Human Reading Assessment

Artificial intelligence (AI) is entering schools at full speed, bringing with it both opportunities and challenges. Human Reading Assessment (HUMAN) aims to explore how technology can be applied in a way that places the student in the centre. During their years in school, students have to take a number of reading assessments. These assessments form the basis of HUMAN, which seeks to respond to the challenges of KI on several levels: At the most basic level, HUMAN will develop and investigate algorithms, and explore how these can be applied in human centered reading assessments. The next level involves examining the content of the learning technology. In this case, the content are the texts and assignments found in the assessment. HUMAN will investigate how learning technology can contribute to designing texts and tasks that are both engaging and supportive of learning. In addition, the project aims to investigate how assessment results should be presented, in order for teachers to use them to adapt their instruction. At its most general level, HUMAN investigates what the application of AI in learning technology means for children's experience of the world and their agency in it, as well as its consequences for children's rights. How AI is implemented and presented in schools has major consequences for the individual's ability to act independently later in life. HUMAN is therefore also about examining the role of the education system for the individual, and for society and working life. The knowledge developed in the project will provide important input for future policy-making. Findings from HUMAN will also be directly applicable for school administration on all levels, for teachers and other actors in the education sector. HUMAN will take place in close collaboration with Norwegian schools. It is an interdisciplinary project, involving researchers in literary studies, literacy and reading studies, computer science and law.

Goal: identify how we can develop a HUMAN digital reading assessment that presents students with tasks adapted to their interests and level of reading skill. Objectives: 1. New knowledge on the design of engaging texts, tasks and teacher interfaces that can be used to assess reading comprehension 2. A ground-breaking new methodology for adaptive reading assessment, combining IRT with machine learning techniques such as recommender systems and natural language processing 3. Analytical tools for analysing philosophical and legal implications that are specific to the use of AI in assessment 4. A validated and reusable framework harnessing results from objectives 1 to 3 to enable an engaging, precise, and usable assessment Research questions: 1. What characterizes texts and items that are motivating for children to read and are suitable for valid and efficient assessment of their reading skills? 2. What kind of information do teachers need about their students’ reading skills in order to adapt their instruction? 3. Which methods (from e.g., machine learning, multidimensional scaling and natural language processing) are best suited to match the right text to the right student, with respect to their skills and interest? 4. How can we analyse students’ experience of AI-based digital reading assessment? 5. What are the legal implications of AI-based digital reading assessments? 6. Does the framework established in the project provide a more engaging and precise assessment of reading skills than currently existing assessments? 7. Does the framework established in the project provide the teacher with more actionable information than currently existing assessments? The project is interdisciplinary. It is anchored in educational assessment, reading science, and computer science. In addition, methods from literature, law and philosophy are employed locally.