public research centres as well as universities and innovation clusters around the world are now working hard to train computer programs that can see and think like humans. The algorithms are available, cloud-computing has made compute readily available, and fairly cheap. The remaining bottleneck for advancing the frontiers of artificial intelligence is data. Progress has been rapid in fields where attempts have been made to train computer programs to make simple, intuitive judgments, such as ?I like this song, (or this post, this video, etc.)?, ?the object in this picture is a car? and ?this word is negatively charged, (or an adjective, or it means spade in Norwegian etc.)?. Data of this intuitive type is captured in popular web applications, extracted from publicly available datasets, or can be generated relatively reasonably by manual categorization work. Progress has been slow in fields where it is possible to ascribe wrong labels, where the concepts are complicated, and expertise is required to distinguish between cases that fall within and without. The usual methods of compiling large datasets do not work for these domains. But the data is valuable. Therefore, researchers and developers use expensive solutions to create datasets anyway. A common solution is to hire experts with several years of education to categorize images or annotate the text. Several researchers create datasets in domains for which they have no formal training themselves, even though they have doctorates in their own respective fields. Scientists have long been interested in developing artificial intelligences that can reason by drawing logical inferences. That is, computer programs that can not only retrieve simple facts from the internet, but programs that are able to draw inferences from facts, and see how different statements are connected or exclude each other logically. The field of research dedicated to this end is called ?argument-mining?, because it is largely dedicated to disguishing good arguments from all the other content that occurs in text. The field of research is largely characterized by the situation described above. It is no small matter to distinguish between sound and erroneous inferences, and to see logical connections between statements. While experts analyze argumentative text at research centres and the computer science department, there is another group that spends time and effort doing the same thing not far away. Universities and colleges in all countries of the world offer subjects in logic and critical thinking. The most important component in most subjects of this type is argument analysis, a practice in which the student analyzes a text, and identifies, reconstructs and evaluates the arguments that are given. In Norway, these courses are mandatory for most students, they are called the ex. phil. courses. In analytical disciplines, it has long been common to use models and graphical tools of various kinds to visualize abstract patterns in numerical data or text. This is also the case for courses in logic and critical thinking. The syllabus literature in these subjects provides many useful pieces of advice, and examples, which make it easier for students to understand difficult concepts and abstract thinking. These pedagogical tools are well-suited for a digital format. In statistics, computer science and many other analytical subjects, computer programs has long since aided students in their learning efforts. However, courses in logic and critical thinking does not benefit from the same range of digital services, despite the fact that the Norwegian ministry of Education and Research has a stated goal that the country?s educational institutions should offer ?simple, effective and reliable digital pedagogical services that renew, simplify and improve the teaching offer?. With funds from the Research Council?s STUD-ENT scheme and academic support from UiO and UiB, we have developed a platform for reading and writing argumentative text that meets the specific needs of courses on logic and critical thinking. Furthermore, we aim to develop the platform in such a way that use generates data that is fitted for training a new line of artificial intelligence that may reason by drawing inferences, and seeing logical interrelations between statements.

Our aim with this project is to make logic and critical thinking accessible to everyone through engaging educational technologies and language technologies. The educational platform is now ready for use, and is already a competitive alternative to other text analysis services. We expect the use of the platform to grow in the time to come, and will expand to new subject areas. If all goes well, the educational offer for text intensive courses in Norway and all English speaking countries will have a digital pedagogical alternative for active learning. Also, the efforts of teachers and students in the educational sector using the platform will give rise to linguistic data that can be used in R&D for language technologies. The platform thus constitutes a research infrastructure as well as a modern pedagogical offering for active learning.